[f-nsp] Use of RX-4 / MLX-4

[i3D.net - Martijn Schmidt]

Hi Bardo, 

One thing you should be aware of is that the RX are pretty much at the end of their life, and no longer being sold in Europe. Brocade isn't implementing new features anymore as far as I know, and one of those missing features is IPv6 VRRP-e support which you'll want to make your server gateways redundant. 

In our experience running lots of ARP and mac addresses on the RX will also melt it down. The MLXe doesn't have this problem. 

I would recommend getting MLXe routers for all locations: go with M-blades for your server PoP (max 256k routes in hardware if you want to use high density blades) and X-blades for your full table BGP PoPs (max 1M routes in hardware). Try to use only 8x10G blades or higher density, you'll get some more perks out of that such as per-VLAN snmp polling and better use of your switching fabrics. 

Best regards, 

Martijn Schmidt 
i3D.net 

On 22 May 2014 09:42:58 CEST, Bardo Cornelissen <b.cornelissen at caveo.nl> wrote:
>Hi,
>
> 
>
>I'm considering using RX-4 and/or MLX-4 routers to connect 3
>datacenters.
>Two locations are mainly for carrier uplinks (BGP IPv4/IPv6) and are
>cross-connected. The third locations is connected to both other
>locations,
>doesn't have to do BGP and serves managed and collocated servers.
>
> 
>
>Currently we use OSPF for default and internal Ipv4 routes between the
>locations, which works great and I would prefer to maintain that setup
>but
>also apply that for IPv6.
>
> 
>
>I'd consider MLX routers for the BGP locations because they would allow
>larger route tables (or at least in hardware) and more peers.
>
>For the server location the RX-4 (2 units) should be sufficient only
>requiring OSPF for no more than about a hundred routes. Now could you
>guys
>conform the RX-BI-MR would be sufficient for the RX-4 units,
>considering I'm
>not going to do BGP on them?
>
> 
>
>Another thing I'm struggling with is that I've been told that the RX
>series
>still push IPv6 through its CPU, causing major performance limitations.
>I've
>also been told this is not the case with the MLX series. However, I
>cannot
>verify this statement anywhere, nor can I track any difference in the
>datasheet that would implicate there is such difference. Is anyone
>familiar
>with this and can tell me whether this is true or a myth?
>
>Kind regards,
>
> 
>
>Bardo Cornelissen.
>
>Caveo Internet BV
>
>
>
>------------------------------------------------------------------------
>
>_______________________________________________
>foundry-nsp mailing list
>foundry-nsp at puck.nether.net
>http://puck.nether.net/mailman/listinfo/foundry-nsp

-- 
Sent from my Android phone with K-9 Mail. Please excuse my brevity.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/foundry-nsp/attachments/20140522/d01aa2d5/attachment.html>