Invites Go Out for Edge Router Test

Service providers can face a tough time picking the right edge router -- partly because they're spoilt for choice and partly because it's tough to figure out which edge routers are most appropriate for particular applications.

Light Reading is hoping to shed some light on this issue by sponsoring the first ever independent test of edge routers. Invitations to participate went out on Friday (April 12) to more than 20 vendors. The testing will be conducted by Network Test Inc. using test equipment from Spirent Communications.

The test will focus on the following issues:

  • IP baseline performance
  • Resiliency/failover handling
  • IP Exterior Gateway Protocol scaleability
  • IP Interior Gateway Protocol scaleability
  • Quality of Service enforcement
  • MPLS VPN scaleability (optional)
  • MPLS Martini tunnel scaleability (optional)

A draft version of the test plan is available at http://networktest.com/edge02/edgemeth.html.

For more background on edge routers, check out Light Reading's archived Webinar on the topic, which will be available soon. And watch out for the associated report, scheduled for publication this week.

As it happens, Light Reading's April Research Poll is on the topic of testing -- and guess what? So far, edge routers and multiservice switches are winning the most votes for the type of equipment respondents would like to see tested.

The poll also illustrates the importance of ensuring that vendors don't pay to participate in tests. So far, 56 percent of respondents say this is "vital to ensure the independence of tests," while another 27 percent consider it "important but not essential." In none of the tests sponsored by Light Reading do vendors pay to play.

Light Reading is planning a major program of equipment and component tests this year. One on IPSec VPNs is now under way, and a test of sub-wavelength grooming switches has already been announced (see Light Reading Announces Switch Test). There are more to come.

— Peter Heywood, Founding Editor, Light Reading
Page 1 / 8   >   >>
null0 12/4/2012 | 10:36:39 PM
re: Invites Go Out for Edge Router Test Will you be publishing the names of those invited to participate ?
skeptic 12/4/2012 | 10:36:39 PM
re: Invites Go Out for Edge Router Test
Looking at the route capacity test, it seems
that you have listened to nothing that was said
during your core router tests and you (once again)
intend to use a method that favors systems
(including juniper) that have virtual memory.

To refresh your memory:

- A router that "learns" and advertises prefixes
that it can't use in forwarding is useless.
It doesn't matter if it can learn millions of
prefixes. If they can't fit in the forwarding
table, the router isn't functioning correctly.

- If a router starts to send route information
to disk and slows down dramatically (as does
the juniper) due to paging, the router is
not operating in an effecitive manner and any
benchmark measured is both misleading and
meaningless from an evalution perspective.

Once again, I would strongly suggest that you
ONLY measure:

- The route capacity that can actually be placed
in a forwarding table on the system.

- Route capacity numbers where the system is
not obviously abusing virtual memory or
hammering a disk to the point where router
performance is clearly degraded.

I really hope if you do not change the methodology
you will at least explain this time what your
reasoning is. As I have said before, the
only number that means anything to a service
provider is how many routes the system can
correctly learn and forward with while remaining
in its normal performance range.
skeptic 12/4/2012 | 10:36:38 PM
re: Invites Go Out for Edge Router Test
Once again, as in the core router test,
the test methodology
seems to ignore BGP route policy. Especially
on an edge router, this is a critical feature
and the performance declines associated with
applying policies (ip address filters,
AS-path regular expressions...etc) are critical
in the evaluation of BGP especially on an edge

Another flaw is that you don't test peering
capacity at all. How many BGP peering sessions
(IBGP or EBGP) the system can sustain is
another important criteria for measuring an
edge router.

Peter Heywood 12/4/2012 | 10:36:36 PM
re: Invites Go Out for Edge Router Test We hope we've sent invites to all companies shipping edge routers. If you've been missed, it's an oversight and you should get in touch with David Newman on [email protected]

On whether we will identify everybody that was invited: yes, when we publish the results.
photon_tim 12/4/2012 | 10:36:36 PM
re: Invites Go Out for Edge Router Test This looks like a test for a small core router.
Is this the final version or are there still additions going in. Like (some ideas only):
* Traffic shaping
* rate limiting
* Termination of hundreds/thousands of low-speed interfaces (DS0/FT1/T1/T3/OC3)
* General scalability (# OSPF adjacancies)
Peter Heywood 12/4/2012 | 10:36:36 PM
re: Invites Go Out for Edge Router Test ...are welcome...encouraged. Feel free to post them here or get in touch directly with David Newman on [email protected]
Holy Grail 12/4/2012 | 10:36:30 PM
re: Invites Go Out for Edge Router Test I guess you could also add billing and accounting features. These are Edge Router specific requirements.

As operators look to offer higher value services such as IP VPN and Martini type layer 2 transport services, then they are going to need to generate accurate bills. This is something that is not needed in a core router, but is needed in an Edge Router. Also Destination sensitive billing for IP transit services that allow the operator to count and bill for IP packets according to where they are being actually sent!
Belzebutt 12/4/2012 | 10:36:28 PM
re: Invites Go Out for Edge Router Test It would be fun to see how those DSL routers compare with the "real stuff" :)

Hey, it doesn't hurt to ask...
commonman 12/4/2012 | 10:36:12 PM
re: Invites Go Out for Edge Router Test Can you tell me what Lucent products will be invited or have made a commitment to show?
dnewman 12/4/2012 | 10:36:12 PM
re: Invites Go Out for Edge Router Test skeptic, I long ago stopped responding to anonymous posts. I'm going make a partial exception in your case.

Some of your statements about the edge methodology are way off base, and some of your comments I agree with completely.

I would be happy to discuss this with you, either privately or on the LR message board. This would make for a better test.

But first, as a member of the Network Test Frequent Critics Club, you ought to disclose any motivation you might have. You know who I am; I always sign my name and my affiliation.

Again, I'm very willing to talk openly if you're willing to do the same.

David Newman
Network Test

Page 1 / 8   >   >>
Sign In