Alcatel Stars in Multiservice Switch Test

COMMENTS

Newest First | Oldest First | Threaded View

Page 1 / 5 > >>

edgecore 12/4/2012 | 9:17:07 PM

re: Alcatel Stars in Multiservice Switch Test ============
Alcatel's lowest score -- 2 stars -- was for a test of 7670's MPLS tunnel reroute performance. It took up to 8.9 seconds to reroute 1,000 paths. Alcatel expects to improve this radically by implementing fast rerouting
============

I am curious as to the structure of the application that does this rerouting, can it be "parrallelized" at all? meaning if you tried it on a control card running 2 or even 4 Motorola PowerPC 7455's symmetrically, would this speed up the performance significantly?

EC

Reply | Post Message | MESSAGES LIST | START A BOARD

skeptic 12/4/2012 | 9:17:05 PM

re: Alcatel Stars in Multiservice Switch Test I am curious as to the structure of the application that does this rerouting, can it be "parrallelized" at all? meaning if you tried it on a control card running 2 or even 4 Motorola PowerPC 7455's symmetrically, would this speed up the performance significantly?
---------------------
The rerouting application doesn't go parallel
very well. The rerouting operations tend to
do operations on the same data set. Reservations
can have effects on other reservations in some
cases as well. I'm not saying its impossible,
but its not a classic divide-and-conquer
multiprocessor type problem.

Alcatel's rerouting takes way too long because
of design flaws in the software they are
(probably) using.

The comment about "fast rerouting" is meaningless.
Alcatel has a scalability issue in their software
and fast reroute doesn't fix it.

Reply | Post Message | MESSAGES LIST | START A BOARD

Tony Li 12/4/2012 | 9:17:04 PM

re: Alcatel Stars in Multiservice Switch Test I'm not sure I agree. If they are running a CSPF for each of those paths, then that translates to about 9ms per SPF. That's not TOO horrible, assuming an 'interesting' topology.

I think they might get a factor of 2 or 4 out of it, but it's not many orders of magnitude off.

Tony

Reply | Post Message | MESSAGES LIST | START A BOARD

skeptic 12/4/2012 | 9:17:02 PM

re: Alcatel Stars in Multiservice Switch Test Tell us why NT did not show off their outstanding performance ?
-------------

I have no idea. Why don't you grow up and
learn to live with the fact that your product
isn't perfect.

Every test I comment on, somehow like you shows
up and says I work for cisco/juniper/CWNT/
marconi/whoever. Great. Live in your dream
world.

Reply | Post Message | MESSAGES LIST | START A BOARD

lite-brite 12/4/2012 | 9:17:02 PM

re: Alcatel Stars in Multiservice Switch Test if no one else will say it, kudos to Alcatel for submitting their equipment. I know you'll all say that this wasn't tested properly, or that was missed out, but what the h**l, they put their box up for test and that's the bottom line! And don't give me that crap about 'why should companies put their gear up for test unless it will result in a sale (i.e. 'be designed in'); the same reason they advertise it!!
l-b

Reply | Post Message | MESSAGES LIST | START A BOARD

deer_in_the_light 12/4/2012 | 9:17:02 PM

re: Alcatel Stars in Multiservice Switch Test Tell us why NT did not show off their outstanding performance ?

Reply | Post Message | MESSAGES LIST | START A BOARD

yea_right 12/4/2012 | 9:17:01 PM

re: Alcatel Stars in Multiservice Switch Test "More than 15 other vendors were invited to participate in Light Reading's test, and declined or didn't reply. There were plenty of excuses, but the fact remains that only Alcatel has proved its switch works in an independent test. The rest haven't."

Come on, face it. It is a failure as far as lightreading is concerned. You guys couldnt convince more than one vendor for the test. That shows a lot.

I think, first you guys have to prove the world that you are credible both when you perform "independent testing" and while writing columns on your website.

Yeah, you can blame all those who didnt participate but why would they spend their $$ in this market for an "independent test" that may mean "nothing". If a test involves only one vendor, is it useful for anyone other than Alcatel ?

Grow up guys, if you want to win the industry, win the vendors, they make the industry. Prove them you are fair and they can trust you, before throwing mud at them.

Reply | Post Message | MESSAGES LIST | START A BOARD

ip-eng 12/4/2012 | 9:17:00 PM

re: Alcatel Stars in Multiservice Switch Test >Every test I comment on, somehow like you shows
>up and says I work for cisco/juniper/CWNT/
>marconi/whoever. Great. Live in your dream
>world.

Skeptic - you're right - its good to see your comments and people should not jump to the conclusion that you are working for the competition.

Reply | Post Message | MESSAGES LIST | START A BOARD

ip-eng 12/4/2012 | 9:17:00 PM

re: Alcatel Stars in Multiservice Switch Test
1)
>why would they spend their $$ in this market for
>an "independent test" that may mean "nothing".

Read the announcement - light reading paid the $$, so why should venders have a problem when someone else is paying for the tests?

These tests deserve special recognition on at least three counts. First, they are genuinely independent. Light Reading paid for them, not Alcatel.

=============================================

2)
>it was tested at EANTC who has no credibility
someone posted this about the 7770 OBX - so why would lightreading pay someone without any credibility?

Looking forward to all your answers.

Reply | Post Message | MESSAGES LIST | START A BOARD

broadbandboy 12/4/2012 | 9:17:00 PM

re: Alcatel Stars in Multiservice Switch Test skeptic wrote: "The comment about "fast rerouting" is meaningless. Alcatel has a scalability issue in their software and fast reroute doesn't fix it."

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

Skeptic, ignoring the software issue, it seems reasonable to me that implementing frr should improve rerouting performance. Isn't that what its supposed to do? With frr, back up paths are precomputed and presignalled. If the reroute can occur at any potential point of local repair, doesn't that simplify and speed up the rerouting process?

Of course, it would require a different test scenario to validate FRR performance, which is supposed to meet or beat Sonet APS.

BBboy

Reply | Post Message | MESSAGES LIST | START A BOARD

Page 1 / 5 > >>