locked
Squid proxy server (redhat) latency issues? RRS feed

  • Question

  • I need to deploy a proxy server on a network servicing about 200 machines.

    If I virtualize redhat Linux to use squid, will the latency be too much?  Bear in mind that the underlying OS and hardware (server 2008 on RAID-5 and 2GB of RAM) must also act as a WDS server.  If it weren't for the necessity of WDS I would make the whole machine a RHL box.

    My boss thinks that there will be too much latency... I on the other hand don't have the experience to know or the time to build a prototype for testing it. 

    • Moved by BrianEhMVP Wednesday, April 23, 2014 3:34 PM
    Wednesday, April 23, 2014 7:16 AM

Answers

  • I would doubt that the latency will be an issue.  Almost every kind of workload is being virtualized these days. Hypervisors in general, and Hyper-V in particular, offer synthetic devices such as NICs and storage controllers along with custom device devices for those synthetic devices specifically so that performance and latency can get pretty close to what you would have with physical devices.  Virtual appliances for various network functions such as proxies, firewall, and load balancers, have been created specifically to run in virtual environments, and they are achieving the necessary performance and latency.  I'm sure there might be specific situations that are particularly sensitive to latency and that could be a problem running virtual, but all the mainstream scenarios are pretty much OK.

    Michael Kelley, Lead Program Manager, Open Source Technology Center

    Monday, April 28, 2014 4:18 PM
    Moderator

All replies

  • Hello, I think it would be tough to answer this question as it would depend upon your environment and hardware. You will have to try and see. The only known problem you might face is that we do not support Linux virtual machines on Server 2008. Linux virtual machines are only supported on Server 2008 R2 and newer releases.

    Please let me know how we can help further.

    Thanks,

    Abhishek

    Friday, April 25, 2014 3:03 AM
  • My question pertains to the latency introduced by the hyper-vs virtual network interface from the instance of the virtual machine to the physical network itself.  The actual virtual OS is irrelevant, for instance, I could have said that I was using a microsoft OS with some kind of proxy software and my question would have been the same.
    Sunday, April 27, 2014 1:06 AM
  • I would doubt that the latency will be an issue.  Almost every kind of workload is being virtualized these days. Hypervisors in general, and Hyper-V in particular, offer synthetic devices such as NICs and storage controllers along with custom device devices for those synthetic devices specifically so that performance and latency can get pretty close to what you would have with physical devices.  Virtual appliances for various network functions such as proxies, firewall, and load balancers, have been created specifically to run in virtual environments, and they are achieving the necessary performance and latency.  I'm sure there might be specific situations that are particularly sensitive to latency and that could be a problem running virtual, but all the mainstream scenarios are pretty much OK.

    Michael Kelley, Lead Program Manager, Open Source Technology Center

    Monday, April 28, 2014 4:18 PM
    Moderator
  • ~~

     the underlying OS and hardware (  . . . on RAID-5 . . .)   =} latency issues

    ~~

       Use RAID-10

    Wednesday, May 7, 2014 9:33 AM
  • ~~


     . . . the underlying OS and hardware (  . . . and 2GB of RAM)

    ~~

     2GB -- is near to _absolutely_ minimum

    Please, add RAM

    Wednesday, May 7, 2014 9:36 AM