The cause for this is not well un derstood one

This preview shows page 7 - 8 out of 13 pages.

660Mbytes/sec. The cause for this is not well un- derstood. One difference between these systems is that XenoLinux uses a 100HZ clock, whereas PLK and stock Linux use a 1000HZ clock 2 . Clock in- terrupts in conventional Linux would need to be on the order of 8 μ s to account for the difference in per- formance. However, that does not account for the difference between the results for host versus guest domain results for Xen. Further investigation into these discrepancies is required. The bottom four rows listed in Table 2 do in- volve the operating system and in XenoLinux’s case the hypervisor. Here the results are mixed. There are two takeaways from these results: (1) overall the performance of XenoLinux is remarkably good compared with stock Linux and PLK, (2) yet mi- crobenchmarks seldom are indicators of overall sys- tem behavior for real workloads [4]. 3.2 Network Performance We examine TCP performance over a private Gi- gabit Ethernet LAN that is unloaded. We use a HP DL320 server with a 3.4GHz Xeon uniproces- sor system that generally is faster than our test sys- tem. This lets us measure receive and transmit per- 2 We were negligent not to eliminate this difference, but are out of time to correct for the submission. Linux PLK Xeno Config FC4 Linux rcv snd rcv snd rcv snd TCP Win 16K 524 593 524 593 348 425 TCP Win 32K 888 938 888 938 678 815 TCP Win 64K 941 941 941 941 787 867 TCP Win 128K 941 941 941 941 888 915 TCP Win 256K 941 941 941 941 902 929 Table 3: iperf performance with TCP windows sizes ranging from 16K to 256Kbytes. formance independently to a machine that can both source and sync data faster than our test system. The iperf benchmark was used to perform these mea- surements. Both sender and receiper applications were configured to use different TCP window sizes, and iperf selects the appropriate socket buffer size that is larger than the max bandwidth delay product. The results presented are a median of four tests run- ning for 10 seconds, transferring between 400MB- 1GB. Table 3 presents the results using the default MTU of 1500 bytes. Both stock FC4 Linux and PLK achieve line GigE rate with a TCP window size of 64Kbytes. In contrast, XenoLinux approaches GigE rates only with large window sizes. While this might seem unremarkable, achieving near GigE rates is a significant achievement for XenoLinux on Xen 2.0. This is because packets flow from the GigE NIC to the host domain that then forwards them to the XenoLinux guest domain. In this way the Xen 2.0 hypervisor no longer needs to replicate iptables-like forwarding support, as was done in Xen 1.0. Guest OS Virtual Devices Hypervisor ... Physical Hardware Control Interface User Software Guest OS User Software Guest OS Control Plane Software ... Physical Hardware User Software Dom0 OS Device Drivers Control Plane Software Guest OS User Software Guest OS Hypervisor Figure 2: Xen 1.0 vs. Xen 2.0 architecture.
Image of page 7

Subscribe to view the full document.

Image of page 8
You've reached the end of this preview.
  • Spring '12
  • GwangS.Jung
  • virtual machine, VMs, Linux kernel, Xen, namespace isolation

{[ snackBarMessage ]}

What students are saying

  • Left Quote Icon

    As a current student on this bumpy collegiate pathway, I stumbled upon Course Hero, where I can find study resources for nearly all my courses, get online help from tutors 24/7, and even share my old projects, papers, and lecture notes with other students.

    Student Picture

    Kiran Temple University Fox School of Business ‘17, Course Hero Intern

  • Left Quote Icon

    I cannot even describe how much Course Hero helped me this summer. It’s truly become something I can always rely on and help me. In the end, I was not only able to survive summer classes, but I was able to thrive thanks to Course Hero.

    Student Picture

    Dana University of Pennsylvania ‘17, Course Hero Intern

  • Left Quote Icon

    The ability to access any university’s resources through Course Hero proved invaluable in my case. I was behind on Tulane coursework and actually used UCLA’s materials to help me move forward and get everything together on time.

    Student Picture

    Jill Tulane University ‘16, Course Hero Intern