Re: Going full PVA?

[Ralph Lange via Tech-talk <tech-talk at aps.anl.gov>]

Well,

From: Stainer Tom <Tom.Stainer at sckcen.be>
Date: Tuesday 8 June 2021 at 15:53
To: Timo Korhonen <Timo.Korhonen at ess.eu>, "tech-talk at aps.anl.gov" <tech-talk at aps.anl.gov>
Subject: RE: Going full PVA?

 

What I meant with CA <-> PVA translators is if some device is providing data over CA is there a way to make a proxy or gateway to wrap this and convert it to a corresponding PVA signal? As you mentioned below, avoiding vendor specific EPICS integration is always desired but sometimes unavoidable and I was thinking in these cases what can you do if you want only PVA….. 

On Tue, 8 Jun 2021 at 17:54, Timo Korhonen via Tech-talk <tech-talk at aps.anl.gov> wrote:

I guess it would be possible to create such a gateway, although I have not yet seen a case where that would be strictly necessary. I can imagine that there are such cases, we just have not seen any real blockers yet, or at least none have come to my knowledge.

For a modular approach... the existing gateway is called IOC.

Create records pointing to the "CA speaking device" and use QSRV to provide them as pvAccess.

That has the advantages of being fully compatible with all your other IOCs, moving its EPICS version with all your other IOCs, not using any specific implementations of CA client or PVA server, allowing to put additional logic into the glue layer (add health monitoring, add logging, combine things, change record names, etc.) without introducing new concepts, languages or implementations and cleanly separating that gateway functionality between different use cases.

At ITER, our practice is to put an IOC in front of any "CA speaking device", just for the sake of conformity and flexibility.

With maybe the exception of the device running an original IOC that we have full control over - OS, EPICS version, database... which I don't expect to happen. High bandwidth low latency applications - think detectors - may become the other exception.

Cheers,
~Ralph