[j-nsp] high 10GE port density in EX switch?

[Paul Zugnoni]

Only thing with the EX4550's is that they're only supported by latest
(read: bleeding edge) code. Nonetheless, I'm adding them in multiple VC's
in two locations where we had only EX4500's. Not the perfect switch, but
the VC tech has been sound for years with EX. If you're really curious,
ask me in two weeks how those implementations go (one is two VC's of 3
ex4550's, and another is two vc's of 1 ex4550 with 6 ex4200's.

Paul Zugnoni




On 2/2/13 3:40 PM, "JP Velders" <jpv at veldersjes.net> wrote:

>
>> Date: Sat, 2 Feb 2013 14:51:02 -0800
>> From: Brent Jones <brent at brentrjones.com>
>> Subject: Re: [j-nsp] high 10GE port density in EX switch?
>
>> There is always the EX4550, 32 10Gb ports in 1U, with a module to add
>> 8 more I believe
>
>Two module slots in fact. With 10GE and VC modules released. The 40GE
>module should debute somewhere this year (so they keep telling me).
>
>If you're going the _physical_ VC route with VCP ports it would give
>you 2.5:1 oversubscription (32 x 10Gbps vs 4 x 32Gbps). Going VCPe you
>can do a bit better with 2:1 oversubscription. If doing that look into
>the traffic patterns to figure out how far you can stretch the Virtual
>Chassis (max 10 members) before creating contention.
>
>I haven't seen official specs for the EX9200 yet, but supposedly that
>should fill the gap between EX8200 and MX I'm told. Apart from Qfabric
>itself, the QFX3600 with MC-LAG might also be a nice consolidation
>layer with its 16 QSFP+ ports between an MX and a bunch of VC'd
>EX4550's. Don't know if the routing on there (with VRRP and filtering)
>is beefy enough for the needs.
>
>In the end it boils down to not just the port-count, but the traffic
>and usage patterns. Unfortunately those aren't always that clear up
>front (or even in hindsight ;D).
>
>Kind regards,
>JP Velders
>_______________________________________________
>juniper-nsp mailing list juniper-nsp at puck.nether.net
>https://puck.nether.net/mailman/listinfo/juniper-nsp