0

We are using Backup Exec 16 for backing up couple of physical servers and approx 60 VMs running on vmware esxi 6.0.

Backup server itself is running on physical box Windows 2012 R2 and has got 2x 10 gbps NICs. All my ESXi hosts have 1x 10 gbps active NIC and 1x 10 gbps in standby mode. Im trying to setup the best scenario from networks point of view. Both NICs from backup are connected to HP Procurve 5400zl core spanning tree root bridge also all active NICs of all my ESXi hosts. On the Backup server side I enabled NIC teaming and here im confused which teaming mode (static,switch independent, LACP) and load balancing mode (address hash, hyper-v port,dynamic) should I use in order to get best network performance.

Here is a simple drawing:

enter image description here My current setup is:

Backup server NIC teaming - Static teaming mode with Dynamic load balancing mode.

On my procurve switch I did this config:

trunk E1,E7 trk3 trunk

I will appreciate any suggestion of how to improve my backup server load balancing and its performance.

  • LACP + dynamic. Nothing you do will make one 20G link. The hashing function will pin a connection to a single link. – Ricky Beam Nov 7 '17 at 20:24
  • Thanks for your suggestion, I did what you recommended me and I see that load balancing works great, but unforunately it did not speed up the backup speed, so problem is probably not on network side. – Xtriker Nov 9 '17 at 12:30
  • Unless you're backing up to disk (array of disks), the network is rarely the slowest part. – Ricky Beam Nov 9 '17 at 21:04
  • Did any answer help you? If so, you should accept the answer so that the question doesn't keep popping up forever, looking for an answer. Alternatively, you can provide and accept your own answer. – Ron Maupin Feb 19 '18 at 20:02
1

LACP trunking is recommended.

Since your traffic flow is mainly from the switch to the backup server and the 5400zl uses SA/DA traffic distribution, the load balancing is completely dependent on the source and destination IP addresses of the packets - each pair of source and destination IPs will always use the same physical link.

If you have many different flows they statistically distribute somewhat evenly. Very few flows will not distribute very well, only by luck or fine-tuning (=selecting the combinations of IP addresses that result in the desired distribution). A single flow will not benefit from trunking at all.

Usually, it's preferrable to not use link aggregation in layer 2 but to use multiple independent NICs with multiple IP addresses and do the load-balancing in the application in layer 3 (=using the different IP addresses).

In the best case, the application can be configured to distribute the traffic over multiple IP addresses. In another case, you use a DNS name in the configuration that resolves to multiple IP addresses with round-robin and the application alternates between them (either by not caching or on purpose).

I'm not sure how to do that with BE though. You might need to do some testing.

| improve this answer | |

Not the answer you're looking for? Browse other questions tagged or ask your own question.