OMM Provider settings to avoid RSSL buffer failures

We have a high throughput provider application that from time to time disconnects after an OMM Error "Attempt to get rssl buffer failed" is triggered. From what we understand, this can occur when there are issues sending data between the provider and the ADH (network blip, capacity issues, ...).

We increased \Connections\<conn>\guaranteedOutputBuffers to 10000.

a) Apart from the information in the RFA Config Guide, is there any guidance for setting connection parameter values?
b) Should maxOutputBuffers be increased to a similar value to the guaranteed buffers?
c) What value do recvBufferSize/sendBufferSize take when not defined (what is the default size)?
d) What is the impact of setting disconnectOnGetSubmitBufFailed to false (does it block/drop message)?

We are using the RFA 8.1.0 C++ sdk.

Find more posts tagged with

trep

rfa

rssl

buffer

rfa-api

Accepted answers

umer.nalla

Hi @sbutler9

Yes, you can also increase maxOutputBuffers - values as high as 20,480 (for both this and guaranteedOutputBuffers) is not unreasonable.

The default value for the recv and send buffers will be specific your local PC/Server environment.

Setting disconnectOnGetSubmitBufFailed would result in the connection being dropped if RFA fails to get enough buffers for submitting messages.

All comments

umer.nalla

Hi @sbutler9

Yes, you can also increase maxOutputBuffers - values as high as 20,480 (for both this and guaranteedOutputBuffers) is not unreasonable.

The default value for the recv and send buffers will be specific your local PC/Server environment.

Setting disconnectOnGetSubmitBufFailed would result in the connection being dropped if RFA fails to get enough buffers for submitting messages.

sbutler9

Apart from the information in the RFA Config Guide, is there any guidance for setting connection parameter values?

umer.nalla

Hi @sbutler9

Unfortunately not - as their are too many factors at play including but not limited to things like instrument volatility, update rates, update sizes - your TREP config, your network etc etc.

It is a case of trial and error to arrive at works best for your situation.

You can also try increasing the number of guaranteedOutputBuffers and maxOutputBuffers on the ADH - it may be enough to get you past some short periods where you are publishing at higher rate than average.

You can also use the throughput and latency testing tools that come with our TREP packages to perform load testing.

EXPLORE OUR SITES