3 Replies Latest reply on Sep 22, 2009 2:57 AM by Mervyn Wong

    Perfromance and customer experiences

      Hope everyone is doing well. I would appreciate it if you can help to provide any information either documented or based on your customer experience on Bladelogic infrastructure network "chatter" across:


      1) The Bladelogic server and client agents.

      2) Bladelogic to bladelogic proxy servers.

      3) Etc.


      These figures are to show a potential customer (6000 servers) the communications between the different components and when they require connection. I know this is a tough question but customer needs to identify which if any of the bladelogic components churn up comms. The other factors I have identified include:


      a) Scheduled jobs

      b) Validation of live data on servers

      c) Live view

      d) Patch/Application/Software deployment

      e) Provisioning


      These are subjective and will require data to be transfered. I just need to know if there are any administrative comms apart from the factors I have identified.


      It would be good to know the approximate customer timing information regarding:


      1) Customer patch deployment including nature of patch

      2) Application deployment, type of application

      3) Daily snapshot/baseline

      4) Provisioning

      5) etc.


      Thanks in advance, The customer will require this information by Monday next week before the customers architectural meeting.


      Mervyn Wong



        • 1. Re: Perfromance and customer experiences
          Bill Robinson

          There's not really data for this, though it gets asked alot.  It's so dependent on the environment and usage.  we can describe what does generates traffic but 'how much' is dependent on alot of variables.


          The agents are 'speak when spoken to', so if you are not running a job against a target, or live browsing it, or connecting w/ nsh, there is no traffic to the target.  There is alot of chatter between the appserver(s) and database.  Unless a job or nsh session is running there's nothing between the appserver and proxy, client and target and client and nsh proxy.


          Any traffic generated from a job is dependent on how fast the network is, how fast the target responds, and what you are doing against the target.


          The only thing missing from your list is 'nsh connections' - using the nsh client to do things against the targets.


          there are a variety of things we can do to deal w/ traffic concerns, like using repeaters for servers across a wan from the appserver to ease deploys, or setting up multiple provisioning datastores, setting up 'network share' repositories.


          typically, the customer is concerned about the appserver to agent communications, and across the wan communications.  that's where we can introduce different infrastrucuture components.  but if you are copying 200 megs of patches across the wan, it's going to take however long it takes to copy 200megs across the wan...  our agent does not add much overhead.

          • 2. Re: Perfromance and customer experiences
            young so


            I try to answer most of your questions:

            Most of communication between the server and blade tool is done 4750 by default.  There is no bandwidth control so, I would recommend that you evaluate your network speed and do some type of network level QOS to keep the network guys off your back.  However, there is no dynamic pushed being done other then the schedule jobs within Blade.  So, you should evaluate what time the jobs are getting done and any other job that are parallel like back ups.  It's just common sense on within data center environment.  Now, if you try to manage servers outside the data center, you might need to evalute each link or just do QOS on all remote links outside of the data center.

            Another thing you can do is to setup sniffer on the network do test jobs, live server audit, and base line etc.  This can be done with VM Ware and test server to get the real time stats.  Here are some software:

            ethereal (Free Sniffing tool)

            portmon.exe (System internal Tool)

            Using these tools you could provided real base line for the customer.

            • 3. Re: Perfromance and customer experiences

              Thanks Guys, that was very helpful.