HPE Synergy

"Load Balancing" in and Active/Active network configuration


"Load Balancing" in and Active/Active network configuration

The question I've been asked, "Does an active/avctive configuration balance traffic across multiple uplinks if traffic from  Compute Module A higher than Complute Module B?"   I've been looking through the "HPE Virtual Connect traffic flow with HPE Synergy" document and I've run into a couple of different takes.

On Page 14.  "Classic Active-Active Configurations" I see this 

"There are no stacking links, and compute modules are responsible for load balancing across ICMs cross-connected to the uplink infrastructure.  Does that mean the load balancing is left to the application or OS?  Am I correct in thinking that uplink traffic balancing does not occur? 

Thank you, 

Dave Thacker 

"Leveraging Synergy across all technology platforms!"--Scott Adams

Re: "Load Balancing" in and Active/Active network configuration

The term "Active/Active" refers to the legacy c-Class Virtual Connect method of allowing all connected uplink ports to be Linked/Active state.  By default, in c-Class Virtual Connect, uplinks that span multiple physical interconnect modules will not all be Linked/Active.  If they were, a loop would be introduced as LACP is not possible in c-Class VC in that mode.  So, the concept of A-side and B-side networks was created.  That would mean within VC, both A and B networks would use the same VLAN ID (external that is), within VC they would be unique.

HPE Synergy VC does not have this concept per-say.  When you design and create a Highly-Available or Redundant Logical Interconnect Group, you would assign all needed uplink ports to the same Uplink Set.  For those ports to be Linked/Active you need to support LACP on the upstream switch side.  This will typically mean a multi-chassis LACP protocol/configuration would be needed; sometimes referred to as M-LAG or MC-LAG.  3COM/H3C/Comware use IRF.  Cisco uses vPC.  Juniper uses Inter-Chassis Control Protocol (ICCP).

If your networking infrastructure is unable to support this, then you would create two, independent, non-redundant logical interconnect groups (LIGs).  You would not use the A-side and B-side networking concept found in c-Class VC.   You can still use LACP on the uplink ports, but they must terminate to the same switch device.  Otherwise, you will see some Linked/Active with others Linked/Standby.

So, how do you achieve load balancing in this non-MLAG, Active/Active type configuration?  It happens on the host side.  You can use Transmit Load Balance (TLB) or Switch Independent Load Balancing (similar to LACP, but without the extremely low port failover capability).

I am an HPE employee

Accept or Kudo