Carriers Weigh Savings With Cisco CRS-1

At the launch event yesterday for its CRS-1 router, Cisco Systems Inc. (Nasdaq: CSCO) demonstrated an array of futuristic services that its new, massively scaleable box will enable telcos to provide -- from videoconferencing to multiplayer gaming (see Cisco Unveils the HFR). Dazzling as those services may be, carriers are also interested in the savings they stand to gain from a piece of equipment that aims to change the way they buy and replace routers.

Typically, carriers purchase routers much as corporations buy PCs, replacing older boxes with newer, faster ones every five to seven years as microprocessor speeds improve. That’s a far cry from the way phone companies have historically bought switches for time-division multiplexed (TDM) voice traffic. Once a phone company installs a TDM switch, it may stay in a network for up to 20 years, during which administrators add incremental upgrades to it.

Cisco is taking a step toward that traditional approach with the CRS-1, which is designed for a lifetime comparable to that of a TDM switch (see Chambers Expects Cisco Dominance). “Once we put it in place, we don’t want to move it for one to two decades,” said Cisco CEO John Chambers in a speech yesterday, echoing the requirements customers gave the vendor as it built the CRS-1. The vision is this: Instead of replacing routers every five years, carriers add new linecards and software modules to the CRS-1 to meet changing demands and introduce new services. But the tool-shed-sized box stays in place.

Will it really work that way? “Let us pray,” says Jack Wimmer, VP of network architecture and advanced technology at MCI Inc. (Nasdaq: WCOEQ, MCWEQ). On the surface, the notion of buying linecards and software updates instead of new routers seems simply like spending the same money on one set of Cisco products instead of another. Not so, says Wimmer, who expects to reap significant savings in capital and operating expenditures from the CRS-1. “It’s a lot more cost effective than fork-lifting a whole box out and putting a new box in,” he says.

With a new box come new management systems, new maintenance, new operations, and new training for administrators, all of which cost more than adding new linecards and software incrementally as demand dictates, Wimmer says. MCI has not yet deployed the CRS-1 but plans to begin testing it once it ships in July.

The flipside of this argument is that carriers might have to pay even higher prices for new linecards, software, and so on, once they've committed to a long-term strategy based on the CRS-1. Right now, however, this doesn't seem to bother carriers too much. Cisco reaped a gross margin of 68.8 percent in its latest reported quarter, and yet it got high ratings from carriers in the market perception survey conducted by Heavy Reading, Light Reading's market research division, last September (see Cisco Delivers, Sees 5% Growth in Q4 and Cisco Winning Market Perception War).

The only carrier that has put the router into production so far is Sprint Corp. (NYSE: FON), which began testing the CRS-1 a year ago. The carrier is using a single-chassis configuration of the box as a core router connected to an OC48 line at its facility in San Jose, Calif., where it has previously used gigabit switch routers (GSRs) from Cisco.

“Where the large benefit comes in is one CRS-1 replaces eight GSRs,” says Kathryn Walker, executive VP of network services at Sprint. Sprint will study its CRS-1 in San Jose for the rest of this year before determining whether to install more of the systems in its network. Once this decision is made, Sprint would deploy the CRS-1 in its "densest locations," says Walker, and "redeploy the GSRs we have today to other places in the network that don’t have as much capacity."

Sprint will pay special attention to whether the savings in capital expenditures from replacing GSRs outweighs the operating expenditures from installing and maintaining the CRS-1.

— Justin Hibbard, Senior Editor, Light Reading

Page 1 / 3   >   >>
tsat 12/5/2012 | 1:42:57 AM
re: Carriers Weigh Savings With Cisco CRS-1
I doubt it, the techology moves to fast. In 12 years, what is a customer going to do, buy another 7 foot tall, 24" rack, or a 6-U box that does the
same thing for 1/8th the power?

I am betting they will prefer not to use the forklift....

There is a reason Mainframes died...

Bush sucks 12/5/2012 | 1:42:55 AM
re: Carriers Weigh Savings With Cisco CRS-1 I agree. This is the biggest con that vendors have used and carriers seem to be unwittingly buy into.

(1) What is the depreciation cycle on these platforms? I would be surprised if it were more than 5 years. Depreciation over 20 years?? Are they nuts??

(2) Shift in architectures.
In 5 years, I can certainly see platforms that converge IP and optical switch technologies and do more cost-effective transit and bypass. Do you really think there are lots of POPs that will need to actually terminate more than 1 Terabits for distribution to customers in that region? Cheaper OOO cards can be used for pure transit with Packet cards used only for actual customer termination. OOO cards may not be reality now but in 5 years ....

Why would anyone want to invest in a fixed architecture and be captive to a vendor for 20 years is completely beyond me. It is time so-called "network architects" in these carrier companies start "architecting" and not fall prey to vendor "marketectures"
materialgirl 12/5/2012 | 1:42:55 AM
re: Carriers Weigh Savings With Cisco CRS-1 The idea of moving into a footprint then milking it for decades is an old one, used by LU, IBM and others.

The idea that you need up upgrade your box every 5 years presumes that Moore's law will continue to operate. This, however, is not the opinion of leading semiconductor vendors at this time. The leakage and the NRE and processing costs of the 90nm process are proving to be roadblocks. This is why INTC and AMD are moving away from clock speeds in naming their chips.

