Experimental Physics and
Industrial Control System

"Kasemir, Kay via Core-talk" <core-talk at aps.anl.gov> · Mon, 25 Oct 2021 17:14:44 +0000

> I’m not sure that arrays really get higher priority anywhere...

Some CA clients (CS-Studio) will in fact create the Channel Access channel for arrays at a _lower_ priority than scalar channels. I believe this was originally implemented for BNL to lessen IOC CPU load.

>IIRC the PVA protocol and gateway were designed to be a lot more efficient at handling large amounts of array data, which is why I would recommend using PVA for areaDetector transport.

The numerous CA gateways used on SNS beamline and the concentrator gateway that Steven just mentioned limit the EPICS_CA_MAX_ARRAY_BYTES because otherwise one remote user looking at a large array (image) would halt updates for everybody else's screen.   

The PVA implementations try to offer "fair scheduling" which Michael once described as trading throughput in favor of latency, ie. throughput may be reduced, but latency should also be reduced. At this time I don't have operational experience with it, but a first test of the latest PVA gateway, https://mdavidsaver.github.io/p4p/gw.html, indeed passes area detector images from the sim detector. Unclear how to quantify, but a basic test of looking at scalars via the PVA gateway and then opening an area detector image display didn't result in the same severe delays of the scalars in the way usually seen with the CA gateway.

-Kay

Experimental Physics and Industrial Control System

Experimental Physics and
Industrial Control System