Trunking with Procurve 2848 & Windows 2008 server

[BillClark22]

Have a Proliant DL380 with a Broadcom 1gb dual-port nic on the mainboard.  I have teamed both NIC ports together using the Broadcom Advanced Control Suite (BroadcomTeam.jpg).  I now want to create a trunk on the Procurve 2848 switch for the ports this server is connected to so it can provide 2gb to the server.  I used the CLI command "trunk 13,14 trk1 lacp" on the switch and here are the results (ProcurveTrunk.jpg).  It appears to be ok, the server it talking to the rest of the network, but when I run a test backup job on the server (it's a BackupExec 2010 server), I only see port 14 with any activity.  So, is this really working the way it should?  Is there a more accurate way to determine if the trunking is functional?  I had thought to see both ports on the switch with close to the same activity.  Need some clarification here.....Thanks.

Bill

[Fredrik Lönnman]

According to your attached screenshots the trunk seems to be in order. What you are seeing are standard nature of port trunks, which is more like load balancing than load sharing. Im not 100% sure about the 2848 switch but probably it has L2 or L3 balancing algorithms which means that traffic between a specific mac- or ip source-destination adress pair will use the same port, it wont be shared over both.

If you run two backup jobs to different machines you will probably see utilization on both your ports.

[BillClark22]

Is there any better option for achieving what I'm trying to do, besides buying a better switch?  I would really like to fully utilize both gigabit links coming into that server if possible.  Don't think adding any more NIC's and expanding the switch trunk would really help would it?  Thanks!

Bill

[Tijl van der Steeg]

That's a limitation of the ProCurve line. Does the Broadcom suite, like the HP NCU, have a setting for "Switch assisted load balancing"?

[BillClark22]

I don't see anything mentioning "switch assisted load balancing", the options I have for team type are as follows: Smart load Balance and Failover, Link Aggregation(802.3ad),  Generic Trunking(802.3ad-draft static) & SLB Auto-fallback disable.  The team is currently set to Link Aggregation.  I'm also using v3 of the Broadcom Advanced Control Suite.

Bill

[Helper]

Hi,

To be sure that your LAG is working, if you start a ping to the server, then unplugged the first is it always working ?
Try the same thing with the second wire after plugging the first one.

If possible, can you give us more information regarding your backup exec data flows.
Is your server is using only one MAC and IP address ?
During your test, how many end-nodes were concerned by the backup exec task ?

On this switchs, the load-balancing is done regarding the couple mac-src/mac-dst, or ip-src/ipdst (i can't remember exactly). So if you have only one couple regarding for every flows then it is normal that only one link is load.

Bye,