Wahoo! By now, and without a definitive confirmation we have a 'tango device server devices record'. I think this sentence is not much clear. Tango is a distributed systems who controls some synchrotrons, and like in unix (where everything is a file) in tango everything is a device. One device is one agent in this distributed systems, and this agents are running on a machine where the main process of this is the device server.
Normally, a tango developer has some devices controlled by a devices server, and our logics guides us to, when you have many devices split this between more instances of this device server. So, if you have ccd camera devices, like 50; and you likes to control them, normally you creates at least a device server for each sector, and this device server controls all the cameras in an specific sector. Easy (I think).
But what happens when you have a pool of devices to control a section of this machine (in this practical case the linac to booster transferline [LTB]) and you don't wanna a split this? Then we arrive to this situation where the pool device server has to manage 240 devices only in the first segment of the ltb.
I need to confirm if in the vacuum control system we don't have something in this level. Waiting for this confirmation, we are able to say (at low voice) that we have a record!
Actualization: This record has been overcome! We realize that the pool was consuming too much cpu when this large number of devices are not doing anything. Known this, a deadlock was solved and a stress test creates 1000 devices with non cpu consumption when they are free.
No comments:
Post a Comment