We have an array of client specs deployed.
Some running hardware as old as Intel 6th gen.
These older systems seem to have a real hard time running ignition, when loading in new elements it can take tens of seconds. Where has newer systems load immediately.
Looking at CPU, Memory, Network and Graphics performance. The older systems all appares to have pleantly of overhead.
Im wondering if anyone has had similar experiences?
perhaps there is hardware acceleration which is being made use of in the newer systems?
If you visit any regular 'ol website on those older systems, does it also load them slowly?
There can be many factors at play here... Perspective uses a web browser for visualization, but any scripting, device access, database functions, and more run on the Ignition gateway, so based on your description, I would venture to guess that other websites will probably also load slowly (for instance, try McMaster's website).
If that's not the case, meaning other websites/web applications load normally but Perspective is slow, I would investigate your Layer 1 between the client and the Ignition gateway, as well as gateway resources (especially memory AND how much of it is allocated to the Ignition gateway).
The last thing I would check are updates on the old machines, especially browser updates and/or Windows updates. If you're running Windows XP, for example, there is only so much you would be able to do, but still, this is worth checking.
Those are some of the first things I would check - the more information / context you add, the better! Tens of seconds to load a page is definitely unacceptable.
Unfortunately all of our hardware is separated from the internet so no option for testing typical websites.
All of our systems have their windows versions and patches maintained so no difference there.
I did some more investigation today, our current perspective instance can consume upto around 2gb of memory on the client.
The older PC is running DDR3 RAM at 1600mhz, where as the newer system is DDR4/5 at 5000mhz.
So my current thought is that with so much being held in RAM the slower RAM is causing the delay in retrieving components when opening new views
That may well be it. Did you document how much memory each PC has, the one with the slower RAM? It can be using 2 GB and only have 4 or 8 GB, and the operating system overhead (Windows...) can be high enough to where your disk is used as a cache. Now THAT could be slow.
Did you determine the 2 GB by looking at Task Manager -> Performance when running Perspective or some other way? That seems a little high, I'm going to monitor a few sessions I have open to see what an average looks like for me.
16Gb on both systems with 8Gb consumed while running. So I dont believe it would be dumping to disk but I had considered disabling the page file to see if that would help the performance.
Our perspective program has two operating modes, 1 view at 1080p or 4 views at 4k (all runnning in the 1 instance).
I observered 2Gb when running at 4k, and ~900Mb when running at 1080p
I think there's something that needs to be clarified unless I missed in when reading this thread. When you say clients, are you talking about just a client with no gateway running on it, or are these like edge devices with perspective running on the same PC as the gateway, or something different entirely?
If you're running a separate gateway and you're just comparing client only machines to each other, then I don't think gateway specs/resources/etc will matter as long as your test is an apples to apples test (meaning you don't have 10 clients with old hardware loaded on one gateway, but when you run the test with a new client there's only 1 or 2 running.) As long as everything from the gateway is the same, it could be the clients themselves. If you're not maxed out on RAM, that's probably fine, but I do wonder about the client/browser. I believe in modern browsers they try to use hardware acceleration, but are these machines running older OSs? What are the full specs on the older machines? Hard drives? Fast network connection?
Yes the gateway is hosted seperately from the client machine.
The older clients are running 4th/6th gen Intel i7 (T spec / Low power), 16gbs 1600Mhz RAM, 120GB SSD and a 1Gbs Network connection.
Performace seems to vairy which is strange, sometime fine, sometimes very slow... we dont see these slowdowns on the newer clients (12th/14th Gen). All of which are connected via the same network and gateways.
Performance seem to be similar whether we are running Perspective Workstation (8.1.42) or via Microsoft Edge.
I had been curious if it was moving non-frequently used memory to page file so have disabled that. However, that looks to have made no effect.
That's strange. As an example yesterday for a client who was questioning if they could run Ignition at 4K without a graphics card, I launched a current Vision project with live data in 4K (quad virtual 1080p screens) with a repurposed Dell Wyse thin client running a 6th gen i5 with 4GB of RAM. While it took longer to initially load, popups, screen changes, etc changed almost as fast as a modern i9 13th gen processor with 64GB of RAM on 1080p.
I suppose to my point regarding the page file, As I understand it a thin client would be running 100% off RAM.
I've disabled the page file, and it 'seems' to be preforming better... Still not at the same level as the modern hardware.
Re-enabling page file and watching the disk usage there are noticeable bumps in reads when changing views, opening pop-ups etc. so it does appear that it is moving some memory to disk, even though there is several GBs spare in RAM.
Disabling the page file entirely does seem... radical. I would like to see if I'm able to prevent specific programs from accessing the page file rather than the system wide setting
It's not really a thin client if you're running a full Windows OS. No matter how the device was marketed. A true thin client essentially has no HD at all and boots off the network. They usually have lower end processors due to the fact that they're just connecting to a server and running a remote desktop session or a a web session on a very minimal "OS". ThinManager clients are what my company typically uses as they run a customized BIOS and boot up entirely off the network. (OnLogic makes some great thin clients for ThinManager).
Are the "thin clients" you're running using a standalone Windows OS (can they boot up without a network connection), or are they truly network booting? If the latter, they should perform similarly to any other thin client since all the processing happens on the server. If they're running a full Windows OS, then it may be thin client hardware, but software wise is a thick/fat client (similar to my example above, but I was using Vision vs Perspective).
We are experiencing differing loading times between our older and new client stations (full windows machines). These are all running on the same network and pointed at the same gateways.
At this point It appears that it is primarily due to windows moving 'unused' memory to the SSD (page file) and retrieving this data from the older SSD over SATA2 is as slow as molasses.
If found that disabling the page file seems to make the loading times a lot more consistent, granted the older hardware is still not as responsive, it doesn't slow to being unusable.
