Diagnosing slow throughput

I'm on site in the process of replacing quite a lot of (IT) network infrastructire and moving to a couple of virtual machines to communicate with C-Gate, replacing two old PCs running C-Gate 1.5.x. The new VMs are running Win7 (32-bit) and TK 1.12.8

Part of the application that runs through C-Gate sends lots of group address ONs or OFFs in succession (~140 groups), effectively a scene just not one run through Clipsal hardware. Switching everything used to take about 3s, now it takes about 20s.

I subsequently tried switching the same groups en-masse through TK and got the same slow results.

Suggestions as to what I can tune? C-GateConfig.txt is the default auto-created one with project.start and project.default set, not other changes.

To rule things out I've just tested the application on a Windows XP VirtualBox VM running on my laptop (Win7 host) directly connected to the Clipsal switch that all the CNIs are connected to, thus ruling out the site IT network and the specific virtual environment used on site (Hyper-V), albeit the machine was still virtual.

The VMs are only for this application so I'm open to anything, eg changing TCP/IP settings.

We have a demo to the client on Wednesday so it would be really nice to know what is going on before then even if I can't fix it before the demo.

 

Tell me about it!

Might be hard to find out now but pretty sure it was the latest available 1.5 version.

Calling application is Citect (well actually C-Lution!) so it uses the CI commands rather than ON/OFF but as far as I know it does wait for responses.

However, note my comment in the original post that if I run the same test from TK it is also slow (ie multi-select the same group addresses, then send ON or OFF to them all in one go). So any issues here are presumably in C-Gate rather than the app (or in the network, or in the C-Gate config, etc), unless there's a reason to expect TK to be slow anyway (ie it's not designed/optimised for throwing a command at multiple groups quickly).

It should do, yes. Also the new Citect app also sends a whole load of new phantom group commands on startup *

I think everything is slower rather than there being any specific delays, although I'd need to run some more tests to be sure.

I'll try to collect them tomorrow when back on site. Thanks!

I'll probably only get one hit at this, so if there's anything I should probably tweak in the C-Gate config first let me know.

* As far as I can tell, C-Gate's setting to auto-create phantom groups doesn't apply to CI commands - or at least if I don't create phantom groups manually then Citect errors. Also: Group addresses that only appear in certain places (eg PIR enables) don't seem to get auto-created (so we have to create them as phantoms) and reading their status via CI commands seems to be very "inaccurate" until we've written to them.

That is with TK 1.12.8.