Trunking between HP5406zl and Netgear FS700TS

[SteveDolphin]

Hello there,

As per my other post querying STP I have a network that looks a little like this:

The workstations are all running at 100. They go into the Netgear (24x100) which then has an aggregated 2 x 1GB uplink back to the procurve.

What I find really strange is that those uplinks seem to hardly ever carry that much traffic. The individual ports will only be showing tens of, not even hundreds, of MB worth of traffic. I find it really strange that 24 machines (24 x 100 = 2.4GB, right?) aren't pulling more data down, and they're all busy all the time so...

I'm thinking that there's a (good) chance that the Netgears are so rubbish that they can't handle the traffic for the 24 workstations and so the uplink never gets used and/or I've done the lagging wrong somehow.

I've configured the trunk as a 'trunk' on the procurve and as a 'static lag' on the netgear, but in theory I could configure both as LACP. Would that be better do you think, or does what I've got already represent the best setup I'll get here.

The aim of the trunk/lag is to get more traffic down the uplink by having more capacity, with a secondary aim of it still working if one of the two cables fails. 

[Vince-Whirlwind]

The traffic stats you are looking at are probably averages over 5 minutes or something - workstations don't tend to use much sustained bandwidth.

If you want to see your utilisation go up significantly:

 - setup several workstations to do a very large (many GB) file trasnfer simultaneously

 - setup a VLC server on a host on your HP switch, then setup several VLC clients on your workstations and use them to stream a video simultaneously.

Static trunk v. LACP probably doesn't matter. Whatever works is fine.

[SteveDolphin]

Hmm, I'm getting the stats from IMC actually: 

Well that's the thing, the standard behaviour is that people open large files all day every day, so that should already be happening I think, but if there's no real difference in the LACP options then perhaps it's fine.

Do you think there's a problem with them being 'trunk' on the HP end rather than LACP on both?

[Vince-Whirlwind]

You can see your traffic stats are averages over 10 minutes.

If somebody opens a 10MB file, that is less than 100Mb/s for 1 second, or 160Kb averaged over the 10 minutes.

So your peak utilisation of about 4Mb in 10 minutes represents around 300MB of network traffic over that period.

If you can increase the granularity of the traffic stats, you will see much higher, sharper peaks.

The link aggregation at your end is probably fine - if you can see traffic coming in over both interfaces. If the other end is LACP and you don't have LACP configured on the HP end, I guess the other end might be using just one link. Just check that.

[SteveDolphin]

Thank you, you're right about the reporting certainly, I've changed the sFlow polling and sampling rates and I'm getting a much clearer picture now, thank you for your help!