Page 2 of 3 FirstFirst 123 LastLast
Results 11 to 20 of 22

Thread: Need to update Cortex PC ...

  1. #11
    Automated Home Ninja Viv's Avatar
    Join Date
    Dec 2004
    Posts
    284

    Default

    Quote Originally Posted by chris_j_hunter View Post
    ah, thanks ...

    module polling intervals - would upping the interval a little, from the standard five minutes, have a significant effect on CPU loading ?
    The polling period whilst defaulting to 5 minutes is randomised in regards when it starts so that requests for module status are spread out to reduce peak traffic.
    Also should a module report its status as a consequence of a change of state e.g. motion sensor triggers or door opens then the polling period is reset.

    To a certain extent the polling period just re-affirms that the module is alive an well. So increasing the polling period will not normally have an adverse effect of behaviour or logging information. Unfortunatlly their is not a global change all module polling period option and I believe you have many modules.

    Do not however increase the polling period of digital inputs used for pulse counting.

    Viv.

  2. #12
    Automated Home Legend chris_j_hunter's Avatar
    Join Date
    Dec 2007
    Location
    North Lancashire
    Posts
    1,670

    Default

    ah, thanks ...

    so, what to try ?

    given most Rooms are unoccupied at any given time, and temperatures & humidities will vary slowly, except where the Sun has fairly direct influence, then maybe it's possible to roughly estimate what proportion of CPU load might be saved by extending (default) polling times ...

    assuming network nodes are the only ones polled (?) - about 70% of the total in our case - then a reasoned guess might be that something like 10% of polled nodes could be reporting changes in any whole cycle (perhaps half that, maybe double that) ...

    ie: doubling the (default) polling interval, from five to ten minutes, might reduce the polling load on the CPU load by something like 40% ...

    non-polled nodes (Macros, the Scheduler, etc) will also produce changes, of course, and changes will bring-on other activities ... so, at a guess, the polling load on the CPU will be about half of the total - ie: doubling the default polling interval might save 20% of total CPU load ...

    hmm, it's tempting to try a bit more - say, an increase from five to fifteen minutes ...

    will also check that retries are still limited to two (when a module - or is it node ? - fails to respond) - in-case the setting reverted when we (recently) updated to latest version of Cortex ...

    Chris

  3. #13
    Automated Home Legend chris_j_hunter's Avatar
    Join Date
    Dec 2007
    Location
    North Lancashire
    Posts
    1,670

    Default

    OK, gave it a go - upping the (default) polling interval from 5 to 15 minutes, for 70% of active nodes (several of the rest involve pulse-counting, and 70% seemed enough to show a difference) ...

    traces before (upper) & after (lower) attached ...

    Untitled.jpg

    so - seems to have made next-to-no difference ...

    ‘twas worth a try - but will now revert to yesterday’s database !

    Chris
    Last edited by chris_j_hunter; 9th May 2013 at 10:17 PM.

  4. #14
    Automated Home Jr Member
    Join Date
    Jun 2008
    Posts
    30

    Default

    Hi Chris,
    i recently experienced problems with high CPU utilization. Problem was related to defibrillator proces, and with Karam advice I changed API version: go to Cortex Tools and Select Cortex API, then change Default API to Version 1 and enable suppress logging 'Pulse ... ' if not already enabled. Then close this menu. Then Close and Restart Defibrillator application.
    Maybe it will help to you too

  5. #15
    Automated Home Legend chris_j_hunter's Avatar
    Join Date
    Dec 2007
    Location
    North Lancashire
    Posts
    1,670

    Default

    many thanks - did check that, and we were already set to Version 1 & the box was already ticked ...

    OTOH, our experiment with increased (default) polling intervals might suggest something's not quite as it should-be ...

    ie: we upped the interval for quite a bit more than a few modules, so it was odd to see just-about zilch effect !
    Last edited by chris_j_hunter; 13th May 2013 at 01:03 PM.

