Cisco's CRS-1 Passes Our Test

As the first commercially shipping router sporting OC768 interfaces, the Cisco Systems Inc. (Nasdaq: CSCO) CRS-1 Carrier Routing System is living up to its billing, according to results of a test commissioned by Light Reading.

The router passed its first-ever independent test with flying colors, giving some credence to Cisco's claims that the CRS-1 is winning hearts among the 14 carriers that have taken it for a spin (see Cisco Stumps for CRS-1). The CRS-1 test was a crucial hurdle to pass, according to Mike Volpi, the executive in charge of Cisco's router division (see the new Light Reading interview: Mike Volpi, Cisco Systems).

The results of the test, which was performed by the European Advanced Networking Test Center AG (EANTC), were published this morning (see 40-Gig Router Test Results for the detailed report). Light Reading and EANTC will present the findings in a special one-hour Webinar on Dec. 9 (details below).

The test was a Herculean effort in itself. (One contributor to the project described it as “trying to measure lightning bolts -- in mid-air.”) It took 140 hours over two weeks and involved a team of 11 engineers. That's just for the test itself; preparations required some heavy commitment from Agilent Technologies Inc. (NYSE: A), as Light Reading begged for the keys to one of the very few OC768 test ports in existence.

Not everything promised with the CRS-1 got covered in the Light Reading test. In particular, the equipment to create a gigantic multichassis router -- to try out Cisco's promised "92 Tbit/s" of capacity -- wasn't yet available. Nor were we able to create a full network of the systems to truly test their worthiness in a real-world environment.

But the testers were able to load one fully populated single-chassis with the Agilent interfaces we had available: 56 x 10-Gig plus 2 x 40-Gig equals 640 Gbit/s. They were also able to create a multichassis configuration with two of the chassis connected to each other. That produced 1.2 Tbit/s (full duplex) of capacity, and -- in case you're keeping score at home -- that's a massive IP traffic emulation.

In the other areas we could test, the CRS-1 performed on target. Those included:

  • IPv4 and IPv6 forwarding (and a mix, of course)
  • Performance under features such as quality of service (QOS) and security
  • Border Gateway Protocol (BGP) peering
  • Multiprotocol Label Switching (MPLS) performance
  • "Hitless" performance during hardware or software upgrades, or during a failure
The methodology was devised by Light Reading with scrutiny from some of the technology bigwigs at Tier 1 carriers.

The project didn't start out as a CRS-1 lovefest. Rather, Light Reading wanted to test all routers with 40-Gbit/s interfaces, a group test similar to those we've performed previously (for examples, see Corporate-Grade WLAN Test Proposal and Content Switch Test). But with Avici Systems Inc. (Nasdaq: AVCI; Frankfurt: BVC7) and Juniper Networks Inc. (Nasdaq: JNPR) saying their OC768 interfaces are still in the works, the CRS-1 became the lone entrant in this category.

That didn't spoil the party, though, because the CRS-1 itself is well worth a peel of the onion. The project started four years ago -- and was probably kicked around as a concept long before that -- as the Huge Fast* Router; and in the years before its launch, it was often the subject of industry whispers but never confirmed by Cisco. The covers finally came off in May, when the CRS-1 got its formal debut at a gala media event discussing the next 20 years of Cisco (see Cisco's Got a Terabit Router Too and Cisco Unveils the HFR).

Of course, that event came with extra helpings of hyperbole, but there's some truth in that "next 20 years" part. As Volpi noted in the aforementioned interview, the CRS-1 isn't financially important -- it won't rate more than 10 percent of Cisco's revenues, and that's if the routers fly off the shelves. The significance lies in the CRS-1 pieces that will define the future of Cisco's carrier business. For example, future carrier-class equipment will run on IOS XR, the modular operating system introduced with the CRS-1. Other features and advancements debuting with the CRS-1 will also see their way into other products, Volpi says.

The full test report -- along with some feedback from those carrier representatives, discussing the points the test missed -- is available here.

— Craig Matsumoto, Senior Editor, Light Reading

* Yes, we know the "F" didn't really stand for "Fast."

On Thursday, December 9, at 12 noon New York / 9 a.m. California / 5 p.m. London time, during a free hour-long live Web seminar, representatives from Light Reading, EANTC, Agilent, and Cisco will present the results of the CRS-1 test and answer questions about them. Telecom Italia will also be participating.

To register for the December 9 Webinar click here.

Page 1 / 4   >   >>
myoptic 12/5/2012 | 1:02:08 AM
re: Cisco's CRS-1 Passes Our Test A few years ago the stars aligned and both Cisco and Juniper (and others) agreed to a Light Reading core router bakeoff. Cisco lost.
Not surprisingly, this time Cisco refused to participate in the multi-vendor test LR originally envisioned. No - Cisco was only willing to play if they could define eligibility criteria which only it could meet.

C'mon LR do you honestly believe that OC-768 is so important that it should be the criteria for eligibility? Can you name a single carrier who has deployed it or has firm plans for deployment?
Didn't think so.

There are some interesting data points here, but the credibility of the whole exercise is undermined by the lack of participation from other vendors. Only participation from other vendors - not a smattering of carrier comments, would ensure a fair and thorough test plan and meaningful results.


gotman 12/5/2012 | 1:02:07 AM
re: Cisco's CRS-1 Passes Our Test >C'mon LR do you honestly believe that OC-768 is >so important that it should be the criteria for >eligibility? Can you name a single carrier who >has deployed it or has firm plans for deployment?
>Didn't think so.

Its eligibility is debatable.. What isn't is the need for 768. You haven't been to Japan, Korea, or talked to some EMEA customers in your company have you? When they start bundling 6X10gig between cities you know they need is there.

tsat 12/5/2012 | 1:02:06 AM
re: Cisco's CRS-1 Passes Our Test
LR did not bill this as a bake-off. This
was just a chance to test-drive the CRS-1.
A true bake-off of high-end routers would
probably be too expensive of an undertaking
for the LR folks in this non-boom day and age.
It is interesting to read, since very few
people have the ability to touch a CRS-1.

However, the fact remains that competitors still
cannot get a hold of a CRS-1 to really look for
and market its defects. Cisco has torn apart
the T640 and probably has engineers dedicated
full time to finding its flaws.

russ4br 12/5/2012 | 1:02:05 AM
re: Cisco's CRS-1 Passes Our Test Hey tsat,

Your posting(s) are number one in google now!

pavlovsdog 12/5/2012 | 1:02:00 AM
re: Cisco's CRS-1 Passes Our Test Question remains if 40G is still economically viable against 10G. Given how fast 10G prices have plunged (i.e. XFP modules), does 40G even make sense?

Does CRS-1 provide better port or power density at 40G than can be achieved at 10G?

Internal to the router, 40G must still be broken apart into lower rates, so there is an extra electrical mux/demux stage compared to 10G, too.
ironccie 12/5/2012 | 1:01:57 AM
re: Cisco's CRS-1 Passes Our Test We're thinking about CRS-1. Can someone please describe the power provisioning for this test? How much of a difference was their in electric bill for the month of this 10 day test as opposed from a normal month?

ironccie 12/5/2012 | 1:01:55 AM
re: Cisco's CRS-1 Passes Our Test Never mind... I found it. It's only 13.2 Kilowatts for a loaded chassis. Next question: How many of those units will it take to cause a California brown out? We plan to install one on stage as soon as Slipnot finishes playing.

ironccie 12/5/2012 | 1:01:54 AM
re: Cisco's CRS-1 Passes Our Test Geez... No help from the crew. I figured it out. Assuming 45,000 megawatts is going to bring everyone to their knees I'm sure we can install about 300,000 units if nobody uses their electric shavers and CA freezes over so we don't have to run the AC's to cool them. Realisticly, we can support about 100,000 of these units on the Internet here if nobody uses their PC's.

reoptic 12/5/2012 | 1:01:54 AM
re: Cisco's CRS-1 Passes Our Test The real measure of the CRS-1 is not in some limited lab trial but in the customer adoption. It has been 6 months since launch and no one has deployed this router despite enormous emphasis by Cisco. The article does not address that question. Why if this thing is so ready for prime time has no one adopted it and Cisco continued to lose share in the core?
materialgirl 12/5/2012 | 1:01:53 AM
re: Cisco's CRS-1 Passes Our Test Here is another question no one has even asked: Why, in the quarter after the CRS-1 was announced, did VZ suddenly show up out of nowhere as a 10% JNPR customer. It seems to me that this is not a coincidence, that VZ was waiting on this beast, saw it and went to JNPR. Am I missing something?
Page 1 / 4   >   >>
Sign In