x
Optical/IP

Chiaro Debuts a Big, Bad Router

It's the spring of 2003 and there's a capital spending slump that shows few signs of roaring back. What kind of startup would think to launch a cutting-edge core router product?

It seems a bit crazy, but that's exactly what Chiaro Networks did this week when it unveiled its Enstara product to mild applause and stifled yawns (see Chiaro Intros IP Routing Platform).

It could be the case of promising technology delivered at an unfortunate time. The technology muscle powering Chiaro's big, fast IP router is impressive. But Chiaro has significant economic hurdles to cross if it wants to survive, much less thrive.

A few weeks ago, Chiaro touted its switching technology, called Optical Phased Array, that helps it switch a large number of ports at nanosecond speeds. This technology, which makes up the guts of the Enstara, is combined with several other technologies to make a system that is covered by some 25 technology patents (and counting).

Chiaro says its router can save carriers money in their core networks because it's so fast and so scaleable that it eliminates the need for an aggregation layer between the network's core and its edge. The company is also touting the system's reliability, saying it can withstand hardware and software failures by switching over to redundant resources without dropping packets (see IP Routing Gets a Restart). This infrastructure allows so-called "hitless" maintenance as well, meaning that you don't have to stop routing in order to tinker with or upgrade the machine.

Such strides are important, says David Newman, president of Network Test. "Service providers like to count uptime by excluding maintenance windows for upgrades. Customers want uptime counted by the clock on the wall. Anything a device can do to enhance uptime is definitely to the good."

Chiaro is also talking up its router's flexibility. The system comes in a variety of shapes and sizes. For small nodes, there's a single chassis, 10-slot version. In bigger nodes, one Chiaro router can comprise several refrigerator-sized boxes, which can be split up over different floors or rooms, the parts connected by optical links. In its largest iteration, the Enstara can scale to up to 315 10-Gbit/s equivalent ports (6.3 Tbit/s of routing capacity in both directions).

So can Chiaro's big, bad router really handle the jobs that would be required of it in a large carrier network? Well, we still don't know for sure. No large carriers have bought the box yet, and the two customers that Chiaro has announced aren't running traditional telecom networks.

Chiaro's first customer is the OptIPuter, a five-year, $13.5 million grid computing project funded by the National Science Foundation (NSF). Its second customer, announced Monday, is San Francisco's IP Networks Inc., a privately held bandwidth provider. IPN evaluated Chiaro's router for two years before deploying, the company says, but details on the configuration, the number of units purchased, and the size of the deal were not disclosed.

There's also the matter of Chiaro's box lagging its competitors in some areas. In its initial configuration, the Enstara is only capable of supporting 100 Gbit/s of I/O (input/output) in a single system and only 200 Gbit/s in a rack, according to a report by Joe McGarvey of Current Analysis. "Though the Enstara can scale up to 315 OC192 ports from day one, it would take roughly 15 racks of telco space to achieve that capacity," he writes. He notes that, in their respective half-rack configurations, the Juniper Networks Inc. (Nasdaq: JNPR) T-640 offers 320 Gbit/s of I/O capacity, and the forthcoming Procket Networks Inc. Pro/8812 will offer 480 Gbit/s of I/O in a half-rack system.

In its Richardson, Texas, offices, Chiaro will proudly walk you through a demo of one of its systems outperforming a Cisco GSR 12000 system in a 64,000-user network when both routers are faced with outages of one sort or another. Routing experts, however, say beating a decade-old incumbent isn't that impressive. "It'd be more interesting to compare Chiaro to vendors who are pushing high reliability and high scaleability routing -- folks such as Avici Systems Inc. and Alcatel SA," says Geoff Bennett, director of Light Reading University.

Chiaro did go through the trouble of hiring BTexact Technologies to put the product through its paces and brazenly says it will release the full results of the tests to whoever requests them.

