[j-nsp] SRX650 Clustering Issue

Pavel Lunin plunin at senetsy.ru
Wed Mar 9 15:18:08 EST 2011




On 09.03.2011 21:47, Stefan Fouant wrote:
> What Ben is saying is that you it is simply not necessary to configure the
> AE interface when doing this on a Clustered device.  Basically, when you are
> doing clustering, you simply add multiple ports from the same node to a RETH
> interface and this bonds them together to create your Link Aggregation
> Group.  To answer your question, load-balancing should happen by default
> amongst all the ports in the RETH on the Active node.  BTW, I've only done
> this on the High End SRX platforms... I'm not sure if this support exists on
> the SRX 650.  So you may want to check with your account rep or do some
> digging to determine if this support is provided on the Branch SRX devices.
>
No, LAGs are not supported in cluster mode for branch SRX.

http://www.juniper.net/techpubs/software/junos-security/junos-security10.4/junos-srx-jseries-support-reference/jd0e3251.html

Though, to be honest, I don't really believe it's much needed for the 
low-end. At least I am sure it is not in most cases. SRX650's 
preformance is 2,5Gig for IMIX. In case you don't use a 
router-on-a-stick topology, two 1-gig reths is something close to this. 
I bet most deployments of such a box allow splitting VLANs into several 
physical reths.

In case you really have all the 2,5 gigs coming, say, from the Inet to 
one internal VLAN and you have no way to split this load somehow using 
more interfaces in the routing scheme, well… 10GE card, which is 
available for SRX650, is your friend. However it's more likely, that you 
need more powerful gear for this. This is not the case for most of the 
deployments of SRX branch, at lease of those I've seen.


More information about the juniper-nsp mailing list