x
Routing

Alcatel Router Revenues Surge

(NYSE: ALA; Paris: CGEP:PA) has some numbers to back up the success it's claiming in edge routing, as the company plans to announce today that its market share surged in the third quarter.

Figures from Synergy Research Group Inc. show Alcatel's 7750 and 7450 models collected $88.8 million in revenues during the third quarter, up from roughly $35 million in the second quarter.

Yes, sales more than doubled in three months.

"That shocked a lot of people," says Ray Mota, the Synergy analyst behind the report. Mota queried some major carriers, though, and he says they're backing up the numbers.

"It's not a spike," asserts Basil Alwan, president of Alcatel's IP division. "This was the decisive quarter to the point where we are a head-on challenger" to (Nasdaq: CSCO) and (Nasdaq: JNPR), he says.

What has Alcatel particularly jazzed is that it took second place in Synergy's "IP edge aggregation routing" category, with 23.6 percent market share in the third quarter -- surpassing Juniper's 19.7 percent but still trailing Cisco's 45.9 percent. (See Alcatel Seizes #2 Position.) Alcatel's IP edge routing market share stood at just 3.1 percent after the first quarter of 2005.

But Juniper notes some extenuating circumstances. In the third quarter, Juniper's M7i and M10i routers were taken out of the service provider category and into the high-end enterprise category. So Alcatel's bump in revenues was accompanied by a decline in what Juniper was reporting: "We moved a substantial amount of revenue out of that category," a Juniper spokeswoman says.

She notes that Synergy still ranks Juniper second in all service provider edge routing, a superset of the IP edge category. Service provider edge routing in the third quarter was led by Cisco with a 48 percent share, followed by Juniper's 27 percent and Alcatel at roughly 14 percent, she says.

Note, also, that the numbers can be sliced up any number of ways. Alcatel's 7450 is an Ethernet box lacking full IP routing functionality. So some might argue it's not suitable for any "IP" category, although it's often sold in tandem with the 7750.

Even with such caveats, the numbers suggest Alcatel's IP division, launched after the TiMetra acquisition, has kicked into gear. Alwan claims the 7750 and 7450 have racked up 90 customers, 50 of them announced, with wins including (NYSE: BT; London: BTA), (NYSE: CHA), and (NYSE: SBC). (See Alcatel Picked for BT's 21CN, Alcatel Wins China Telecom Deal, and Scaling IPTV: Progress at SBC .)

Alcatel and Redback Networks Inc. (Nasdaq: RBAK) both appear to be on the rise in the broadband edge, says Heavy Reading analyst Rick Thompson. (See How Redback Won BellSouth.) Alcatel's port density and the integration of policy management have helped it in the video market, in particular, he notes.

Alcatel's recent strength in IPTV wins might have been a factor in Cisco's decision to acquire Scientific-Atlanta Inc. (NYSE: SFA), a move that could boost Cisco's prospects in service-provider video. (See Sci-Atlanta: Cisco's IPTV Lifeline?.) But Alcatel officials note that triple play wins account for only half their router revenues, implying the 7750 has proven attractive in normal routing cases as well.

Alcatel may have to work hard to maintain its presence in the IP edge. "That space is extremely challenging," Mota says. "They have to stay innovative, like Juniper in its earlier days."

Alcatel is trying. The company has been adding software features to the platforms and has increased Layer 2 support on the 7750. (See Alcatel Adds to MSE , Alcatel Enhances IP Tech, and Alcatel Taps Layer 2.)

— Craig Matsumoto, Senior Editor, Light Reading

materialgirl 12/5/2012 | 2:51:52 AM
re: Alcatel Router Revenues Surge Here is a shot from outer space on market segmentation. Service providers want a cheap, fast, network that is simple to manage. This implies using switches. However, fixed routes get hard to manage as they must be changed manually over time as traffic patterns change. Automatically flexible routes, via routers, are easy to manage, but expensive in that the processing required in the box gets costly.

So, what they want is the "right" mix of cheap forwarding, with easy-to-manage routes. By definition, this seems like a hybrid beast.
konafella 12/5/2012 | 2:51:53 AM
re: Alcatel Router Revenues Surge So I would summarize this thread with two conclusions:

1) everyone has their own definition of a router vs a switch, and GMPLS and MPLS L2 PE capabilities clearly confuse the issues further.

2) most agree that 7450 deserves classification as a router

So what about the Alcatel 7250? Is it also a router?

kf
douggreen 12/5/2012 | 2:51:58 AM
re: Alcatel Router Revenues Surge Mark,

So, how about a crack at a reasonable set of market segments for routers and switches defined by application (network type, SP business model, network location)? To be useful, it has to be granular enough, but not contain too many categories.

I could do it for transport equipment, but you are much better suited to do it for routers. Care to give it a try?
Mark Seery 12/5/2012 | 2:51:59 AM
re: Alcatel Router Revenues Surge Doug,

please see my previous comments on segmenting by application.
douggreen 12/5/2012 | 2:51:59 AM
re: Alcatel Router Revenues Surge Mark,

As I have argued many times before, while segmenting markets by technolgy can be a useful market segmentations basically skip the one segmentation that makes the most sense: Segmentation by application. These are products that actually compete for each other for a certain type of business, regardless of their underlying technology.

I first examined this in an article in XChange magazine 5 years ago (when the market was more complex, with BLECs, CLECs, etc.). If you are really bored, here is the link...

http://www.xchangemag.com/arti...

This is not usually done because, although it is a lot more useful, it is hard to do and requires a knowledge of the various carrier network architectures, and even their business models. In my experience, this is just too much to ask of most of the people who do market analysis.

So, I really don't care much about the switch/router debate unless it is put into the context of a specific application space.

Mark Seery 12/5/2012 | 2:52:00 AM
re: Alcatel Router Revenues Surge my last thoughts on this subject............

extensions vs definitions:

802.1q is not the definition of what an ethernet switch is, it is an extension to what the definition of what an ethernet switch is.

likewise, VPLS is not the definition of what an IP router is, it is an extension to what the definitin of what a router is.

just like you could implement 802.1q on a non Ethernet switched network, you could implement VPLS on something that is not a IPv4 router (and that has been done).

however, at the end of the day, the clear intent for MPLS was that it be an extension to what the definition of what a router is, and the clear intent of the VPLS working group is that a PE is "commonly an edge router".

that VPLS implements a layer 2 switching function (by some people's definition of switching) does not make the platform any less of a router. it also does not make the platform an Ethernet switch. To the best of my knowledge, the IEEE has never been particularly comfortable with what essentially amounts to a half-bridge implementation, and technically virtualizing the bridge function over a network like this, does not meet the IEEE definition of what a bridge is.

whether the underlying router that supports a VPLS meets the needs of any buyer, is a decision for the buyer to make and has nothing to do with whether something is technically a router or not.
Mark Seery 12/5/2012 | 2:52:00 AM
re: Alcatel Router Revenues Surge TMC1,

The use of global vs local addresses as a distinction between routing vs switching is a derivitive of the difference between connectionless and connection-oriented, i.e. in a truly connectionless network you can take a packet out the network and inject it back anywhere else in the network and it will find its way to the destination. this is one defintion of the difference between routing and switching, though as you say clearly not shared by every one here.........
tmc1 12/5/2012 | 2:52:00 AM
re: Alcatel Router Revenues Surge Desi,

You touch on something i was going to mention that i learned in my career and helps to keep routing vs. switching clear. i did not mention because it seems that most of the people here would confuse the issure but global vs. local addressing has always been a key difference between routing and switching as i have learned it.

i was taught/learned that when you are just looking up and modifying local L2 headers in the cell/frame then it is switching. when you are required to do a lookup against a globally unique address database then it is routing. FR DLCIs and ATM PVCs are only unique locally, MAC addrs are more complicated but could be construed this way because different VLANs can have the exact same MAC addrs with no problem as can multiple broadcast domains within the same network so they are only unique to broadcast domain. MPLS labels are another example of this - only locally unique to the node.

Anything where routing is involved is globally unique such as IP addrs, IPX addrs. If two devices exist globally with the same # it is a problem. Remember non-routable protocols were those without unique L3 addressing such as SNA and Netbios and they had to be bridged everywhere. This along with the design and architecture has always allowed me to keep routing and switching pretty well separated but an mpls router can do both just like an L3 switch can do both. These products do seem to blur the boundry but i have no problem calling one a router than can switch mpls and the other an L2 switch that can do some routing.
Mark Seery 12/5/2012 | 2:52:00 AM
re: Alcatel Router Revenues Surge desi,

1) anyone has the right to say that because a router does not support multicast, it does not meet there requirements as a router, it is of course still a router.

