Score:0

Does Throughput calculated adds up using multiple (all active) physical adaptor

pk flag

I had a network adaptor of speed 20Gbps. By stressing this adaptor using one of benchmarking tools (netperf is the tool which I used), I was able to get ~18Gbps which is well accepted.

Now, after installing one more adaptor of the same speed i.e. 20Gbps, and stressing both adaptors using the benchmarking tool, I was expecting ~36 to ~38Gbps. But still able to get the same ~18Gbps.

Does this mean, throughput gets calculated by considering average traffic on both physical adaptors and doesn't sum the values & always doesn't cross the boundary of a single adaptor link speed?

PS: The setup looks something like this. Total of 32 netperf sessions with sufficiently large socket buffer size and message sizes in send/receive calls which are 256K and 16K respectively. Each adaptor being used by 16 sessions simultaneously.

John Mahowald avatar
cn flag
Are the links aggregated and if so how, or are they different interfaces with different IPs and you are putting load on both?
Score:1
jp flag

From the Netperf Manual:

3. The Design of Netperf

Netperf is designed around a basic client-server model. – –

Once the control connection is up and the configuration information has been passed, a separate “data” connection will be opened for the measurement itself using the API's and protocols appropriate for the specified test.

As it is testing the throughput between those client and server it depends on that connection. In order to test the two adapters together they should be load balanced in a way that utilizes both adapters for a single connection. That would probably not be a very useful configuration for anything else but this benchmarking.

Chandan Hegde avatar
pk flag
the configuration I am testing is for benchmarking and I have edited my question post w/ setup detail. Total I am using 32 netperf sessions and each adoptor is stressed by 16 sessions simultaneously.
mangohost

Post an answer

Most people don’t grasp that asking a lot of questions unlocks learning and improves interpersonal bonding. In Alison’s studies, for example, though people could accurately recall how many questions had been asked in their conversations, they didn’t intuit the link between questions and liking. Across four studies, in which participants were engaged in conversations themselves or read transcripts of others’ conversations, people tended not to realize that question asking would influence—or had influenced—the level of amity between the conversationalists.