[Ace] Gateways for Test Lab IOCs

[Anthony Cuffe]

I am not sure what networks your devices are on but if it is the new srf networks, there is very little on them.  They are probably less crowed than their previous network.  One thing I would point out is that the PC may have gone from a 1G connection to a 100Mb connection.  I doubt this is the problem, but we could test it.  It might be the order of the CA address list.  That level of latency almost sounds like it is missing and re-trying.


+-------------------------+
| Anthony Cuffe           |
| voice  : 757 269-6213   |
| e-mail : cuffe at jlab.org |
+-------------------------+

On 08/28/2014 11:24 AM, Brian Bevins wrote:
> Hi Christiana,
>
> If the same code is behaving differently on a different network, then we
> may need to get some input from the ACE guys. There may be some reason
> why the operations networks in the Test Lab are inherently slower or
> there may be a lot of traffic on those networks. 100+ msec sounds like a
> lot of latency.
>
> --Brian
>
> On 08/28/14 11:13, Christiana Wilson wrote:
>> Hi Brian,
>> I'm going to pass this along to Tom Powers who wrote the code that is behaving badly.  The trigger we were using for a scan, when the systems were on the CUE network, were passed through very quickly (~2ms) to the IOC.  We were pointing to the gateways.
>>
>> Since I have moved the PC and the IOC over to the operations network, the triggers take a long time to trigger the scan (~100+ ms).
>> I do not think therehas been any code change.
>>
>> Tom or I need to look at the software and determine if it is doing PV gets in a tight loop rather than using a monitor.  I bet that is the problem.
>>
>> I will get back to you on this.
>> thanks for the advice,
>> Christiana
>>
>> ----- Original Message -----
>> From: "Brian Bevins" <bevins at jlab.org>
>> To: "Christiana Wilson" <grenoble at JLab.org>
>> Cc: "Matthew Bickley" <bickley at JLab.org>, "Pam Kjeldsen" <kjeldsen at JLab.org>
>> Sent: Thursday, August 28, 2014 10:15:40 AM
>> Subject: Re: Gateways for Test Lab IOCs
>>
>> Hi Christiana,
>>
>> Inserting a gateway between a CA client and server is unlikely to improve performance except in two cases. If there are many clients, pointing them all at a gateway will reduce the load on the server. If a single client is very badly behaved (such as doing PV gets in a tight loop rather than using monitors) then pointing that client at a gateway can help. Another bad behavior for a client is repeatedly trying to connect many unresolved channels. This loads down the iocs.
>>
>> Otherwise, having a gateway between a single well-behaved client and server will not help the server.
>>
>> What is the actual symptom you are seeing? Are you seeing large CA-caused CPU loads when connected to the ioc console? Does the ioc really slow down? Our just the updates received by the client machine?
>>
>> --Brian
>>
>>
>>
>> On August 28, 2014 9:47:13 AM EDT, Christiana Wilson <grenoble at jlab.org> wrote:
>>
>> Hi Brian,
>> Is it possible to set up gateways for PVs that are specific to the test lab?
>> It appears that communicating directly with the IOC is slowing down the VTA RF systems.  Specifically, the pc srfvta-805 is experiencing large delays communicating with iocpcarcfld2.
>> Christiana
>>
>
>