2) you suggested there was such a thing as "IP routing protocols" and that any product that uses them is an IPv4 router, which is clearly not true. the fact that you don't care that a control plane uses a data plane (and this has nothing to do with inline) suggests you don't fully appreciate how irrelevant the control plane protocol is to the data plane - specifically in the case of MPLS where an additional control plane protocol has to be used to make it work (or extensions to a protocol in the case of BGP4 label distribution).

3) some forms of LDP-based MPLS looks very much like hop-by-hop forwarding, there is not a lot of confusion about that as far as I know.

A product that does not have a IPv4 data plane and uses GMPLS to establish wavelengths is not a IPv4 router (calling it a router is throwing a blanket over something to try and make a point). Likewise, a product that uses an IP+MPLS data plane to set up LSPs is not necessarily a IPv4 router (though usually it is) - but you should not from a technical perspective be insisting that it is. There are examples on the market where it is not.

If PNNI SVCs are routing then so are PVCs. Please consider the difference between a management plane, a control plane, and a data plane. You are referring to the difference between something that is dynamic and something that is static (by a subjective view of what is dynamic - a management plane could be used to make something dynamic, though that is probably not a great idea). As discussed widely in this thread, most view the diffrence between switching and routing in a different way than what you have suggested (even if I would tend to disagree with these views).

I would agree there are many interpretations of the word routing, and as an example some of the other protocols you lists were considered by some not to be routable.

flyingsausage 12/5/2012 | 2:52:01 AM
re: Alcatel Router Revenues Surge whatever terminologh & regilious arguments we may discuss for years,
the 7450 is anyway much better than a router, or a simple Ethernet switch, just play with it in your lab and you'll understand.
desiEngineer 12/5/2012 | 2:52:01 AM
re: Alcatel Router Revenues Surge Mark,

That's a heckuva long post. I'll try and dissect it.

MS: 1) you are suggesting (intentionally or not) that because the IGP development has been done for MPLS then the hard work of building a IPv4 router is complete.

Yes. In a sense, I question that a box has to run BGP to be a router. Or PIM (as one earlier post suggested, a router must support multicast routing).

Take classical MPLS. After the hard work of digesting an IGP is completed, that forwarding state is captured in a next-hop label forwarding entry for a FEC. That's just geek for a bunch of routes map to a next-hop, and here's my shorthand for that route: it's a label.

MS: 2) that the control plane is tightly coupled with the data plane.

I don't care that a control plane uses the dataplane. That only means protocols run inline with data. That doesn't define routing.

The control plane drives the data plane actions.

I don't quite know what to do with this point. It doesn't refute or relate to anything I said.

MS: 3) that the scenario you describe is analagous to what you wish to contrast it with (i.e. what is good for the goose is good for the gander)

So we come to the nub of the question. MPLS, and in particular, the various applications of MPLS, cause a bit of a problem. They don't look like hop-by-hop forwarding, the way IP does things, with RSVP being semi-connection oriented, and PWs creating these strange edge boxes that use routing protocols to do switching functions.

I think GMPLS LSRs are routers. I think looking at the control plane information of an aggregate network topology, and then determining the forwarding path of traffic based on that information is routing.

Ethernet switches run multi-node control plane protocols, and aggregate that knowledge into a topological structure, a spanning tree. But they forward based on local knowledge of incoming and outgoing ports, VLAN contexts, and MAC address learning. That's very different.

Here's another example: PNNI based SVCs vs. PVC = routing vs. switching.

If you wire up a cross-connect with local information, that's switching. If you wire it up with some aggregation of information derived from distribution of topological information via a control protocol, that's routing.

That's my perspective.

BTW, other defintions of routing exist: in laying down streets, wiring, components; in other protocols, e.g., connection-oriented SNA, XNS, DECnet. I think they fall closer to my generalized definition of a router.

-desi
lightwait 12/5/2012 | 2:52:02 AM
re: Alcatel Router Revenues Surge Well, if area codes should be used for IP
let's have local number portability for IP!
Imagine moving your IP address from Sprint to SBC.
(was that thud a BGP guy fainting?)
...all in jest.

Whatever you want to call them, the 7x50's are selling.
Mark Seery 12/5/2012 | 2:52:02 AM
re: Alcatel Router Revenues Surge >> Forget about what you call the edge box. Why do you call the core box a Router?

(a) Because it could do BGP?
(b) Because it is a cisco or juniper box?
(c) Because it is completely dependent on IP routing protocols to do what it does?
(d) You fill this in so it isn't like I rigged this question to have only those 3 answers.

Personally, I think the answer is (c).<<

Desi,

When you select (c) you touch on three different issues all at once.

1) you are suggesting (intentionally or not) that because the IGP development has been done for MPLS then the hard work of building a IPv4 router is complete

We could argue about the nuances of that for the next 10 years, so I will make things simple and just concede the point (for certain applications).

2) that the control plane is tightly coupled with the data plane.

this would probably take another 50 messages to thrash out, but let me just say I don't believe the control plane has anything "technically" to do with the data plane, other than it has to produce information that supports dataplane actions.
most people don't even recognize that a control plane uses a data plane, this is how separate they are (with the exception of design assumptions that are made, that perhaps should not be - and are generally found not to be irreversible)

please note for example:

http://www.postel.org/rbridge/...
http://www.ietf.org/rfc/rfc119...

also note discussions on the CCAMP WG concluding that GMPLS can support an out of band control plane, which by definition infers the data plane used by the control plane need not necessarily be IP-based (assumptions by control plane implementers not withstanding), and that the control plane could itself be any protocol (observing which you might ask why did it take so long for this to be acknowledged if "techies" never "fuzzy" issues as someone else claimed).

3) that the scenario you describe is analagous to what you wish to contrast it with (i.e. what is good for the goose is good for the gander)

technically what you describe is an (MPLS) label switch (by common usage of the term switch, and by the IETF's inference by selecting the term "LSR"). this comes more from an examination of the dataplane than of the control plane (IMO). Though of course the control plane constructs the forwarding table for the data plane - it is the semantics of the address in the label that would lead most people to conclude that this is not a connectionless address format (with most common usages of MPLS) and therefore, in the context of WAN applications, most likely to lead to the assessment that this is a switching function. also by common definition of a switch operating at either layer 2, or not layer 3 (where "layer 3" is considered to be IP - remembering that the OSI model is most applicable to end stations, and not very robust when it comes to describing what happens below "layer 3" at intermediate nodes). So technically, if such products only ever switched MPLS labelled packets, the case would be strong for putting those products you refer to into a segment called something like "core MPLS switch" (and who knows, that may indeed happen one day - and you would be wrong if you thought I had not had that conversation already)

but my answer to your question is d) (note I am not suggesting you rigged it).

when Juniper entered the market with the M40 it was used in a very limited way. if an analyst had observed that, and therefore put it in a different category, some number of months later that analyst would have looked like an idiot (remembering many analyst work on a 12 month research agenda as a courtesy to their clients - though it probably does not feel that way at the moment!). the juniper M40 market entry is analagous to the CRS-1 market entry because the CRS-1 is a new software and hardware technology base

in both cases we can examine the positioning of the products (application and pricing), and the intent of the companies selling those products (to add or remove function), and what the customer believes they are buying and what they are buying it for (not only today, but sometime in the future). Based on this, we can make an educated guess about where that product should be positioned. we can also ask what is the majority case. BGP4 free core is happening more. is it the majority case over the next 12 months? will an overwhelming number of packets that are processed by a core router be MPLS labelled over the next 12 months? are these products used by a broad range of ISPs and carriers? are these products used for other things / what is the expectation of the customer in terms of being able to migrate it to the edge in the future? what about carriers and ISPs that need to do IPv6 routing - what platforms are they using? and lastly of course is the capabilities of the product - which is of course your point.

the situation you are referring to is a company coming to market with a feature deficient product and then enhancing it over time. this is not an analagous situation to the one you wish to contrast it with. But I won't elaborate any further because I did not get on this message board to bash anyone, but to simply have a conversation about what a switch is and what router is. Given the contrasted scenario, the market place has some questions and is asking them. This is healthy, normal, unavoidable, and necessary. That said, this conversation has been occuring behind the scenes for a number of months, and I think most people (that are directly effected) have already moved on, with the exception of those still scratching their heads about what are the right "terms" to use are - which while technically challenging (if you want to put it in terms of this is a router or a switch) is not so challenging if you want to put it in terms of application usage and focus. the bottom line is that carriers have a certain amount of money to spend, and xyz company is getting so much, and abc is getting something else - this truly is the bottom line. everything else we do is just a tool (one of many - but perhaps the most visible) to try and create insights in to what is happening in the market place. from time to time, those tools need to be refined.

feel free to contact me offline if you want to discuss those things I am not willing to elaborate on here. i have a feeling from the question you asked you know how to contact me, or know someone who does.

p.s. this thread has been mostly about "terms". I have argued that such terms need to be analyzed from multiple perspectives. but if you are interested in using the term "router", you might want to check with the body that popularized the term, with how they would name various devices: http://www.ietf.org/rfc/rfc402... . these products *are* LSRs and LERs, the question being debated is are they more LS than IP_Rs, and more PE-S than PE-R (in terms of how they are actually being used in networks). I think the market knows what the answer is today - and even if the answer is PE-S, then you have to strategically consider the platform that it is being delivered on. How such platforms will be used in the future, is by definition unknowable - but based on other tools/research, you might for example feel you can make an educated guess if you are looking out 12+ months - as an example (starting of course by an examination of the totallity of the ways they are being used today). Only time will tell if that guess is right or wrong. While you are letting that "technology" guess ride, you can have a different conversation about application usage.
paolo.franzoi 12/5/2012 | 2:52:07 AM
re: Alcatel Router Revenues Surge 3. Since the internet has grown organically from many competing providers, the topology is hierarchical on a per provider basis, not on a political/national/geographical basis. (how many networks cross the atlantic ocean _in parallel_ between NYC & London? Many. Are single networks which serve multiple central european nations hierarchical per nation? I doubt it. Hell, my cell phone is topologically 2000 miles from my laptop even though they are on the same desk...)

