8 Replies Latest reply on Sep 17, 2012 2:49 PM by gregatkins

    Is monitoring the underlying Operating System Important?

    lundrog

      Last week I wrote on why Applications are important. This week I want to touch on the importance of Monitoring the underlying operating system, or OS.

       

      Is monitoring the Operating System Important?

       

      Yes, without an operating system our applications would not exist. I am always amazed at the lack of attention our server operating systems get.

       

      In regard to the Operating System, or OS for short, lets discuss performance trending. Ask yourself these questions about a important application’s OS in your environment.

       

      • On a daily basis how much CPU does the OS consume?
      • Does the CPU usage change during end of month or start of month?
      • Does the CPU usage change at different times of day?
      • What does Physical and Virtual Memory usage look like?
      • Does the Physical and or Virtual Memory usage change correspond with the CPU usage?
      • What about Disk, do you have any idea on the growth per volume, or disk/lun over time?

       

      My point is that with all the challenges of working in the Data Center in 2012, most of us do not have time to be in running perfmon in each server. As Virtualization rockets our server count, in most cases our Server, or System Administration Staff count says the same. However, the Business needs in many cases require things like the above to be reported on, or alerted on.

       

      Another viewpoint is as follows. Let’s say, you have a IIS server , and users are calling reporting sluggish web pages.So, being the Admin, you remotely login to the server, check for errors, finding none; check the system performance metrics via perfmon. Assuming the usage is not 100%, or you don;t know good vs bad; you're stuck either making a judgment call on what normal is, or passing the buck on to the network or storage team. However, if you can go back and look at a trend report, or even better, a graph of the resource in question, you can rule the cause out, or determine it is the cause,

       

      Here are some examples from Solarwinds Server and Application Monitor.

       

      avgresponsetime.PNGcpu max.PNGdiskpercent.PNGMemory Used.PNGtopcpubypercentload.PNG

       

      Just the ability to know what your underlying operating system calls normal is a huge step for a lot of folks in my opinion, Then if you couple in features like alerting, reporting, and trending, it is the first step into maintaining a health Data Center, and guaranteeing that our Applications have a solid foundation to live upon.

       

      Next time I will dive into Application Monitoring. Until then!

       

       

       

       

      Roger Lund

        • Re: Is monitoring the underlying Operating System Important?
          joelgarnick

          Definitely agree...if you're not monitoring the OS, you've got nothing to go on when problems come up.  Having historical information that is easy to access is absolutely essential to proper troubleshooting.  Without them you have no visibility into what normal operation is and end up just theorizing and trying things until you hit on something that makes a positive change.

          • Re: Is monitoring the underlying Operating System Important?
            byrona

            I think that monitoring the OS is just as important as the applications.  The OS as well as the network, the hypervisor and storage systems are all the building blocks that your applications run on and if they are not performing properly then it will impact your application.  If you don't monitor all of the building blocks then you have no way of sorting out where your bottlenecks are and  what is causing performance problems for your applications.

             

            We just recently had a customer that was experiencing problems with job completion times on a SQL server after virtualizing the SQL server.  It turned out that the problem didn't have anything to do with SQL at all but the OS being starved for memory by SQL and the back end network not being fast enough.  Providing more memory to the system and limiting how much SQL was allowed to use as well as upgrading to a 10Gb storage network solved the problem.  It was having both the OS and network performance data as well as the application performance data that allowed us to see the entire picture.

              • Re: Is monitoring the underlying Operating System Important?
                mdriskell

                I recently had the opposite problem.  SQL was starved as it had been limited to 20GB of ram on a 32GB system.  In looking we saw that the OS was never consuming more than 4GB so we had an additional 8GB to play with.  We allocated 6GB more to SQL (leaving 2 free for the OS) for a total of 26GB and the SQL server is performing much better.  It was just slightly starved for memory.  We saw all this by a combination of monitoring the SQL Server application with SAM and the base OS from NPM.

              • Re: Is monitoring the underlying Operating System Important?
                Sohail Bhamani

                I always evangelize about a layered monitoring approach to all of my clients.  SNMP to monitor the physical hardware is the foundation.  An OS level template is next as this is what can share about pages/sec, page file usage, swap usage and other OS type metrics.  You add an application level template next and you end up with a pretty good view of an application.  This can be pushed further for web apps by adding SEUM to act as the user layer to our monitoring onion.  Netflow kind of sits along the side of all of these layers adding the on the wire view to it all.

                 

                The big picture is what we should be after.

                 

                Sohail Bhamani

                http://www.loop1systems.com

                • Re: Is monitoring the underlying Operating System Important?
                  gregatkins

                  It is absolutely essential in my opinion, especially with all the applications I deal with at work.  It is a critical element in helping me troubleshoot issues that happen with an application, or between an application and the OS...and then toss in the the variables of the network itself or any hardware on the server.

                  • Re: Is monitoring the underlying Operating System Important?
                    chriswahl

                    It's interesting to see how many different approaches there are to monitoring the speeds and feeds of a server. Much of this is being exposed over time in the virtualization world, as the hypervisor is able to pull in more data than before (such as via VMware tools), and then exposed further via APIs. Sadly, I see a lot of shops that don't take advantage of this, and do reactive monitoring (post mortem) instead of proactive monitoring such as the tool you demonstrate here.

                    • Re: Is monitoring the underlying Operating System Important?
                      gregatkins

                      We will be doing a pretty radical change from single hardware hosts to a NetApp, Cisco Nexus, VCenter environment. The new virtual tools available from Solarwinds allow you to pull a ton of SNMP information all laid out in a nice interface.  I think virtualization is going to make IT pros take a step back and re-evaulate how and what they monitor.

                       

                      It is the same thing for testing..before we would need a seperate physical test network off our production network.  Now with virtualization, we can set up a private VM network on our production network in a half the time and with no additional hardware.