4 Replies Latest reply on Mar 31, 2010 2:48 PM by chris.lapoint

    Split database load for NPM/Netflow

    AndyBB

      We primarily use NPM for alerting and event monitoring of around ~1500 devices and have been approached about adding netflow analysis for an additional ~900 devices.

      We are concerned about the additional load placed on the database server, primarily in terms of I/O and finding it difficult to scope what hardware upgrades will be required, and how well it will scale with such a high number of devices. It is extremely important that it doesn't negatively impact the primary purpose of our solarwinds which is the alert and event monitoring. 

      Ideally we'd like to setup a 2nd database server that we could use solely for netflow to ensure it does not impact performance of our current setup. The only other option would be another fresh Solarwinds install but the cost of this and ongoing maintenance of two installs isn't ideal.

      Are there any plans of the horizon for allowing segregation of database load across different servers for different products ala Netflow? It would definitely help when scoping/planning for growth in large networks.

        • Re: Split database load for NPM/Netflow

          Hi Andy--

          I've marked this for the product manager to see about adding to the feature request list.

          Thanks,

          M

          • Re: Split database load for NPM/Netflow
            chris.lapoint

            We are concerned about the additional load placed on the database server, primarily in terms of I/O and finding it difficult to scope what hardware upgrades will be required, and how well it will scale with such a high number of devices. It is extremely important that it doesn't negatively impact the primary purpose of our solarwinds which is the alert and event monitoring. 

            The number of devices isn't as important as knowing the number of flows per second.   The best way to determine this is to install Orion NTA, export NetFlow to the machine, and review the Windows performance counters.   It's not necessary to have NTA actually collecting/storing the flows.

            Here's a reference NTA implementation from a customer who's been able to see 40,000 flows per second (60,000 flows per second in peak).

            Orion Server
            Intel Xeon Dual Quad Core @ 2.00GHz
            4.00 GB of RAM
            Disks – DELL PERC 5/I
             6 136GB 15K – RAID 5
            8706 elements
            279 NetFlow interfaces (spread across 6 devices)

            Poller 1 – Same HW as Main
            9309 elements
            Poller 2 – Same HW as Main
            10893 elements

            SQL Server
            Server 2008/SQL Standard 2005
            Intel Xeon Dual Quad Core @ 2.50GHZ
            8.00 GB of RAM
            Disks  -DELL PERC 6/i
            2 300GB 15K – RAID 1 (Operating System & DB Log)
            4 300GB 15K – RAID 10 (DB files)

            Are there any plans of the horizon for allowing segregation of database load across different servers for different products ala Netflow? It would definitely help when scoping/planning for growth in large networks.

            This is something we're actively exploring.   In parallel to this longer-term architectural change, we continue to make significant performance improvements in each release to mitigate the I/O load on the primary Orion DB.  Just as an example, we added Top Talker Optimization in NTA 3.6.