If processor speed gains indeed slow, then the lifecycle of a footprint can be extended, for nothing much better exists to take its place. That is, if the footprint goes in in the first place. That is CSCO's challenge. Big is not necessarily better, as INTC is learning with the Itanium.
optical_dude 12/5/2012 | 1:42:54 AM
re: Carriers Weigh Savings With Cisco CRS-1 I don't think 5 years is right. It's way too short. CRS-1 scales impressively to 92Tb/s. Lifecycle is determined by the operator and those guys like to amortize over decades, especially if the equipment is still useful and generating revenue.

My 3 cents
arch_1 12/5/2012 | 1:42:50 AM
re: Carriers Weigh Savings With Cisco CRS-1 A scalable router should allow the user to add a new chassis without taking the existing chassis out of service. If Cisco designs new chassis that are fabric-compatable with the old, then by the tiemthe router has grown to its maximum capacity, it may be 20 years old without ever being taken out of service. The last chassis to be added will have 64xOC-192 in a 3U form factor.

The CSR-1 fabric uses a Benes topology. I know nothing of their design, but a Benes is theoretically arbitrarily scalable by adding stages. They may be able to add stages ten years from now, again without taking the router out of service. This would permit them to exceed the 72-chassis limitation. In 20 years, at 100%/y growth, a site that now needs one chassis will need 1M chassis, so we better hope the chassis get smaller :-)
linuxz900 12/5/2012 | 1:42:46 AM
re: Carriers Weigh Savings With Cisco CRS-1 Agreed. We carriers will gladly use a device with a mean-time-to-failure rate of 20-30 years, like a mainframe!

On a side note to the original post, mainframes are not dead.
My company (Sprint) just installed 3 new ones. One of them runs some IFL (Linux-only) engines and our testing shows it beat every other platform we have for Linux, which is just about every major player on our datacenter floors.

I knew about the testing for the CRS-1 in Cali last fall, don't think we're going to go crazy and install them everywhere without seeing some ROI in both the near and long term.

We do have plenty of "owned" devices from the 80's that are providing great service and benefit and are just amazing to be running still. Maintenance is much cheaper than replacement cost.

My opinions are my own.
Tony Li 12/5/2012 | 1:42:46 AM
re: Carriers Weigh Savings With Cisco CRS-1
Yes, you can add stages, but the real question is about the overall architecture. If you add stages and other chassis, but cannot send bits to those new chassis from older edge chassis, then you haven't been able to grow smoothly.

The architectural scalability of the fabric addressing is key, and Cisco clearly undershot on this one.

arch_1 12/5/2012 | 1:42:44 AM
re: Carriers Weigh Savings With Cisco CRS-1 Responding to Tony:
As I said, I don't know the details of the CSR-1 fabric architecture. They may or may not have made provision for later addition of stages. If they did not make such provision, then a CSR-1 can never exceed 72 chassis with 64xOC-192/chassis. At 100%/yr growth, a site that currently needs a full single chassis will exceed the capacity of a CSR-1 in less than seven years. Last I looked, seven is less than twenty. If Cisco did allow for more stages, then mathematically they gain an additional N years per stage, where N is the out-degree of a fabric switching element.

Of course, there is a huge difference between computations based on the mathematics of a Benes topology and the actual constraints of their architecture. If anything on an individual port card must have knowledge of all other port cards, the maximum size of the CSR-1 will depend on its least-capable port card rather than on its fabric.
icenine 12/5/2012 | 1:42:39 AM
re: Carriers Weigh Savings With Cisco CRS-1 Do I sense a lack of nerve, or a drop in sense of humor?

What happened to the HFR name? What did that stand for anyway?
andreasb 12/5/2012 | 1:42:38 AM
re: Carriers Weigh Savings With Cisco CRS-1 This seems to be relevant (given the interest in machines with long life cycle), and I do not want it to sound like a commercial:

Dune Networks (VLSI vendor) is currently shipping a fabric and TM scaling from 20Gbps to 40Tbps, which enable the design of products with 10 years of life cycle

- The fabric is based on CLOS interconnect, but as mentioned in previous discussions, the routing strategy on top of the interconnect is the important aspect.

- 32 half-rack IO chassis of 640Gbps each (and maybe 1.2Tbps within a year or two) can be connected via several fabric chassis.

- All TM devices (existing rates and future rates) work together by design.

- Nothing prevents future chassis to be much denser but still connect to older chassis (as IGÇÖm sure is the case with the CRS-1).

- Nothing prevents future line cards to be of higher rate (or denser) GÇô as long as the backplane is designed to support these rates

- All of the above mentioned configurations are non-blocking and the rates are full-duplex. Finally, this technology is available for anyone to use...

- Needless to say we believe that switches/routers should be designed like TDM equipment (it does not make a lot of sense that the mechanic box will limit the capacity of the switch).

At present, the majority of the designs using these devices are for 20G to 640G or even up to 1.28T product lines, but it is nice to know you can scale...

Andreas (Dune Networks employee)

PS: I would like to apologize for posting almost the same message twice. I do not intend to do it again.

Page 1 / 3   >   >>