----------------------

That is the reason why. The rest of it is all true for phone systems as well (see Country Codes). It is just that the way the network grew up was ad hoc. It seemed easier to build a system that did not change people's addresses, than it did to build one that did.

What did the phone network do? It prepended numbers to allow country calling and area calling...i.e. it changed its numbering plan. That was an alternative and it was rejected.

I am not saying that one is better or worse. What I am saying is that there is no better or worse. There is, just is. But both networks are equivalent for what they do. The Internet community did not want to impose order. In many ways it could not impose order. So, it is what it is.

However, your comments apply to both networks (gee - need to know how to forward information, check, need to have redundancy - check, multiple parallel networks - check, networks with dubious topologies - check). Both schemes have worked and proven successful.

If you read my messages at the start of this, my claim was that these schemes were NOT the cause of complexity. Multi-layered QoS is the cause. Then, we got religious zealots who want to claim that the phone network does not work or is one thing or the other.

seven
spelurker 12/5/2012 | 2:52:08 AM
re: Alcatel Router Revenues Surge "'You can make addressing match topology or make the topology match the addressing.' -- Yakov"

That's true in a connectionless network. But isn't it the case that the reason the PSTN is able to cope with so many exceptions to strict hierarchical addressing (e.g. local number portability) is because the AIN can be do a lookup on a per-call basis? Doing that for every IP packet gets rather messy (we don't want a DFZ consisting of hundreds of millions of /32s).
---------------------------------------------------
There's a few problems that are faced in the internet:
1. Because lookups are done on a per-packet basis, all routers must know a summary of all addresses
2. Because redundancy is performed at L3, and links (esp. on a worldwide scale) can go up & down fairly often, that huge database has to be continuously synchronized
3. Since the internet has grown organically from many competing providers, the topology is hierarchical on a per provider basis, not on a political/national/geographical basis. (how many networks cross the atlantic ocean _in parallel_ between NYC & London? Many. Are single networks which serve multiple central european nations hierarchical per nation? I doubt it. Hell, my cell phone is topologically 2000 miles from my laptop even though they are on the same desk...)

Based on these, there's really no way for the internet to become area-code based, regardless of what the UN may think. This is why IP mobility schemes always seem to involve tunneling protocols, and thats why BGP is still the routing protocol of choice.
Mark Seery 12/5/2012 | 2:52:09 AM
re: Alcatel Router Revenues Surge Desi,

I can say more tonight, have to get some work done today, but quickly:

You are talking about a situation where a product is using IP control plane protocols AND mpls control plane protocols. Not just IP control plane protocols.

Consider the situation where a product is using IP control plane protocols and GMPLS protocols to establish a wavelength or a STS-1.
desiEngineer 12/5/2012 | 2:52:10 AM
re: Alcatel Router Revenues Surge Mark: "-there is no industry consensus on what to call frame/packet forwarding activities based on MPLS/PWE3. it is not an Ethernet switch at that point - for sure (and as I stated before it does not even necessarily faithfully emulate one). Is it an MPLS Switch? Is it an MPLS router? Is it a LER (where ER stands for edge router)?"

Ok, let's take a "real" example. Provider uses a T640 or a CRS-1 (note, everyone calls these Routers with a capital R) in the core of a network that does IGP shortcuts, BGP-free core, PWs, VPLS, what have you.

Basically, all that Router is doing is label switching, and that too, only tunnel label switching.

That edge router is doing all the BGP, PW signaling, etc.

Forget about what you call the edge box. Why do you call the core box a Router?

(a) Because it could do BGP?
(b) Because it is a cisco or juniper box?
(c) Because it is completely dependent on IP routing protocols to do what it does?
(d) You fill this in so it isn't like I rigged this question to have only those 3 answers.

Personally, I think the answer is (c).

-desi
konafella 12/5/2012 | 2:52:11 AM
re: Alcatel Router Revenues Surge So Mr Seery's states essentially that if it routes at L3 it is a router, it just might not be applicable for many service provider applications, etc.

And TMC1 states that if it is "VLAN-centric" then it is basically an ethernet switch with (sometimes questionable) routing capabilities added.

I couldn't agree more with both of you. In fact, in the service provider market for these products, I like to divide them into two distinct categories:
1) Routers
2) Switches that can route

Let's assume were talking heavy iron here, not little 1RU devices.

In the former category, if a product is being accepted by service providers globally, then it will be very capable of running full Internet BGP among other things.

In the latter category, if a product being accepted by service providers globally, it is not necessarily an endorsement of the products routing capability whatsoever; you really have to understand the applications and the routing performance limits. Some can do the job. While others are merely switches with nice data sheets.

kf
Mark Seery 12/5/2012 | 2:52:11 AM
re: Alcatel Router Revenues Surge Konafella,

I agree generally with the spirit of what you are suggesting. Let me "fuzzy" things up a little though:

-there are valid routing applications for which BGP is not a requirement (in all networks).
-running "full internet BGP" could still be a fairly minimalistic view of the requirement depending on how you interpret that statement.
-there is no industry consensus on what to call frame/packet forwarding activities based on MPLS/PWE3. it is not an Ethernet switch at that point - for sure (and as I stated before it does not even necessarily faithfully emulate one). Is it an MPLS Switch? Is it an MPLS router? Is it a LER (where ER stands for edge router)?..........
-what if the switch in question has routing capabilities as good or better than other routers? is it still a switch that can route?

it does come down to how it is being used. but MPLS and PWE3 have created some difficulty in relating old terminology (switch/router) to new capabilities.

"a market is a conversation"..............
giles0 12/5/2012 | 2:52:12 AM
re: Alcatel Router Revenues Surge "'You can make addressing match topology or make the topology match the addressing.' -- Yakov"

That's true in a connectionless network. But isn't it the case that the reason the PSTN is able to cope with so many exceptions to strict hierarchical addressing (e.g. local number portability) is because the AIN can be do a lookup on a per-call basis? Doing that for every IP packet gets rather messy (we don't want a DFZ consisting of hundreds of millions of /32s).

In some ways DNS can be seen as an analogue of the AIN (in the PSTN case the only sort of addresses they had were phone numbers, so it was a natural choice for the AIN to use the same addressing format as the PSTN, whereas with DNS we use human-friendly addresses since our computers typically ship with QWERTY keyboards as well as numeric keypads).

I guess with ENUM this all comes full circle as we put E.164 addresses in DNS...
light-headed 12/5/2012 | 2:52:14 AM
re: Alcatel Router Revenues Surge of course no Ethernet switch has to know anything about VLANs to be am "ethernet switch" (I buy Ethernet switches all the time that know nothing about VLANs)
-----------------------------

Mark,

You are adding ambiguity to something that at a technical level is pretty unambiguous. Those Ethernet Switches you buy are either Hubs (not Switches) or actually "vlan-centric" in which case they only support 1 vlan/broadcast domain which is untagged.

While there are some products that blur the lines many of the products clearly fit one category or another. L3 switches seem to cause the most blurring. I agree with you that the 7450 is absolutely a router by most criteria. (Juniper and Cisco will do their utmost to make this unclear and push it to another category - They don't like competition on their turf)

--------------------------------------------
Which just points to the bigger truth that there is a very small number of things a product has to do to technically past the test of being an "ethernet switch" or an "IPv4 router".