BTexact's comments were (surprise) very positive. In the test's executive summary, BTexact's technicians wrote that the product is "very stable and easy to use." Some key carrier features, however, such as Multiprotocol Label Switching (MPLS), quality of service, and multicast capabilities weren't included in the test.

Will the core router market be large enough to support another startup? Chiaro CEO Ken Lewis says carrier deployment decisions on adding more functions and capacity to core networks will start happening this year – and he's confident that Chiaro is in the running. "We feel we're on the leading edge of a window that's just opening up now," he says, while tugging at the buckles of his straightjacket.

"I think the barriers to entry in the remainder of the core market are very high," says LRU's Bennett. "Put yourself in the position of a buyer at Verizon Communications Inc. or Qwest Communications International Inc.. Why would you take the risk of not buying Cisco Systems Inc. or Juniper? What would make you try another vendor?"

— Phil Harvey, Senior Editor, Light Reading

Page 1 / 4   >   >>
veluthuru 12/5/2012 | 12:01:52 AM
re: Chiaro Debuts a Big, Bad Router
We had moved to the Dallas area from California couple of years back to join a local startup.

Locally, Chiaro was touted as having the most seasoned routing people. What I found when I interviewed was a bunch of developers
with 15-20 years experience - they would not let me in since I did not have as much "experience".
This kind of attitude pretty much speaks for most of the startups in the Dallas area.

Perception is more important than substance - this attitude is coming from the Nortel's of the region and is hard to change or have a successful startup culture.

I doubt there is any hope for this company. I would be very interested in hearing from the likes of Skeptic.

Well, I am heading back to California as soon as
I can.
jim_smith 12/5/2012 | 12:01:50 AM
re: Chiaro Debuts a Big, Bad Router BTW, why are you so interested in skeptic's answer? Is he based in the Dallas area? Do you know him/her personally?

The party is over for the folks that you interviewed with. They will be replaced sooner rather than later by programmers in (alphabetical order) China, Eastern Europe, India, and Russia. Future companies won't turn you down because of your lack of experience, but they will turn you down if they can't afford you...
skeptic 12/5/2012 | 12:01:48 AM
re: Chiaro Debuts a Big, Bad Router I doubt there is any hope for this company. I would be very interested in hearing from the likes of Skeptic.
-------------

Chiaro is competing for business with caspian
and procket now. On the positive side, they
have announced two small sales (with names) and
they went through an outside evaluation. On
the negative side, as the article suggests their
density isn't all that great. And they will
be fighting it out *to the death* with caspian
and procket for what core router opportunities
there are. None of these companies can afford
to lose any of what business there is.

Chiaro's pushing reliability hard with service
providers. If their story proves itself out
in real lab trials, it will be an advantage.
But the history on that isn't real good. Avici
talked big and then fell on its face when people
did real tests.

I've written off hyperchip until they make some
sort of annoucement about something. They
can't have much money left.





changeisgood 12/5/2012 | 12:01:46 AM
re: Chiaro Debuts a Big, Bad Router >>The Enstara can scale up to 315 OC192 ports from day one.

hey, i am *somewhat* familiar with the big ip backbones and 315xOC192 in a POP will not be needed until 2099. their business model created years ago is no longer valid and they will die soon.
skeptic 12/5/2012 | 12:01:42 AM
re: Chiaro Debuts a Big, Bad Router hey, i am *somewhat* familiar with the big ip backbones and 315xOC192 in a POP will not be needed until 2099. their business model created years ago is no longer valid and they will die soon.
-------------------

I think what you and others miss when looking
at core routers is that there is a difference
between "capacity" and what gets deployed. All
the companies in this space are claiming huge
scalability. The question for chiaro/procket
/caspian/et al is how flexabile the equipment
is in deployment and the upgrade path when
scale needs to be added.

The last generation of core routers had inflexable
backplanes which limited their life or at the
least made upgrades really complex. The new
generation of core routers, from all vendors
including cisco (HFR) and juniper (the external
switch & interconnect), is focused on
expandable backplanes and modular equipment.
They all claim different levels of scalability
and have different approaches but they all
involve modular expansion.

