5 Replies Latest reply on Apr 19, 2013 1:03 PM by Jim Denk

    Maintaining performance for demanding environments

    Jim Denk Wayfarer

      We are moving to a high availability scenario running on two installations of windows server 2008 R2. The spool is running from an external drive and the DB server is on a separate machine.

       

      We have large printing volumes between classes, so we see distinct spike times for printing. Moreover we have a wide range of documents and document types.

       

      We have a fairly large volume of print jobs with spool files ranging from 50K to over 2GB.

       

      We had to restart the print server this morning and clear the spool files to get the CPU to cease being pegged at 100% and get Pharos back up and running. I recognize that this is something that may happen regardless of infrastructure given that we have no control over the jobs being printed.

       

      I was wondering if anyone out there has a solution to providing or approaching 100% availibility for printing. I'm concerned that in a high availability cluster scenario, the print server will still technically be responding at 100% load, and even if I script a forced failover that the documents causing the problems will be resent.

       

      I've toyed with trying to find a manner for redirect large documents to a separate server so that the majority of the jobs remain unaffected by the few larger ones that are causing problems. Although, I'm not sure if you have to suck the performance hit on the server just to identify the document size.

       

      I'd love other peoples ideas on how to have or give the appearance of 100% uptime.

       

      Jim

        • Re: Maintaining performance for demanding environments
          Yadin Flammer Navigator

          I'm not sure why you are seeing this issue.  We have 24/7 printing and have for years.  There are 400+ users printing everything from 150 page color documents to 42"x6' full color posters.  Even during finals when the printers are backed up for hours due to the volume and size of jobs being sent, the server doesn't really break a sweat.  What you describe sounds like a system issue, or you're running on 10 year old hardware.  We use a 3 year old Dell server with no issue, and all services are on one machine, not split up between servers as you describe.

          • Re: Maintaining performance for demanding environments
            Edmund Greene Wayfarer

            I had similiar questions when installing Pharos in our environment.  I, too, work at an academic instituion.  I wanted to know what could be done about back up and redundancy and Pharos' answer was "We've never had a problem."  Not quite the answer I was looking for.  I put in a feature request for some kind of back up and/or redundancy, but that was 4 years ago.

             

            What I did in my environment was to install 2 principle servers and 2 print servers.  Then I put half my queues on one server and half on the second server.  I installed "backup" queues on each server also, so in case one server was to go down, users could print to the backup queues.  This does not handle already queued jobs but it is better than nothing.  The biggest problem with this setup is keeping everything in sync.  I have to make sure that all primary queues have backup queues on the other server.  I also have to have a way to switch the release stations in an area to point to the backup server in case of a problem.  I use Ghost for the latter.

             

            This solution is not ideal, but the closest I could get for some kind of redundancy.  I've only had to use it once so far but I was glad it was there.

             

            Ed

            • Re: Maintaining performance for demanding environments
              Scott Olswold Guide

              Jim,

               

              A monitoring service (something like BMC Patrol, although this should not imply a recommendation) to monitor services and/or TCP/UDP ports is a great
              option. If a service exceeds a utilization threshold for some period of time, you can force fail-over to the other node, which will clean up the issue on the current node.

               

              Printing is a somewhat reactive process model in most operating systems; there's very little that is predictive. I hope that this helps.

               

              Regards,

               

              Scott

              • Re: Maintaining performance for demanding environments
                Scott Perkins Wayfarer

                Jim,

                We had issues like this when we first started but we doubled the number of CPU's available and increased memory in the servers which took care of the problem.   We are also running 3 print servers to distribute the load and rarely have an issue.  We run all our servers in a VM which makes increasing system resources fairly easy and recovery is easy if anything fails.

                • Re: Maintaining performance for demanding environments
                  Jim Denk Wayfarer

                  We are currently running with 6GB of RAM and 2 processors in a VM environment.

                   

                  We have 25 physical queues coupled with 19 virtual queues. We hold jobs for three hours before they are deleted.

                   

                  Given the Pharos deployment guidelines we fall into the "medium sized" deployment range as far as I can tell.

                   

                  When I bring the clustered print server online, I will have faster CPUs, faster disk and faster RAM, all of which I hope will help.

                   

                  I think some of the bigged culprits we are seeing for the performance hit are very large documents that the students, who in an effort to reduce page costs, print with 4 pages to the sheet and the spool size for these documents skyrockets.

                   

                  Thanks for all the helpful suggests on what is working for people.

                   

                  Jim