And hence the fight about whether something is a switch or a router is interesting, but ultimately not as important as a discussion of what features are needed for any given application.
-------------------------------------------------

Agreed. However, it is very much in the interest of some to blur the lines to discredit the competition or force their revenue to show up in a different category. This has been a non-issue in the technical community. However, the analyst community with the help of vendor marketing have come up with some very creative differentiations to help sell reports and have made the waters a little murkier in the process. The techies ignore these reports because they are just a business/marketing execise. Those less knowledgeable, get confused by them.
Mark Seery 12/5/2012 | 2:52:14 AM
re: Alcatel Router Revenues Surge Turing,

I agree terms can be confusing, connection-oriented, connectionless, switch, router,.... so while they are helpful when communicating within a context (for example a community) that uses them the same way, in a forum such as this the participants appear to be so diverse that they actually get in the way of communication.

I fully understand why you would say that if something inspects IP packets (and forwards based on the addressing etc) then it is a router, but it is a particular type of router, an IP router (specifically if it conforms to RFC 1812 - perhaps one of the best documents ever written - it is an IPv4 Router). The term "router" could be used generically with respect to other protocols, as can the term switch (and the requiremets of such products would vary accordingly).

But to say something is a switch if it does not (do IP-based forwarding), is simpy to reference an accepted usage, not a provable fact. Also your use of L2 and L3 hints at a context in which you operate - not a critcism, just pointing to a frame of reference from which you are viewing the problem (for example, when talking about products that support Ethernet over PWE3 over MPLS over Ethernet or even IP over MPLS (or IP+MPLS if you prefer) over Ethernet the issue of talking about L2 and L3 is somewhat more complex. Likewise when TMC1 talks about something being VLAN centric, he is referencing a view of what in his/her opinion the market has accepted as being a switch - of course no Ethernet switch has to know anything about VLANs to be am "ethernet switch" (I buy Ethernet switches all the time that know nothing about VLANs) on the other hand some people would not buy an "ethernet switch" for some applications if it does not. Which just points to the bigger truth that there is a very small number of things a product has to do to technically past the test of being an "ethernet switch" or an "IPv4 router". There is a much larger number of things a product has to do to pass the buyers test of whether they would use any given product for any given application, and that test will vary depending on the application. And hence the fight about whether something is a switch or a router is interesting, but ultimately not as important as a discussion of what features are needed for any given application. it is obviously also important to understand of all the things a product technically supports, what of those many options are being used in production networks.

So to talk in terms I hope you can relate to, many IP/MPLS routers built for carrier networks are not "Ethernet switches", even when they implement VPLS. While VPLS emulates an Ethernet network (over another network), these products do not have the same default behavior, for example since they support so many different modes of operation, you have to select one on a port, and potentially do a great deal more configuration, before anything gets forwarded, whereas with a simple Ethernet switch you just plug things in and the devices find each other and start talking. Additionally, in some cases they do not perform a relay function (an option on some implementations) so they do not pass the IEEE test; and that does not even start to address various disagreements about filtering etc.

Also with respect to the term "switch" the industry struggles to understand what to call modes of operation based on MPLS label switching. Some will argue it is switching, some will say it is IP/MPLS routing, some will refer to fairly meaningless terms such as layer 2.5 (which is a reference to being a shim between two other layers - but does not really illuminate the role of that shim). I know this is a wish-washy nuance, but in my opinion, some modes of MPLS operation are very similar to IP routing, and some modes of MPLS operation are similar to ATM switching, so you just can't say a product is a router or a switch based on the fact that it supports "MPLS". You have to drive the industry to consensus (not there yet) or you have to be more specific about the mode of operation you are referencing.

So just to make things even more confusing some products on the market support ATM-based forwarding and control planes, MPLS-based forwarding and control plane, IP-based forwarding and control plane, and arguably PWE3-based forwarding and control plane. So there may be some products on the market you can look at and confidently say that is a "switch" or that is a "router", but there are others you can not. Most of these other types of products are made for large operators. But even in the enterprise market it is not always cut and dry, especially if you consider that even though a "switch" for a given vendor might have a worse "router" functionality than other products that same vendor makes, it still might have better "router" functionality than other products on the market, and more to the point the "router" capabilities might be perfectly adequate for some tasks.
Mark Seery 12/5/2012 | 2:52:14 AM
re: Alcatel Router Revenues Surge light-headed,

products that learn thousands of MAC addresses and forward based on a learning process are not hubs - by definition. if you want to suggest that they are internally a 1 VLAN broadcast domain, fine. the point is, "Ethernet switches" come in many shapes and sizes.

every segment of this industry fuzzy things up for self interest, famously even the "techies" during the bubble. analysts are no exception of course. however, the inference of your statements is you can never learn new insights by looking at a market a different way, is IMO, false. and demonstrably false. because even a single operator buys different routers, switches, ADMs, SBCs, WDM terminals, for different purposes. that there is a constant desire to find a way to differentiate is a given. to suggest that analysts never have to change the way they look at a market due to changing conditions is false. you have to look at every situation and judge it on its merits.
desiEngineer 12/5/2012 | 2:52:15 AM
re: Alcatel Router Revenues Surge BGP vs. the phone system. Boring.

Juniper #2 vs. Alcatel #2. That's a lot more relevant and interesting. Can we dice the IP Edge in a more significant way so that Avici can come in #2?

Is there something we can do to make BATM #1?

-desi

Ps. There was this one article I read that said Juniper said they wuz still #2. It made Juniper sound so defensive. Hilarious.
Sisyphus 12/5/2012 | 2:52:15 AM
re: Alcatel Router Revenues Surge
Seven -

you paint a picture of perfection when it comes to the telephone network numbering. The fact is that if you look at its history it had its very painful scaling issues. They had to exentd the numbering range, the area code splits were hell to implement and made numbering schemes less transarent, and mobile networks right now continued the trend.

I would disagree that IP Addressing and phone numbering set out to solve the same problem. IP was *not* desinged to address worldwide, universal connectivity - I don't think that was the initial vision, but of course I am one of the few that didn't invent the Internet! :-) But one should point out the phone numbering scheme started off as a static thing and you got your number and phone assigned whenever it pleased your operator. And why is North America "1" and, for example, Spain "34", why is that so natural? I think it's through force of habit.

Granted IP Addressing is less intuitive, because its hierachy seems less transparent for most, and it started off somewhat compromised space-wise. And while IP6 may provide some more transparency, I don't expect it to be *that* easy to be, either. Anf honestly, as long as devices get an address I don't care all that much what it looks like if I can abstract it and ideally call it by a true name, something the phone network doesn't do. But this is all a bit of a philosophical discussion.

That said, MAC addresses should be even worse as an addressing scheme for a public infrastructure, no?
konafella 12/5/2012 | 2:52:16 AM
re: Alcatel Router Revenues Surge Seven, a couple of points to consider. You certainly have the proof in the pudding (and not in the eating, YR ...) and nobody can argue that.

But the complexity of our NANP has to be much greater now than it was 20 years ago with the impact of LNP (and all the complex SCP call intervention stuff that goes on behind the scenes). I would think that the administration of LNP must be eroding the "simplicity" advantage of PSTN's NANP (and global equivalents) over BGP, no?

And with all the access-independant VoIP services being offered now, the area code and NNX (uh, or was that NXX?) are very much decoupled from the lcoation of the user - they really only indicate the location of the PSTN gateway. So your location based argument is less compelling than it might otherwise have been.

The mobility aspect certainly complicates matters. Will make for some interesting dialogue, no doubt.

kf
paolo.franzoi 12/5/2012 | 2:52:17 AM
re: Alcatel Router Revenues Surge
The IETF may lose the argument to the UN. And recall Judge Green did not put in place country numbering which impacts the International community.

My point is simply this. It is a choice in evolution. The fact that the Internet evolved as an ad hoc institution makes it have certain postives and negatives. But the reality is it is solving the exact same problem that was solved by the phone network. It solved this problem with the constraint that their needs to be no central function that will make the network inoperable when part of it is blown up in a nuclear blast.

Now, people can defend history and how we got there from here. My point stands: BGP is more complicated than country codes and area codes. I disagree that it would make routing more complicated and less scalable. The phone network, which includes every cell phone has existed before and at greater scale than the Internet. So, I have an existance proof of what I am saying.

People may not like the existance proof. They can argue lots of things. But at the end of the day, phone calls are made.

Unfortunately, there is a lot of religion here. Lots of rebellion against something that was already in existance. And rightly so. But one can not deny the working nature of it.

seven
Tony Li 12/5/2012 | 2:52:19 AM
re: Alcatel Router Revenues Surge
seven,

Yes, it is in the phone network. The phone network is not a good analogy for the Internet from a routing perspective. The phone network, for example, was developed in a monopoly situation where the topology was well controlled. Subsequently, Judge Green put regulation in place to ensure that the topology of the network conformed to the original addressing plan.

The Internet is not so controlled and without regulation you cannot constrain the topology to match the addressing imposed by the area code. Thus, the routing information would have to expand to carry exception information that would effectively negate any scalability benefit from the area code.

Thus, as I said before, area codes would make BGP simpler, but they would actually make the Internet routing architecture more complicated and less scalable.

