Imperva Cyber Community

communities_1.jpg
 View Only
  • 1.  Environment Sizing

    Posted 04-16-2021 12:20
    Good afternoon;

    Based on Imperva's appliance docummentation GWs capacity can be described by the amount of TPS it can run. For example, in case of GW6500, it can process 18000TPS. Taking it as a reference can we ask our database administrators to give us any data to help us to size the Imperva's environment. In case of Oracle, would the statspac be useful?


    Regards;
    #DatabaseActivityMonitoring

    ------------------------------
    Marcos Sanz
    Security Project Manager
    Telefonica Spain
    Madrid
    ------------------------------


  • 2.  RE: Environment Sizing

    Posted 04-19-2021 13:15

    General guidelines on the capacity of our On-Prem solutions can be found here https://www.imperva.com/resources/datasheets/Imperva-DS-SECURESPHERE-APPLIANCE-Web092118.pdf

    As these sizing exercises require a better understanding of the data centers, the number of servers, the number of cores, the type of the DBs to by monitored, the configuration of the GWs, among many other considerations, we strongly suggest finding a direct recommendation from our Sales Team for sizing of the solutions.

    ------------------------------
    Diego Roa
    Product Support Engineer
    Imperva Inc
    Redwood Shores CA
    ------------------------------



  • 3.  RE: Environment Sizing

    Posted 05-05-2021 14:38
    As Diego mentioned there are several factors that affect sizing.and you should work with a sales engineer to determine what is right for your environment.

    With a v6500 having a max tps of 18000, you really only want to run it at 90% or 16200 tps.  
    Theoretical load is based on cpu count times 100, so on the gw you could have 162 cores.

    Theoretical load is not accurate, I have seen a 2 core AIX server have more throughput than a 16 core Windows server.

    If you have clustered databases as active/passive then that could reduce overall load as only one is running at a time.

    In gateway clusters the total theoretical load is (# of gw's -1) * gw tps

    And all of this is affected by database query and response size and query rate.