In other words paenevirtualization achieves name

This preview shows page 5 - 7 out of 13 pages.

In other words, paenevirtualization achieves name- space isolation by layering a set of filters on top of the underlying logical resources, and it achieves re- source isolation by layering a set of limits on top the underlying physical resource schedulers and al- locators. In effect, this paper reports the differences between duplicating and hiding various components of the OS. 3 Performance This section compares the performance of a paravir- tualized system (Xen + XenoLinux) and a paenevir- tualized Linux Kernel (PLK). As a reference point, we run many of the benchmarks used to evaluate those systems also on a stock Linux kernel. We re- peat a number of the single VM benchmarks as de- scribed in Barham et. al. [1] and Clark et. al [3], which illustrate the base performance of these two systems. These experiments run within a single VM and exercise the whole system to characterize the relative performance of the two approaches for a range of server-type workloads. Our Xen configuration consists of Xen 2.0.7 with XenoLinux based on a patched version of the vanilla 2.6.12 kernel from kernel.org in both the host (dom0) and guest domain (domU) for which results have not yet been reported elsewhere. Un- less otherwise stated, the results reported for Xen are for a single guest domain. Our PLK config- uration consists of a Linux 2.6.12 kernel patched with changes from FC4-1398, Vserver 2.0.1 and a new CPU scheduler developed at Princeton. To account for the difference between different vari- ations of 2.6.12 used by XenoLinux and PLK, we also present the base performance numbers of stock FC4-1398 2.6.12 Linux kernels. The Linux dis- tribution populating the filesystem is the latest re- lease of Fedora Core 2, thereby letting us focus on comparing the architectural difference between para- and paene- virtualization. Where possible, we use the exact same filesystem to avoid variances due to using different portions of the disk. All experiments are evaluated on an Intel SE7505VB2 motherboard with a 3.06Ghz Xeon uniprocessor, 4GB RAM, Intel E1000 GigE Ether- net, and a single Seagate Barracuda ST3200822A 200GB 7.2k RPM ATA-100 IDE disk. The Xeon processor has a 512KB L2 and 1MB L3 cache and 5
Image of page 5

Subscribe to view the full document.

Linux PLK Xeno Config FC4 Linux select TCP 5.91 5.92 4.60 mmap latency ( 32MB) 425.00 428.00 1300.00 mmap latency ( 64MB) 846.00 844.00 2575.00 fork process 127.57 130.90 378.08 exec process 462.58 488.23 697.42 sh process 1565.57 1711.90 1741.60 CS ( 2p/ 0K) 1.76 1.87 3.12 CS ( 2p/16K) 1.85 1.96 3.12 CS ( 2p/64K) 2.07 2.09 3.29 CS ( 8p/16K) 2.37 2.52 3.92 CS ( 8p/64K) 11.24 11.87 14.07 CS (16p/16K) 3.31 3.41 4.88 CS (16p/64K) 23.65 26.35 28.10 prot fault 0.68 0.69 1.07 page fault 2.31 2.35 4.00 Table 1: lmbench latency OS benchmarks - times in μ s hyperthreading disabled. To compare performance between Xen 1.0 and Xen 2.0, we also repeat the base performance evaluation on the identical hard- ware used by Clark et. al. to reproduce the results reported by Barham et. al. 3.1 Base Operating Systems Benchmarks We ran the development benchmark subset of McVoy’s lmbench [11] version 3.0-a3 tool, as those experiements target particular OS subsystems.
Image of page 6
Image of page 7
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