This topic has been endlessly debated in the IETF and the rough consensus from the routing area has been quite clear.

'You can make addressing match topology or make the topology match the addressing.' -- Yakov

Tony
paolo.franzoi 12/5/2012 | 2:52:20 AM
re: Alcatel Router Revenues Surge
uguess,

See the phone network. It exists and has standards. Done.

seven
uguess 12/5/2012 | 2:52:21 AM
re: Alcatel Router Revenues Surge >My statement (which you agree with apparently) is
>that Area Code routing is simpler than BGP.

Seven,

If you are so sure, why not propose it at the IETF and show us the details?

uguess
paolo.franzoi 12/5/2012 | 2:52:22 AM
re: Alcatel Router Revenues Surge
Mark,

I agree with you completely on what you have said. I personally believe that only 2 models of QoS can be deployed en masse: Constant Bit Rate and Best Effort. Trying to tune networks to deliver intermediate QoS types can be done on the small scale (like for enterprises) but not effectively on the large scale.

I agree that ATM networks were good at laying out long hold connections. They could have been good at short hold connections, but two things intervened:

- The first applications were for long hold connections and these lasted for almost 10 years.

- The short hold applications (or lost cost interface applications) were run on these first products without appropriate software work to make the products play in this market. Think about why an BRAS had to exist at all. It didn't. People just didn't update their products to match the new market needs.

I am trying to differentiate technology issues from product issues. I am not trying to say that Ethernet or IP are bad. Everything has its place. Trying to be a zealot and push a single technology to be the end all solution is just a fool's game.

By the way, sorry for the hyperbole about network technicians. But you should travel to say, Namibia and try to do complex telecom installations there.

seven
paolo.franzoi 12/5/2012 | 2:52:22 AM
re: Alcatel Router Revenues Surge
Tony,

Area codes already worked once and is deployed on a massive scale including wireless (add in country codes as well).

So, your argument does not hold water. It would be a significant paradigm shift, but one makes choices in architecture. Both systems evolved the way they evolved. My statement (which you agree with apparently) is that Area Code routing is simpler than BGP.

seven
chook0 12/5/2012 | 2:52:23 AM
re: Alcatel Router Revenues Surge I love the switch vs router flame. It is as old as Methuselah.

Classically, a switch is anything that forwards something from one port to another based on some control function (be that a cross-connect function or an address in a packet.)

In telephony you have class 4 and class 5 switches, etc.

You have switching functions in SONET cross connects.

A multiport bridge is a kind of switch. (But as people correctly pointed out, we didn't start calling them switches until the smart guys at Kalpana wanted a new name for their cut-through multiport bridge.

The classical 2-port bridge is *not* a switch because there is no switching decision involved.

What we today call a router is a kind of switch.

They are all switches.

Sometime in the mid 80s people started sellng commercial products that switched based on L3 addressing (e.g. Cisco AGS, CGS, IGS, the DECRouter, etc.) where previously we had had fuzzballs and unix boxen, and they called them "Routers" but in fact until then the word "router" was just another word for "switch, and a L3 switch was known as a "gateway".

In other words, all the terminology we have today was given to us by the marketeers. An engineer still has to look at what the device really does and what its capabilities are, unfortunately.

To say the 7740 is not a router is just as asinine as saying that just because it is a router it is in the same class of kit and is suitable for the same mission as a T640 or CRS-1. Neither statement survives even the most cursory scrutiny.

--chook
turing 12/5/2012 | 2:52:24 AM
re: Alcatel Router Revenues Surge when an ethernet switch is doing forwarding based on VLAN ID (I mean using it as a label rather than as a broadcast domain), what do you call that BTW? connection-oriented or not? and hence, when modern products have so many different modes, trying to say they are a switch or a router is placing a very limiting box around the conversation.

I am saying using the terms connection-oriented vs not is the confusing piece, at least for me. Because they're connection-oriented at some layers but not at others. For example 2 pure L2 ethernet switches with a VLAN trunk between them - the VLAN trunk could be defined as connection-oriented from a broadcast perspective.

Like in your example of a box forwarding based on VLAN tag instead of MAC address - it is a switch if it forwards all ethernet packets of VLAN 1 to VLAN 2, or encapsulates them into VLAN 2 tag, because then it is not the "end" of the IP broadcast domain at all, does not look into the ethernet payload, etc. It just happens to map broadcast domain ID 1 and 2 together as a bridge, such that they are one broadcast domain. In fact in some indirect ways the original 802.1Q VLAN spec accomodated this with "shared" VLANs, but that never really became popular.

It is a router for me if it only forwards VLAN 1 to VLAN 2 based on IP addresses, changes the source MAC to be its own, needs to be IP addressed to respond to ARPs as a gateway, etc.

That's not to say boxes aren't hybrids of the 2 - a layer-3 switch is exactly such a device, where the switch and router are logical functions integrated into one system.

To me it's more an issue of IP. If the device needs to understand IP to forward packets, and connects subnets (ie, broadcast domains), it's a router. If it is layer-3 agnostic, it's a switch.
desiEngineer 12/5/2012 | 2:52:26 AM
re: Alcatel Router Revenues Surge I think he's saying one man's dessert topping is another man's floorwax :-)

So if a box does MPLS is it a switch or a router? Why is it that something that does RSVP-TE and LDP, and needs to use ISIS and OSPF (with appropriate TE extensions), that allows the establishment of LSP based on bandwidth and link constraints, that uses CSPF, that supports fast reroute, that does GRE tunnelling, that just so happens to do all these things so that it can carry ethernet frames from left to right, is an ETHERNET SWITCH?

Does that sound like an ethernet switch? Just because it had to be feature complete by supporting VLAN swapping/stripping/insertion, or MAC learning (purely for VPLS), or whatever little ethernet function, doesn't turn it into an ethernet switch, fer cryin' out loud :)

Gee, I wonder if it had carried ATM frames, and happened to do the MFA's PW-to-ATM OAM interworking blah, but didn't do PNNI, would you have called it an ATM switch.

I wonder if it had carried SNA packets or cells or whatever, would it have become an SNA router (or whatever IBM called SNA transporters).

I suppose if the 7450 carried little envelopes with stamps on them in pseudowires, we'd have called it a postman.

It's a router, a new breed of router that doesn't do every IP routing protocol, but the ones that are significant in its corner of the network.

Likewise, the 7600. Not the 6500. That's what I think Mark was trying to say.

-desi
netboarder 12/5/2012 | 2:52:26 AM
re: Alcatel Router Revenues Surge Recent carriersG«÷ RFx are showing that although L2 (Eth/VPLS/VPWS/MPLS/PW) is more and more required for metro aggregation and ethernet services, IP routing, with substantial scalability figures (route table sizes, protocols, perfromance), is still a requirement. I think it reflects carriersG«÷ risk mgt processes, which currently donG«÷t recommend betting solely on L2.

The catch here (for the vendor community) is that carriers are not actually willing to pay more for such functionality, and expect almost L2 pricing (some simply say it, and some disclose the prices they got from Huawei - the latter catches your attention much quicker and stronger ;-)).

What it means for vendors is that they need to have a modular cost structure to be as optimal as possible for having both L2 and L3 capabilities, but with the cost targets of L2G«™ Alcatel smartly focused and cost reduced the 7750 to become a more optimized L2 7450, which proves to be a pretty good classification for now.

Going forward, considering silicon technology, and market requirements, one can just build a G«£7750G«• with the same cost as the G«£7450G«• (perhaps with a small premium), converging the two distinct categories into one. A routing (real router) and switching box can be developed as a single product with some hw and (definitely) sw modularity for optimizing the achievable margins.

light-headed 12/5/2012 | 2:52:26 AM
re: Alcatel Router Revenues Surge so you're saying it's not a dessert topping?

;)
Mark Seery 12/5/2012 | 2:52:26 AM
re: Alcatel Router Revenues Surge Light-Headed,

I view my job as helping people to achieve insights, not to be a dictionary (i.e. the definitive reference for what a term means - I have an opinion to be sure). Therefore, I have concluded that I should be giving my clients insights based both on technology and on application/usage; if you look at these markets through one dimension you are simply not going to understand it (IMO - and in this sense there are some things TMC1 point out that I probably should have agreed with in my reply).

The 7450 *is* a router. The 7450 *is* a MPLS/PWE3 and GRE/PWE3 plaform for transporting Ethernet frames. It is both. Not one or the other.

