[j-nsp] Miercom Competitive Performance Testing Results: Cisco ASR9000 vs Juniper MX960

Richard A Steenbergen ras at e-gerbil.net
Fri Sep 25 10:13:47 EDT 2009


On Fri, Sep 25, 2009 at 08:11:04AM +0300, Pekka Savola wrote:
> On Thu, 24 Sep 2009, Richard A Steenbergen wrote:
> >Well, we saw something similar (that I've described on this list several
> >times) but not exactly what you described, starting somewhere in 7.x and
> >continuing until 8.5 where it was fixed on most platforms (including
> >MX).
> 
> For reference, this thread from Feb 2007 describes it at least once:
> http://www.mail-archive.com/juniper-nsp@puck.nether.net/msg00188.html
> 
> We saw this issue as well, but it apparently (for us), resulted in in 
> RIB sync slowness only; there was no (apparent, or at least major) 
> packet loss.
> 
> Somewhere around 8.x -- maybe 8.5 or earlier -- it got worse instead 
> of got fixed.  Now it's causing significant packet loss as well. 
> Maybe we have a cornercase that got worse.  Or there really are some 
> architecture specific differences that might explain the difference in 
> behaviour (in particular I'm speculating about VERY old, memory/CPU 
> constrained T320 FPCs).  But the bad news is that JTAC's response is 
> that this is working as designed.

Well we didn't actually run 8.5 anywhere, but we were told that was
where the fix was implemented. We skipped directly from 8.4 to 9.0 (and
then very quickly to 9.1, 9.2, and eventually 9.3, running away from one
massive bug after another) on MX. I'm absolutely positive this issue got
fixed in 9.0+ though, trust me we had a BIG party to send it off. It
caused easily 3 years of living hell back when it affected M160 and
early MX use, and nobody would believe us that it existed.

What code are you running now? If you're still running older code (and
aren't running logical-routers) my recommendation is to try 9.3R4. I
can't vouch for newer code nearly as well (we did find some serious bugs
in early revs of 9.4 and 9.5 at any rate, I just can't speak to the
current revs), but we've beat the crap out of 9.3R3 and 9.3R4 and found
it to be one of the most stable MX images to roll out since the box
shipped. Give that one a shot and see if it doesn't fix your issue.

-- 
Richard A Steenbergen <ras at e-gerbil.net>       http://www.e-gerbil.net/ras
GPG Key ID: 0xF8B12CBC (7535 7F59 8204 ED1F CC1C 53AF 4C41 5ECA F8B1 2CBC)


More information about the juniper-nsp mailing list