Link Aggregation Options - vSphere 5.5 - Essentials Plus

[doofoo]

What Link Aggregation options are available with vSphere 5.5 w/ Essentials Plus.

Thought was to have 4 x 1Gbe and 2 x 10Gbe links for a small bundle.

2 x 1Gbe going to Switch A Gig-e
2 x 1Gbe going to Switch B Gig-e
Aggregated 4Gbit - Serves failover functions.


1 x 10Gbe going to Switch-C 10Gig-E Storage Network
1 x 10Gbe going to Switch D 10Gig-E Storage Network
Aggregated 20Gb - Serves failover functions.

Is LACP or some other aggregation supported in Essentials Plus in 5.5? From my research, it seems like it's not included because it's missing vDS. Do I have any alternatives?

Mostly looking at this from an aggregation, and fault tolerance point of view. Fault tolerance being more important than the extra aggregated bandwidth. Do I have other options?

 

[k1pp3r]

I'm on 5.0 but I use Route based on IP Hash and load balance src-dst-ip on my switches.

 

[doofoo]

I'm on 5.0 but I use Route based on IP Hash and load balance src-dst-ip on my switches.

Yeah it looks like that's the only options. Not true aggregation, but close enough.

How well does it handle the failover on the NFS mounts?

 

[k1pp3r]

I use fibrechannel so I can't answer that, sorry.

 

[Vashypooh]

I know with iSCSI and such, isn't it best practice from an MPIO standpoint to allow each to be standalone?

Atleast that is how it is with Hyper-V, 802.3ad on iSCSI is seen as detrimental.

 

[Child of Wonder]

You can't do LACP from a VMware host unless you're using a vDS or Nexus 1kv. When it comes to simple failover, VMware does a great job of this since you can simply assign multiple adapters to a vSwitch.

In your scenario, do your VM guests require 10Gb for their network traffic or would 1Gb suffice?

If 1Gb would suffice and you're only using NFS (no iSCSI, then you'd want multiple links dedicated to iSCSI for MPIO) then you could do this:

2x1Gb vSwitch0 - Management
2x1Gb vSwitch1 - VM traffic
2x10Gb vSwitch2 - NFS and VMotion
- NFS using 1st adapter active, 2nd adapter standby
- VMotion using 1st adapter standby, 2nd adapter active

Each traffic type has proper failover, 2Gb combined bandwidth for VM network traffic (VM NICs are randomly assigned to 1 of the 2 1Gb host NICs by default), 10Gb for NFS, and 10Gb for VMotion.

 

[doofoo]

You can't do LACP from a VMware host unless you're using a vDS or Nexus 1kv. When it comes to simple failover, VMware does a great job of this since you can simply assign multiple adapters to a vSwitch.

In your scenario, do your VM guests require 10Gb for their network traffic or would 1Gb suffice?

If 1Gb would suffice and you're only using NFS (no iSCSI, then you'd want multiple links dedicated to iSCSI for MPIO) then you could do this:

2x1Gb vSwitch0 - Management
2x1Gb vSwitch1 - VM traffic
2x10Gb vSwitch2 - NFS and VMotion
- NFS using 1st adapter active, 2nd adapter standby
- VMotion using 1st adapter standby, 2nd adapter active

Each traffic type has proper failover, 2Gb combined bandwidth for VM network traffic (VM NICs are randomly assigned to 1 of the 2 1Gb host NICs by default), 10Gb for NFS, and 10Gb for VMotion.

Yeah I guess that will be my option. I scale out VM traffic to like 4 x 1Gbe.

I actually have 6 x 1Gbe, then 2 x 10Gbe - 10Gbe is just for storage.


2x1Gb vSwitch0 - Management
4x1Gb vSwitch1 - VM traffic
2x10Gb vSwitch2 - NFS and VMotion
- NFS using 1st adapter active, 2nd adapter standby
- VMotion using 1st adapter standby, 2nd adapter active

Makes sense.. Saves me some money on the switches on the 10Gbe side.

 

[doofoo]

Any thoughts on the Powerconnect 8024F or 8132F for straight 10Gbe layer2 NFS traffic? I have heard good things when not using any advanced L2 or L3 feature-sets (Just dumb switch).

 

[Wrench00]

I hate dell switches myself the only thing you need to worry about is the cache on the switch. With NFS I tend to use Jumbo frames as it makes a substantial difference in my environments. Just make sure they got a boat load of cache which since these are 10gb they probably do.

 

[doofoo]

I hate dell switches myself the only thing you need to worry about is the cache on the switch. With NFS I tend to use Jumbo frames as it makes a substantial difference in my environments. Just make sure they got a boat load of cache which since these are 10gb they probably do.

I would never use Dell switches (Outside of the Force10 gear) for anything other than strictly Layer2 (single VLAN) as I would be doing here. I just don't see the sense wasting money without a need.