BladeSystem Virtual Connect
Showing results for 
Search instead for 
Did you mean: 

VC performance question

Trusted Contributor

VC performance question

Abhishek had a customer ask about VC pperformance vs. Pas-thru Modules:



Are there any performance benchmarks available for VC, in terms of impact on throughput? A customer is interested in the comparison between situations where a VC is used and where a pass-through is used.

Along the same lines – what is the level of LACP support on VC and how does LACP affect the throughput seen by the servers?


In principle, LACP should allow aggregation of available VC uplinks into a trunk, hence is it possible to connect N wires across VCs and get a trunk of Nx between them; ie. Say if we link up 8 wires between 1Gbps ports between VC1 and VC2, will the servers behind them see a total throughput of 8Gbps?


How far does this solution scale – can we trunk the 24-port 10Gbps VCs to get an effective 240Gpbs combined throughput?



Paul replied:




While LACP does make a large bandwidth pipe from the VC module to the upstream switch, the standard dictates that traffic from any given server is restricted to a single uplink.  If that uplink should fail, then that server’s traffic would failover to another uplink in the LACP group.  SO, from a particular server’s perspective, LACP is a availability play, not a bandwidth play.



Ken also got into the discussion:



Adding to what Paul said…  in aggregate across multiple servers connected to the VC module, you will be able to take full advantage of the bandwidth provided by the LACP group.



Any additional comments regarding this subject? Let us know your thoughts.