Comments on: HP C-Class Blade Chassis Review Part 2: The Cisco/Brocade Interconnect Switches

By: Brent Ozar

Brent Ozar — Fri, 15 Jul 2011 11:26:09 +0000

Hi, Robert. Sorry to hear about that. I don’t have any quick fixes – after all, we’re talking about something the manufacturer’s support can’t even figure out – so this sounds more like a consulting engagement to do troubleshooting. I’m available for that, but I didn’t want to turn this into a sales pitch – I bet you’re looking for easy, free fixes, as I would be doing in your shoes. If you get to the point where you want to engage us for consulting, email us at help@brentozar.com. Good luck!

By: Robert Bruce

Robert Bruce — Fri, 15 Jul 2011 05:53:42 +0000

I have a serious bandwidth problem when using 4/24 brocade switches as interconnects in the C-7000 chassis. The bandwidth I’m getting is about 10Mbs (or less) for each port across the entire chassis. I have 10 bl460c G1 blades running ESXi 4.1 U1 with Emulex HBAs.

HP level two support and Brocade cannot seem to find a problem. They are looking for errors and are not finding any.

It seems you have extensive knowledge of the chassis design and I’m hoping you can point me in the right direction.

Thanks

By: Frank

Frank — Fri, 14 Jan 2011 12:10:51 +0000

Hi Fernando,

wether this is normal or nut depends on two things which relate to each other.

Point one is the Bonding driver configuration of your blade. If you bond interfaces on both switches and have a active-passive configuration this might apply if, point two, there is no connection (uplink) to a somehow shared backbone with an appropriate VLAN setting.

So if you want this to be correct and make everybody communicate with everybody, check the bonding (aka adapter teaming) on the blades and then check your vlans and connections.

Regards F.

By: Tom

Tom — Thu, 13 Jan 2011 22:45:03 +0000

The advantage of the Cisco 3120 or the likes is that they can be stacked and managed as one logical switch, with a shared back plane.

Really the only advantages I see of the VC is one of management. You can move configurations around and have a nice GUI that server admins understand.

The VC are limited on their link aggregation, number of vlans they can actively support and I am not really impressed on what they can do from a cos/qos point of view.

The 64 million dollar question is what do the guys that like switches do when they want 10 GB? HP is not offering the Nexus 4K so if you want switching you are left with pass throughs to an external switch or the Procurve 6120.

From the little that I can find the 6120 does not stack, I can Link aggregate it to the switch in the corresponding bay but that is far from stacking, its just a link to another switch. If you want 4 10 Gb ports in each blade (which believe it or not some of my vmware costumers are asking for) I have to manage 4 separate switches in each blade enclosure.

I have customers looking at running View with 125+ desktops per blade, thats close to 2000 VM and I am going to need more than 2 10G connection to handle the networking from a client point of view, not even taking into account if the data is on nfs.

I understand that I look at a little differently because I work on the system and networking side but if you want to work with vmware you have to get up to speed on networking. Data center networking is going to keep changing at a very fast pace. Things like trill and otv are going to change tons of this.

I think HP is trying to make the VC go up against what the UCS can do. I am not sure thats the right move given the architecture of the C7000.