Fibre Channel (SAN)

Reply
Occasional Contributor
Posts: 6
Registered: ‎05-10-2011

True bandwidth utilisation

We have 4 x DCX's (Swithtype 62.3) each with 6 blade slots in them. From what I read they have 256 Gbit/sec of slot bandwidth between each blade and the core switching blades.

 

What I want to know given that all of the blades are fully populated (48 slots per blade are in use) is how can I see what is actually happening on each of the slots. The reason I ask is that we have over 20 c7000's connected upto the DCX's with 4 hyper-v blades hooked into each c7000. This provides many hyper-v clusters all serving hundreds of VM's.

We also have multi storage vendor arrays in the fabric. hp-eva, svc and netapp.

There is also cross site replication using metro mirror between SVC.

Most of the storage is local to the servers but there is also an LSAN to an EVA to provide storage for 1 of our svcs.

There are 2 trunks of 4 x 8gb links connected to our backbone.

 

I'm not great at expressing what i'm after but I guess i'm keen to workout if the  port traffic is exceeding slot bandwidth.

 

 

Valued Contributor
Posts: 931
Registered: ‎12-30-2009

Re: True bandwidth utilisation

I'm not great at expressing what i'm after but I guess i'm keen to workout if the port traffic is exceeding slot bandwidth.

 

With ^^^ being your goal, you need to know of much traffic is locally switched ie traffic that is handled by one ASIC.

 

The switch you refer to is an 8GB platform, so I therefor guess the blades you use are FC8-48 ones.

This blade is oversubscribed 24 (external ports per ASIC ) to 16 (internal ports per ASIC) IF traffic needs to be routed to other blades OR the other ASIC on the same blade.

So in theory 48 external ports @ 8Gb requires 384Gbps slot bandwith.

 

Most of the storage is local to the servers

 

I don't know how to interpret local, do you mean what Brocade means with local switching?.

 

 

 

 

Occasional Contributor
Posts: 6
Registered: ‎05-10-2011

Re: True bandwidth utilisation

Thanks for replying - yes these are FC8-48 (sorry omitted that). Ok so I need to look at the portsets on these so see the traffic handled by 1 asic so 1 asic - port inde 8 -23, 40-47 and the other asic being 0-7, 24-39.

 

What I mean by local to the storage is my storage is housed in the same data centre building as the hosts. So the majority of the workload I assume is performed via local switching (maybe i've assumed to much here).

 

IF traffic needs to be routed to other blades OR the other ASIC on the same blade - would this occur for example in a cluster setup - i.e 1 node is on 1 portset (asic1) and the other in another portset (asic2) ?

 

 

Valued Contributor
Posts: 931
Registered: ‎12-30-2009

Re: True bandwidth utilisation

What I mean by local to the storage is my storage is housed in the same data centre building as the hosts. So the majority of the workload I assume is performed via local switching (maybe i've assumed to much here).

 

In Brocade terms local switching means traffic doesn't need to leave the ASIC.

On a FC8-48 for instance traffic from port 0 to port 25 is locally switches.

Thus is does not need to cross the backplane.

 

Traffic from port 0 to port 45 however crosses the backplane as both ports are on different ASIC, regardless of being one the same blade or not.

 

would this occur for example in a cluster setup - i.e 1 node is on 1 portset (asic1) and the other in another portset (asic2) ?

 

If there's traffic between nodes 1 & 2 then yes.

Join the Community

Get quick and easy access to valuable resource designed to help you manage your Brocade Network.