[f-nsp] NetIron MLX-4 vs Juniper MX240

Richard A Steenbergen ras at e-gerbil.net
Fri May 7 11:52:36 EDT 2010


On Fri, May 07, 2010 at 07:21:35AM -0400, Scott T. Cameron wrote:
> Some of the routing engine tasks done on the SRX series are really,
> really slow. I have my SRX series taking a full route table from my
> upstream providers today on BGP.  This process can take about 5
> minutes to complete.
>  Obviously, this is insane.

Heh yes SRX is painfully slow. I have an SRX210 in my basement for my 
home connectivity and it takes a good 30 seconds to commit even a basic 
config. We run MX's with 10k+ line configurations, a heavy dose of 
commit scripts, and on-commit RE sync, and it only takes around 10 
seconds at maximum (on RE-2000).

> How fast are the Juniper MX series at taking / injecting routes?  How is the
> failover convergence?

The juniper-nsp list is probably a better place for a detailed
discussion on this. Generally speaking Foundry is very "lightweight", so
it tends to perform quite well on CPU intensive tasks like this, but at
the expense of features and functionality. Feature-wise there is
absolutely no comparison, Foundry gets stomped hands down, but like I
said if you run a simple network and don't need advanced features
Foundry can become interesting based on price and the ability to
reliably deliver line-rate traffic (something that can't be said for
6500/7600 :P).

> For the CLI, I'm not sure that I like JunOS better than the IOS clone. 
> My former career was entirely Cisco, so the IOS clone is like an old,
> familiar friend for me. I have more or less gotten over the learning
> curve with JunOS, though.  And I do like it is a familiar FreeBSD
> system underneath, even allowing you to go in to a shell.

Once you go JUNOS you can't go back. There are things we do with policy
statements and commit scripts that are worth their weight in gold, that
couldn't be duplicated under IOS let alone under Foundry if their lives
depended on it. But the learning curve is steep, and for people who are
more comfortable with IOS-like CLI Foundry can provide an acceptable
solution. For my needs (running a complex network) we found Foundry L3
to be completely unworkable, and had to pull them. But we turned around
and sold our MLX's to a customer who had far simpler needs and they've
been quite happy. It entirely depends on your network and your needs.

> Some of their "studies" show that the MX-series routers have some trouble
> during failover.  I am taking it with a grain of salt, just curious of
> anyone has had those experiences.

Cisco paid those guys a lot of money to write absolutely absurd bullshit 
about the competition they are most afraid of, and MX certainly has them 
running scared. There have been plenty of discussions on j-nsp about 
exactly which parts are complete and total fabrications, you might want 
to check the archives from "Miercom".

> Last but not least, support.  Both vendors have terrible technical
> support when you have a bug, in my experience.  I loathe having to
> open a case.  Is it different for support on the router series?

Not that Juniper support is perfect (you'll find me ranting about it 
quite a bit on j-nsp :P), but MX is handled by a completely different 
support group from SRX/J-series (which is very enterprise support 
focused, i.e. far more resources devoted to answering "RTFM" questions 
:P).

-- 
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 foundry-nsp mailing list