The question at those big ip networks isn't
so much *what it scales to* as *how easy can
I add capacity*. All the players in the core
have to answer that question.


Belzebutt 12/5/2012 | 12:01:30 AM
re: Chiaro Debuts a Big, Bad Router In one corner we have Tony Li, in the other corner we have an optical switch fabric... who would win in a fight?

The optical switch fabric is based on an optical phased array and can switch ports at nanosecond speeds, but Tony Li helped design the GSR and the M40 and has several RFCs under his belt... it's a close call!
signmeup 12/5/2012 | 12:01:28 AM
re: Chiaro Debuts a Big, Bad Router But isn't clustering exactly what Chiaro is betting on? A single Chiaro chassis can only support 100 Gb/s (in/out), so where is the 315 OC-192's coming from?

The real question is who needs 315 OC-192's right now anyway? It appears that most of the major players have been focused on adding more density in smaller spaces, not adding less density in larger spaces.

The problem comes down to investment protection. Everyone want a box that can scale from 1 Gb to 20 Tb without having to replace chassis, linecards, media adapters, route processors, ect. Given the densities Chiaro is purporting, I'm not you could house the thing at 20 TB - we're gonna need a bigger boat....

signmeup
skeptic 12/5/2012 | 12:01:28 AM
re: Chiaro Debuts a Big, Bad Router , both of these comanies have announced clustering capabilities to scale to many times that number.
--------------
There are lots of people who don't particularly
like the clustering approach. Depending on
how its done, it might work. But as often as
not, vendors are pushing "clustering" as a
stop-gap either because something didn't work
or something else is in progress. Cisco seems
to revive clustering as an idea whenever they
need to fill an inconvenent gap. But it never
happens.

As far as approaches, juniper has the most
chance of making clustering work. Procket's
story on clusters is so vague that I'm not
sure its real. And caspian (which is sort
of a cluster approach) comes with micro-flow
baggage.


signmeup 12/5/2012 | 12:01:28 AM
re: Chiaro Debuts a Big, Bad Router The main problem I see with Chiaro is that with all of its fancy optical phased array switch fabric it tends to be about 5 times as big physically with less density as other solutions. Anyone remember Monterrey Networks? Now that was a beast!

Taken directly from the article:
"the Enstara is only capable of supporting 100 Gbit/s of I/O (input/output) in a single system and only 200 Gbit/s in a rack."

"Though the Enstara can scale up to 315 OC192 ports from day one, it would take roughly 15 racks of telco space to achieve that capacity,"

So if we compare that to existing solutions available today, Juniper's T640 supports 640 Gb/s and Procket's 8812 supports 960 Gb/s in a HALF RACK! In addition, both of these comanies have announced clustering capabilities to scale to many times that number. Procket can reach 315 OC-192's in 3 1/2 racks - that's 76% less rack space and power required!

Given the fact that we all know what Tony has accomplished in past lives, I don't see how Chiaro is even in the same ring with these guys.....

signmeup
changeisgood 12/5/2012 | 12:01:25 AM
re: Chiaro Debuts a Big, Bad Router >> The real question is who needs 315 OC-192's right now anyway?

or when will someone need 315 x OC192 layer3 in a box? i would bet that some of the biggest IP POPs in the world have less than 12 x OC192 for WAN links, use NxGbE or 10GbE LAN-PHY for the layer 2 intraPOP aggregation (peering, edge boxes, distribution layer stuff) on the drop side of their core routers. then you split those WAN links over two boxes - because service providers have traditionally wanted two boxes or more side-by-side for resiliency. i wish more than anybody that we will see 300 x OC192 (turned on) in any core router.
Page 1 / 4   >   >>
HOME
Sign In
SEARCH
CLOSE
MORE
CLOSE