  6. #16
    Automated Home Legend Karam's Avatar
    Join Date
    Mar 2005
    Posts
    817

    Default

    Just to put things in perspective I think this might be one of the largest installations coupled with the smallest computing platform . My last information was around 250 modules totalling 1088 individual networked objects and over 1500 serviced objects in all when macros and such included. Like Vivian suggested I think the polling period will probably have little impact -already starting off with a base line of a relatively long few minutes. I am aware that you will potentially have camera handling in the near future and I can tell you that there will definitely be an additional impact from these..

    A couple of things you can test: stop the network and check CPU % in this state. It should be pretty near zero but if not might suggest that the 'idle' camera objects in your database are actually not being so idle. To check this you can temporarily delete camera objects in structure view eg. on AUIs and IVC digitiser object (don't save the database after such a deletion!) and see then if the CPU usage drops. Also from this position you can visit the Tools | Network supervisor | Service Tools and Temporarily disable all network nodes. Now hit the network run button and your network will run but without communicating with any physical modules - you can think of it as simulation mode. What is the CPU usage now? You probably need to try with Cortex minimised and expanded to see if any difference in each case as I think you suggested on your set up this has a significant effect.

  7. #17
    Automated Home Legend chris_j_hunter's Avatar
    Join Date
    Dec 2007
    Location
    North Lancashire
    Posts
    1,670

    Default

    stopping network, dropped CPU usage to a few % ...

    running it & then deleting both video windows gave around a 15% reduction ...

    deleting both cameras, as well, gave a further 15% reduction ...

    meaning it was then running at around 60% ...

    so then minimising the Cortex window would presumably have brought it below 50% ...

    temporarily disabling network nodes reduced the level to around 60%, or 15%, it seemed to vary, each time, and after a while, and re-enabling it seemed to need several attempts ... not sure why (we kept the network running all-through) ...

    NB: we've not yet connected the cameras to the IVC card - so no actual images being processed (video windows are plain blue) ...

  8. #18
    Automated Home Legend Karam's Avatar
    Join Date
    Mar 2005
    Posts
    817

    Default

    As a general comment, deleting objects and disabling/enabling network nodes is better done when the network is stopped. In the latter case if you disable the network nodes whilst the network is running it still leaves all the nodes in an initialised state i.e they will still report signal changes even though these may not then result in any output actions, whereas doing so before network start means they will be left in a 'quiet' state since Cortex de-initialises upon network shutdown (if the nodes are not disabled prior to the shutdown). So I guess disable network nodes might be better described as 'do not send messages to network nodes'.

    I think even with no cameras attached elements of the video components will be taking up resources hence my suggesting that you delete the camera objects altogether just to clarify things a bit.
    Last edited by Karam; 17th May 2013 at 09:52 AM. Reason: typo

  9. #19
    Automated Home Legend chris_j_hunter's Avatar
    Join Date
    Dec 2007
    Location
    North Lancashire
    Posts
    1,670

    Default

    interesting - thanks to the the insight gained from doing these experiments, and the subsequent feedback, little by little we're gradually building our mental picture of how it all works - 'though, it has to be said, we live in fear of corrupting out database (it's always a relief when we get back to the saved version) !

    NB: we did fully delete the camera objects ...
    Last edited by chris_j_hunter; 16th May 2013 at 09:21 AM.

  10. #20
    Automated Home Legend Karam's Avatar
    Join Date
    Mar 2005
    Posts
    817

    Default

    Yes of course its always a good idea to manually back up files you have laboured on , but also Cortex will make an automatic backup before it saves a database. I suppose this doesn't prevent you from making two successive saves of a 'bad' database.

    Although Cortex will prompt you to save a database before exit if it thinks changes have been made, the save must be by a deliberate user action or confirmation (though the backup creation is then automatic). So if you are in doubt about the changes that may have occured during a session then just exit Cortex without saving.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •