4 Replies Latest reply on Feb 28, 2016 4:58 PM by bigtom

    Two machine deployment questions

    bigtom

      I have been looking at testing a two machine deployment. I have some questions though. I could not find a lot of information on this.

       

      I am assuming CWP is run on the Worker. Is it? All the documentation is focused around WebDirect.

       

      What are the effects of the location of each machine. LocalDB+RemoteWeb, RemoteDB+RemoteWeb, RemoteDB+LocalWeb. how does the distance/latency affect performance. Any advantage to VPN with remotes?

       

      Why did they stop at only one WebServer worker machine?

       

      Basically any advice or experience will be appreciated.

        • 1. Re: Two machine deployment questions
          user19752

          I think you already read this https://fmhelp.filemaker.com/docs/14/en/fms14_getting_started.pdf

          CWP is run on worker.

           

          It may be there was too few users to continue support 3 machine deployment. FMI made standby server (instead?).

          • 2. Re: Two machine deployment questions
            bigtom

            user19752 wrote:

             

            I think you already read this https://fmhelp.filemaker.com/docs/14/en/fms14_getting_started.pdf

            CWP is run on worker.

             

            It may be there was too few users to continue support 3 machine deployment. FMI made standby server (instead?).

            I see now on page 22 there is a reference. I missed that.

            • 3. Re: Two machine deployment questions
              bigtom

              I suppose the problem with the 3 machine deployment (1xDB and 2xWeb) might be the port configuration. It would be interesting to see something that made multiple web servers available so the number of connections per server could be increased or the amount of resources available to each connection could be better managed.

               

              I have found that the recommended number of connections via web direct on a single machine deployment are about spot on. I ran up 6 connections on a Win2012 2-core 4GB FMS and got the following result:

              Screen Shot 2016-02-24 at 4.02.46 PM.png

              When I pushed up to 7 connections it hit 100%

              Screen Shot 2016-02-24 at 4.16.57 PM.png

              It looks like the connections use more CPU than RAM as it scales up. Solution used for testing was the FM sample file with WD optimized layouts. Performance on OSX might be a little better. As a baseline the machine uses average of 2% CPU and 70% RAM at idle to run the OS and FMS with WebPublishing on and no connections. Adding 2 active connections runs up to 25% and 4 active hit 75%. Seven still hits 100%.

               

              These are average numbers. There is some CPU used to keep the connection going but the once the layout is rendered the CPU usage goes down. Tested was average resource monitor of users moving through layouts in the sample file.

              • 4. Re: Two machine deployment questions
                bigtom

                I am still looking for and suggestions on how latency between the DB and web servers affect the solution usability.