Now if we look at application/usage I observe that Juniper gets *most* of its revenue from one type of usage, and the 7450 gets *most* of its revenue from another type of usage. Alcatel has the right to claim the 7450 is a router. Juniper has the right to claim that there is more to the comparison than simply saying they both sell routers. We have always, and will always (well for the forseeable future), segment *routers* in to different buckets for good reasons. Probably the better way to deal with this is to stop talking about routers, and start talking about IP/MPLS/PWE3 platforms - though that is a mouthful and resistance to change is strong. Some people like to use the term MSE, I happen to not, but won't get in to that here.
Sisyphus 12/5/2012 | 2:52:26 AM
re: Alcatel Router Revenues Surge The thing is that it doesn't really matter what the 7750/7450 are definition wise, the fact is that the combo is causing some hurt and people in Cisco and Juniper wonder what truck hit them when it comes to several recent broadband aggregation plays, which *used* to be pure edge router plays and now have invited in a new breed of product. *That* is the news, a recent trend in whatever it is defeating traditional edge routing architectures.

Then again, we'll have to see how long-lived that trend is. Maybe Alcatel continues to play it as smart as they have (and I'd love to hear how much of it has been by design and how much by accident and luck). Maybe the scaling limitations of an L2 approach in large SP aggregation networks surface, and the game's up and people go back to the IP incumbents.

Maybe a deciding factor is simply the business approach - the ability to cater to the remaining ATM stalwarts in their language with a system integration proposition and with old school telco contracts (heavy discounts being the cherry on top) that the packet-heads didn't quite deem necessary to keep conquering the universe.

It's very interesting:

(1) Awesome job by Alcatel and the Timetra guys thus far' give them credit for executing at all levels and fully exploiting the chinks in Cisco's and Juniper's IP infrastructure armor.

(2) Then again, some preliminary battles may have been waged, but the war for the aggregation is far from over. Instinctively, I see L2 in the smaller scale deployments, but I think L3 is better suited for the large scale ones ultimately. Will Alcatel tame that beast as well? If so, their blend of telco-tradition with a flexible packet oriented product portfolio should make them formidable going forward...
Mark Seery 12/5/2012 | 2:52:27 AM
re: Alcatel Router Revenues Surge seven,

on the issue of solving all problems with one network, I agree and disagree. In a sense, SONET/SDH solves all problems with one network, though some better than others. it is when you try to provide a high level of optimization for all network problems that you create complexity.

to suggest people who had difficulty reading or writing did most of the provisioning of ATM networks is to neglect to mention that the largest ATM networks were in fact highly automated via the management plane. management planes are very good for long-hold connections, but not so good for short-hold connections and this is where the analogy with the telephone network breaks down. comparing ATM PVCs to the telephone network makes as much sense as comparing the telephone network to SONET/SDH. That all said, I would also like you dispute it was difficult to scale (for long-hold connections).

add good managements systems to a BGP network, and take away a number of the options, and you have a manageable network by people who "had difficulty reading and writing" (though I am reluctant to refer to people in this way).

as to your other post that BGP would kill for the simplicity of area codes, well you might be right, especially given that Tony agrees (so not going to go out on a limb on that one!) but the experience we had with CLECs (addressing/switch code allocations) when they were blooming like mushrooms gives me reason to pause and wonder how would this system have handled a world where ISPs where appearing at an arguably even higher rate. perhaps it is all administration issues, but I am not sure, there might be some hidden complexity here......

as for QoS. if you are willing to make the simplifying assumption that there are only two modes supported by business models: guaranteed and best effort (and you have the right constructs for grooming and aggregating guaranteed services) - not sure QoS is that complicated.
tmc1 12/5/2012 | 2:52:27 AM
re: Alcatel Router Revenues Surge Note Cisco still calls many of its enterprise routers, switches, for no other reason than historical consistency / how other people use the terms.
-----------------------------------

Mark,

With all due respect, having worked with cisco routers and switches for over 9 years i can tell you that this is simply untrue in most cases. Cisco has routers that come from router design groups (1600,2500,2600,7200,7500, GSR, CRS1) that are made for terminating L3 sessions and running protocols. Many of these are descended from the original cisco routers running IOS (AGS, etc.) and have routing in their DNA.

They also have L2 and L3 switches made by switching groups (2948, 4000, 6000, 5500, 6500, 7600) that have descended from the acquisitions of ethernet switch makers kalpana, crescendo, grand junction, et. al. and they have ethernet switching in their DNA. They all ran non-IOS CLIs at one time and they are ALL VLAN-centric in design. They borrow the protocol stack (if they do L3) and other features from the routing group and run a version of IOS for switches.

Cisco is the most guilty of re-using and updating their old architectures instead of starting from scratch so many of these legacy differences continue to exist in new products.

Also, cisco most certainly calls the 2600, 7200, 7500 which are their most successful enterprise routers as "routers" today.
light-headed 12/5/2012 | 2:52:27 AM
re: Alcatel Router Revenues Surge Mark,

I thought that you were some analyst that was supposed to be helping to define these products not make us more confused. If you are that "Mark Seery" could you please enlighten us? Is the 7450 a router, a switch, a floorwax or a dessert topping?

Inquiring minds want to know!
Mark Seery 12/5/2012 | 2:52:27 AM
re: Alcatel Router Revenues Surge TMC1,

if you want to have a view point about the quality of an implementation then you are entitled to feel that way if that is your conclusion. however, the products you refer to are technically routers as well as switches - and that is where using such terms becomes problematic.
turing 12/5/2012 | 2:52:28 AM
re: Alcatel Router Revenues Surge But I would hardly call an ethernet switch connection-oriented.

The distinction of router switch physically is still important and does still mean hardware differences for vendors - especially if you mean ATM switches vs. routers. I find there is also still a big difference in the cost structure between routers and ethernet switches and although some of that is just positioning and markets, some of it is real hardware differences.

I agree with you in principle. But to me when a vendor tells me they're a router it has a very different meaning for a network than a switch, for most technical groups of the organization.
turing 12/5/2012 | 2:52:28 AM
re: Alcatel Router Revenues Surge When these new "all-ethernet" devices were coupled with cut-through forwarding modes, they didn't want to be lumped into the old "bridge" product category.

Actually if I recall right, they started using the term when the started being able to forward from one port to another while simultaneously forwarding between two other ports.

Cut-through I thought was a bit later, no? (cut-through being the ability to start transmitting out the destination port before waiting for the whole receive packet to arrive, instead of store-and-forward)

They also desperately wanted to distance themselves from cpu-based bridges with only a few ports, as they had lots of ports and hardware ASICs to do bridging much much faster. Somehow router vendors never figured out how to make that distinction - its still a router whether it's a Juniper T640 or a Proteon GlobeTrotter.

tmc1 12/5/2012 | 2:52:28 AM
re: Alcatel Router Revenues Surge More flames for the fire:

There is also considerable difference in the HW and SW design also. While an L3 switch can do routing, it is usually lacking in some of the things that are required by a router. It has been leveraged from an enterprise switching environment, is VLAN-centric and cost is a very important design limitation. In many cases an L3 switch lacks the memory sizes, interface support, reliability design, testing and development that are needed for high-end edge or core routing functions. The people developing L3 switches frequently do not understand the requirements of high-end routing or do not have enough dedicated resources working full time on making such a product. They just slap some new "high-feature" NPA/FPGA blades in and try to add MPLS, etc.

I have worked on both successful L3 switching products and routing products and they are like night and day in design and testing because the mission of each is so different.

For this reason I agree that no enterprise switch should be considered a true router although it can do simple routing just fine. The 7450 is a true router under this criteria IMHO because it is leveraged from a good, proven router (the 7750) and not from an enterprise switch.
Mark Seery 12/5/2012 | 2:52:28 AM
re: Alcatel Router Revenues Surge Turing,

So perhaps to refine my comments the term "switch" has been used differently in the enterprise market than the carrier market. In the carrier market the term has been used generally to refer to something that was connection-oriented, whereas in the enterprise market the term was used generally as a marketing term to infer performance and price (perhaps as a result of a hardware-based implementation in the early days). Note Cisco still calls many of its enterprise routers, switches, for no other reason than historical consistency / how other people use the terms.

if the term router and switch mean different things to you, I would suggest that is as a result of your experience with specific implementations. ATM switches use a control plane that is at some level of abstraction essentially the same as a router. That a connection-oriented paradigm (most commonly associated with WAN/carrier switches) is said to preserve packet order etc. (one of the different ways people react to the word switch vs router) is more myth than reality, IMO; and if there is any truth to it, more implementation than protocol.

when an ethernet switch is doing forwarding based on VLAN ID (I mean using it as a label rather than as a broadcast domain), what do you call that BTW? connection-oriented or not? and hence, when modern products have so many different modes, trying to say they are a switch or a router is placing a very limiting box around the conversation.
Mark Seery 12/5/2012 | 2:52:28 AM
re: Alcatel Router Revenues Surge I think each is a fungible word that can be used to mean anything you want. I suspect that the word router is rooted in a focus on the theory at the base of the IP control plane (graph theory/LSDB, distance-vector, path-vector, etc.) and it just happened to be associated by public use with a certain forwarding contruct (connectionless IP). On the other hand, I suspect the word switch is rooted in the observation of physical devices used to mechanically and electrically create connections between input and output ports. Therefore we are more likely to refer to something that is connectionless as being a router and something that is connection-oriented as being a switch. There is also a tendency to refer to a hardware-based forwarding function as being a switch, which is the other part of the move from "bridge" to "switch" in addition to the other valid points that were made previously in the thread (even if some "switches" were implemented in software). IMO, many of the chassis based products in today's market (certainly in the carrier market) are so protocol agnostic at the core of their system architecture, that they can be anything, and any combination of anything (scarily so, in fact there are over 8 different protocol IP/MPLS/PWE3/ATM/Ethernet combinations in the market today). So thinking of something simply as a switch or a router is something we have to relegate to a luxury we once had in a galaxy far, far, away.

as another viewpoint, IMO, the IP world is dominated more by computer scientists/software people, where as the Ethernet/TDM worlds are more dominated by EE/hardware people - and hence you are more likely to hear the word 'switch' in those communities, IMO.

at the end of the day we are all just trying to get a packet in one port and out the other. You say tomato, I say...........connectionless vs connection-oriented is probably the best way to differentiate IMO, but it is not absolute, and I am sure some people would object violently to the concept of an Ethernet router. Personally I like to think in terms of trails/tunnels (or client/server relationships) vs relay functions (where the link-layer information is stripped at every hop - with the exception of repeaters/regenerators). These are more important distinctions IMO, and at the root of the most important insights in to what is needed in architectures.
mtb826 12/5/2012 | 2:52:29 AM
re: Alcatel Router Revenues Surge The fact that the 7450 functions as a PE for Ethernet services; switches VPLS or PW services over GRE, LDP, or RSVP-TE (I believe that is called an LER); support POS as well as Ethernet interfaces, can also switch MPLS (I believe that is called an LSR ); supports all required IGP protocols within the core to function as an LER and LSR; does not maintain a global L2 FIB or support STP except in the context of VPLS; VLANs have only local significance; why would you not call it a router? If (or should we say when?) Juniper or Cisco had such a box you'd call it a router, wouldn't you?

Face it folks, Alcatel is in the game.
turing 12/5/2012 | 2:52:29 AM
re: Alcatel Router Revenues Surge I think routing/switching is no longer objective, but subjective terms.

I still believe the classical definitions:
A switch forwards based on L2 (802 MAC, ATM VPI/VCI, whatever) addresses in an L2 bridge table, and doesn't fragment/reassemble the packets, doesn't change contents of packets, and for all intents and purposes shouldn't care whether it's L2 packets are carrying IP, IPX or Appletalk. In ieee 802's case (e.g., ethernet switches) it's a bit more specific in that it shouldn't change MAC addresses, but can partition based on vlans, could particiapte in spanning tree and its derivatives as a bridge node, etc.

A router does not have to run a single routing protocol, just like switches don't have to run spanning tree. (although in the latter case usually because people don't want it to) A router merely has to forward based on the ip dest address, changes the L2 addresses since L2 is terminated by it, technically should decrement the TTL, could/should do fragmentation/SARing or at least enforce an MTU, etc. For ethernet L2 it would be the broadcast domain edge, and appears as an ethernet LAN endpoint.

An L3 switch in my book is a router, or at least a logical equivalent of a switch where the separate VLANs are "connected" by an internal router engine. It's just that some people think it's an L2 switch even when they make it route between VLANs, and others think it's an L3 switch even when they only have 1 VLAN configured on it or no inter-VLAN forwarding.
Tony Li 12/5/2012 | 2:52:29 AM
re: Alcatel Router Revenues Surge
Actually, BGP with area codes would push us into the realm of geographic addressing. While that would greatly simplify the protocol itself, the regulation necessary to implement the approach would be intractable. The result would be a series of exceptions, which would lead to a combination of both 'area code' and 'longest match' forwarding, which would be a level of complexity far above where we are today.

Tony
paolo.franzoi 12/5/2012 | 2:52:30 AM
re: Alcatel Router Revenues Surge
ATM PVCs were not a requirement of ATM networks. It was a poor implementation and the technology supported SVCs and PNNI. So, I come back to it was the boxes and not the technology. The boxes were designed to support handsful of high value connections and apply zillions of rules to these connections. The technology did not require that to be true. The application of the technology was such to deploy this over high value interfaces of very small numbers.

When ATM was taken to consumer access, this needed to change and the boxes did not change. That is not the technology's fault. Which is actually your claim. If DSLAMs had used PNNI and SVCs to set up new users than there was no need to set up ATM PVCs. ATM PVCs were not a requirement for that application.

I am simply stating that number of point to point connections is the reason that a particular technology failed. I dispute this claim. I point to my example. ATM did not automate its connections. That was a product problem.

I think the voice network is a great analogy...BGP wishes it had the simplicity of area codes.

seven
Sisyphus 12/5/2012 | 2:52:30 AM
re: Alcatel Router Revenues Surge > The statement was that ATM is hard to manage
> because of the number of point to point
> connections. I dispute this point absolutely
> at any layer. ..

But Seven -

ATM PVCs had to be set up manually. SVCs never quite made it, MPLS came too late to save the day and aid the scaling issue. It was indeed an N**2 problem, and had scaling issues. I saw it first hand...

The voice network supports millions of dynamic connections indeed, but those are set up and torn down by users, there is no administrative overhead involved in setting up those connections - just in setting up a static network topology where connections get set up by users as they need.

There is a fundamental difference, and I still think the voice network is not a god analogy, not for L2 Ethnernet nor for ATM... and not for L3, for that matter.
paolo.franzoi 12/5/2012 | 2:52:30 AM
re: Alcatel Router Revenues Surge
Sis,

The statement was that ATM is hard to manage because of the number of point to point connections. I dispute this point absolutely at any layer. I dispute that it was dificult to scale as it was done with many employees who had difficulty reading and writing. Call me when you can manage a BGP router network with such employees.

The complexity of managing these networks is QoS and trying to solve all problems with a single infrastructure.

seven

matahari 12/5/2012 | 2:52:31 AM
re: Alcatel Router Revenues Surge Juniper is claiming the gain is due to M7i and
M10i moving to high-end enterprise. This statement
actually puts Juniper in a defensive.

- If M7i and M10i are considered enterprise, this
means their claim to have gained in enterprise
is not due to the J-series. This would seem
the J-series sales are disappointing.

- What happened to the M320, Juniper's new edge
multiservice box? Another indication the M320
has been a disappointment.

That's two duds out of two "new" products
from their router group.
photon2 12/5/2012 | 2:52:31 AM
re: Alcatel Router Revenues Surge Well since the we've degraded to bridging vs switching, what I've been looking for (for awhile) is the comprehensive difference today between a router and switch. My definition has been the line where WAN services take place, BGP must be enabled beyond 2 peers, and multicast support is available. But I'm sure that's not all of it....any suggestions in a simple checklist without the 'vendor bent'?
konafella 12/5/2012 | 2:52:32 AM
re: Alcatel Router Revenues Surge When bridges evolved into multi-ethernet port devices, the start-up leaders in the segment (Kalpana, etc) had to find a fancy new term for their product so they coined the term "ethernet switch". Prior to that most bridges were bridging ethernet into serial WAN connections. When these new "all-ethernet" devices were coupled with cut-through forwarding modes, they didn't want to be lumped into the old "bridge" product category.

After all, as discussed here, it's easier to get significant market share when you define a brand new category, right?

kf
tmc1 12/5/2012 | 2:52:32 AM
re: Alcatel Router Revenues Surge signit,

multi-port bridging and switching have become synonomous... sorry if that offends your sensibilities. Other L2 technologies (ATM/FR) are also frequently referred to as switching technologies. Don't even get me started with L3 switching which is a really confusing term (remember BRouting?)

and yes, you are splitting hairs.

sigint 12/5/2012 | 2:52:32 AM
re: Alcatel Router Revenues Surge tmc1:
anything that forwards at L2 based on MAC addr is a switch
__________________

Sorry to split hair - but isn't bridging the historically correct word for this function?

Hopefully, correctly formatted this time.
sigint 12/5/2012 | 2:52:33 AM
re: Alcatel Router Revenues Surge tmc1:
anything that forwards at L2 based on MAC addr is a switch
_________________-

Sorry to split hair - but isn't bridging the historically correct word for this function?
light-headed 12/5/2012 | 2:52:33 AM
re: Alcatel Router Revenues Surge "They have to stay innovative, like Juniper in its earlier days."
-----------------------------------------------

The implication being that Juniper is no longer innovative. NO, that's not possible. They wouldn't be behaving more like a certain large company on Tasman now...

;)
light-headed 12/5/2012 | 2:52:33 AM
re: Alcatel Router Revenues Surge Sisyphus,

I really agree with you if you are talking about all the garbage that companies like BT, Nortel and the ITU/IEEE are trying to add to ethernet like Traffic Engineering. Most of it is terrible and needs to be ignored. You can understand why the bellheads want to do it - more complex, turn ethernet into atm because they have already done as much damage to mpls as they could, etc.

Nortel, of course, wants to continue to change the playing field and start over because they missed the boat and cannot compete with other products based on implementation of standard features. Nortel's whole data strategy seems to be just keep trying to move everyone to proprietary solutions that they have a head start developing.

When some technologist/bellhead at a company like BT gives them some attention it just makes it worse for everyone!
Sisyphus 12/5/2012 | 2:52:33 AM
re: Alcatel Router Revenues Surge > So is the phone network unmanageable due to
> all the point to point connections? Of course
> not.

Not really sure where you see the analogy to Eth L2 there. I don't see it at all. Fundamentally different network philosophy with a clear split between signaling and forwarding plane, ergo its simplicity. Circuit switching is L1 all the way through, my friend, *not* a packet technology. Telephony *solved* the scaling issue in an intuitively simple (albeit complex due to scale) way. The more functionality they keep adding to L2 Ethernet stuff to fit service provider aggregation infrastructures, the more it resembles IP with its somewhat awkward forwarding-control overlaps. The huge question is whether all the mechanisms that are being implemented to mitigate the (negative) effects of huge ARP tables are truly going to scale up... there's a reason why IP infrastructure routing protocols are what they are, only they had far more time to mature and stabilize. It is amusing to see the Bellheads are now building up stuff in a major scale with a lot of stuff that stands on questionable standardization ground and proven poor interoperability... but this is my personal opinion, I'd be summarily executed and shot in the neck if I voiced this in a meeting... :-)
paolo.franzoi 12/5/2012 | 2:52:34 AM
re: Alcatel Router Revenues Surge
So is the phone network unmanageable due to all the point to point connections? Of course not.

ATMs failings were the requirements for PhDs to configure the QoS requirements of a complex network. Additionally, the topology management function of PNNI was overly cumbersome. Finally, ATM boxes were never tuned for Access. They simply just pushed core boxes out to the edge (see additionally Cisco routers moving from Core to Edge).

I believe there is room for both Layer 2 and Layer 3 products. Layer 2 is effective in Access and Aggregation. Layer 3 provides a segmentation which allows these Layer 2 areas to be managed in an effective way. Layer 3 folks are also unwilling to sell their products as commodity pieces of hardware.

Imagine what would happen to Cisco if software was free, and software upgrades were free for 10 years. This is the world of access.

seven
Stbl 12/5/2012 | 2:52:34 AM
re: Alcatel Router Revenues Surge
The 7450 is perhaps debatable though it certainly routes with OSPF/ISIS support and it runs Ethernet over MPLS. It supports Etherent and SONET/SDH interfaces.

The 7750 is a full on router with all the interfaces and scalable BGP, Multicast and 2547. And the 7750 is winning some big deals if you read the releases.

Bottom line: ALA has routing and JNPR/CSCO must be feeling the heat. About time.
photon2 12/5/2012 | 2:52:35 AM
re: Alcatel Router Revenues Surge Agree the strategy has been great for Alcatel...frankly it fits "the ATM heads" who liked layer 2...Ethernet is simple and BGP is hard, so these products appeal to the 'average' network manager is not as complex. However, just as the ATM networks were originally considered not complex with pt to pt connections, they grew unwieldy and unmanageable precisely because of all the connections. VPLS and next gen VPLS will not really solve this....My take is that smaller network providers and smaller enterprises will chose Layer 2 products, and larger providers and larger enterprises will stay with Layer 3 products. With some melding in the middle....but it won't be all one or the other.
tmc1 12/5/2012 | 2:52:35 AM
re: Alcatel Router Revenues Surge This really comes down to what your definition of a router is. If you are cisco or juniper then you will use your influence to make sure it is only the products that are sold by cisco or juniper.

By definition under the OSI model, any product that forwards at L3 based on IP addr info is doing routing, anything that forwards at L2 based on MAC addr is a switch. In addition ATM and FR forwarding occurs at L2 and is switching. This also includes L3 switching products that can perform either function and also run routing protocols. Most of these device actually perform routing in enterprise networks where traditionally a cisco 7500/7200 may have been used.

However i agree that a product that is used to do L2 switching and can also perform some routing function is not a router. This also would eliminate cisco's 6509 and 7600 "routers" which are essentially enterprise switches but they sell as routers into enterprise and SPs/RBOCs. For a while the 6509 may have been the best performing/feature rich router that cisco actually sold (circa 2001-2002) which makes it pretty confusing because it is an L3 switch.

Juniper and cisco want to keep the bar as high as possible so they want to define a router as a product that runs all the routing protocols, has channelized interfaces, ATM/FR support, POS, BGP4 and all IGP support, Multicast Routing, etc. This helps them defend their "turf".

The 7450 is not vlan centric, does not actually support VLANs and is not an L3 or L2 switch. Neither is it sold as such. It is architectually the same as the 7750 and has the same chipsets with some cost-reduction elements. It does not support BGP4 and has less interface support than the 7750 but is it a router. It is a router designed for ethernet services for people that do not want/need to pay for a full-fledged MSE-class router.

Instead of cisco and juniper trying to claim it is not a router, maybe they both should develop a product aimed at this specific sector (and not just a re-badged switch i.e. 7600). It seems that the strategy has been very successful for Alacatel... successful enough that cisco just spent $7B to try and compete with Alcatel.

paolo.franzoi 12/5/2012 | 2:52:36 AM
re: Alcatel Router Revenues Surge
Agreed with Photon2. What Alcatel is doing very well in is Metro Ethernet deployments.

seven
Sisyphus 12/5/2012 | 2:52:36 AM
re: Alcatel Router Revenues Surge
Technicalley true, however the 7750/7450 combo is being used to deliver a variety of services via a highly scalable VPLS architecture, and this "L2 VPN" scaling abiity allows it to gnaw away at areas that have been traditional Cisco and Juniper router deployments.

It's still in the stars, though. The current upswing comes through heavy and desperate discounting -by ALA- that accompanies the VPLS value proposition, and perhaps overcomes concerns about its ability to truly scale to large service provider deployments. The L2 pitch is also a smart cultural fit with the ATM Bellhead stalwarts, who have been now in the DSLAM access and thus have a thing or two to say with aggregation deloyments.

This will be quite the fascinating battle to watch... Alcatel is approaching it smartly, but of course it's also their only chance to position themselves in the emerging infrastructure. However, I personaly have some reservations about Ethernet's ability to truly scale to this level. Wasn't it suposed to be about plug&play provisioning simplicity? Looking at VPLS now -labels in label hierachies to support QoS and business services and reduce scalability issues- is sure doesn't look that simple. In fact, L3 stuff looks far more decipherable - in fact it looks like L2 is now re-inventing the whole L3 wheel at a rapid pace right now. Whether it's a good thing or not we'll see, and Alcatel's success depends on how quickl the market comes up with an answer.
photon2 12/5/2012 | 2:52:37 AM
re: Alcatel Router Revenues Surge The 7450 is NOT a router...no BGP, no multicast...if you are not using MPLS it's just a enet switch...it should not be counted in the router category at all. No competition for Juniper in my mind...
mtrehearne 12/5/2012 | 2:52:38 AM
re: Alcatel Router Revenues Surge Does JNPR have a box that competes head to head with the Alcatel 7450?
russ4br 12/5/2012 | 2:52:38 AM
re: Alcatel Router Revenues Surge It's a cause of concern more for JNPR than for CSCO.

JNPR has greatly benefited from a "dual-source" approach by carriers. In certain situations, JNPR was given a piece of business just in order to "keep CSCO honest". Now that ALA has a credible solution in the IP space, they might starting taking business away from JNPR - in the "dual-source" category.

It's curious that JNPR also tries to underplay the ALA gains in marketshare, saying that they moved a "big piece of revenue" to other category. Not surprisingly, JNPR moved the revenue to "high-end enterprise", in order to artificially get "powerpoint market share numbers" to shore up their claims that they are having traction in the Enterprise besides the NetScreen product line.

-russ
digits 12/5/2012 | 2:52:40 AM
re: Alcatel Router Revenues Surge Alcatel says this isn't a blip. Juniper says the niche sector numbers don't tell the whole story. But can Alcatel join Cisco and Juniper to make the leading pack a trio?
photon_tim 12/5/2012 | 2:52:41 AM
re: Alcatel Router Revenues Surge Is Huawei reporting any router numbers?
They appear to be a credible forth contender.

Tom
paulchenzhong 12/5/2012 | 4:08:40 AM
re: Alcatel Router Revenues Surge In short, 7450 is L2 Etherhet switch, 7750 is L3 router.
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE