VirtualBox

Ignore:
Timestamp:
Feb 19, 2015 2:28:29 PM (10 years ago)
Author:
vboxsync
Message:

Manual: Expanded troubleshooting chapter.

File:
1 edited

Legend:

Unmodified
Added
Removed
  • trunk/doc/manual/en_US/user_Troubleshooting.xml

    r51964 r54295  
    549549    </sect2>
    550550
     551    <sect2 id="ts_host-freq-boost">
     552      <title>Performance variation with frequency boosting</title>
     553
     554      <para>
     555          Many newer multi-core processors support some form of frequency
     556          boosting, which means that if only one core is utilized, it can run
     557          faster (possibly 50% faster or even more) than the rated CPU frequency.
     558          This causes measured performance to vary somewhat as a function of the
     559          momentary overall system load. The exact behavior depends strongly
     560          on the specific processor model.
     561      </para>
     562
     563      <para>
     564          As a consequence, benchmarking on systems which utilize frequency
     565          boosting may produce unstable and non-repeatable results, especially
     566          if benchmark runs are short (on the order of seconds). To obtain stable
     567          results, benchmarks must be run over longer periods of time and with a
     568          constant system load apart from the VM being tested.
     569      </para>
     570    </sect2>
     571
     572    <sect2 id="ts_host-freq-scaling">
     573      <title>Frequency scaling effect on CPU usage</title>
     574
     575      <para>
     576          On some hardware platforms and operating systems, CPU frequency
     577          scaling may cause CPU usage reporting to be highly misleading. This
     578          happens in situations when the host CPU load is significant but not
     579          heavy, such as 15-30% of the maximum.
     580      </para>
     581
     582      <para>
     583          Most operating systems determine CPU usage in terms of time spent,
     584          measuring for example how many nanoseconds the systems or a process
     585          was active within one second. However, in order to save energy, modern
     586          systems can significantly scale down CPU speed when the system is not
     587          fully loaded. Naturally, when the CPU is running at (for example) one
     588          half of its maximum speed, the same number of instructions will take
     589          roughly twice as long to execute compared to running at full speed.
     590      </para>
     591
     592      <para>
     593          Depending on the specific hardware and host OS, this effect can very
     594          significantly skew the CPU usage reported by the OS; the reported CPU
     595          usage can be several times higher than what it would have been had the
     596          CPU been running at full speed. The effect can be observed both on
     597          the host OS and in a guest OS.
     598      </para>
     599    </sect2>
     600
     601    <sect2 id="ts_win-cpu-usage-rept">
     602      <title>Inaccurate Windows CPU usage reporting</title>
     603
     604      <para>
     605          CPU usage reporting tools which come with Windows (Task Manager, Resource
     606          Monitor) do not take the time spent processing hardware interrupts into
     607          account. If the interrupt load is heavy (thousands of interrupts per second),
     608          CPU usage may be significantly underreported.
     609      </para>
     610
     611      <para>
     612          This problem affects Windows as both host and guest OS. Sysinternals tools
     613          (e.g. Process Explorer) do not suffer from this problem.
     614      </para>
     615    </sect2>
     616
    551617    <sect2 id="ts_host-powermgmt">
    552618      <title>Poor performance caused by host power management</title>
     
    590656  </sect1>
    591657
    592   <sect1>
     658  <sect1 id="ts_win-guests">
    593659    <title>Windows guests</title>
    594660
     
    889955  </sect1>
    890956
    891   <sect1>
     957  <sect1 id="ts_lin-x11-guests">
    892958    <title>Linux and X11 guests</title>
    893959
     
    9751041  </sect1>
    9761042
    977   <sect1>
     1043  <sect1 id="ts_sol-guests">
    9781044    <title>Solaris guests</title>
    9791045
    9801046    <sect2>
    981       <title>Older Solaris 10 releases hang in 64-bit mode</title>
     1047      <title>Older Solaris 10 releases crash in 64-bit mode</title>
    9821048
    9831049      <para>Solaris 10 releases up to and including Solaris 10 8/07 ("S10U4")
    9841050          incorrectly detect newer Intel processors produced since 2007. This
    985           problem leads to the 64-bit Solaris kernel hanging or crashing almost
    986           immediately during startup, in both virtualized and physical
    987           environments.
     1051          problem leads to the 64-bit Solaris kernel crashing (and usually causing
     1052          a triple fault) almost immediately during startup, in both virtualized
     1053          and physical environments.
    9881054      </para>
    9891055      <para>
     
    9951061   
    9961062    </sect2>
     1063
     1064    <sect2>
     1065      <title>Solaris 8 5/01 and earlier may crash on startup</title>
     1066
     1067      <para>
     1068          Solaris 2.6, 7 and 8 releases up to and including Solaris 8 4/01 ("S8U4")
     1069          incorrectly set up Machine Check Exception (MCE) MSRs on Pentium 4 and
     1070          somene later Intel CPUs. The problem leads to the Solaris kernel crashing
     1071          (and usually causing a triple fault) almost immediately during startup, in both
     1072          virtualized and physical environments. Solaris 9 and later releases are
     1073          not affected by this problem, and neither is Solaris 2.5.1 and earlier.
     1074      </para>
     1075      <para>
     1076          The recommended solution is upgrading to at least Solaris 8 7/01
     1077          ("S8U5"). Alternative solutions include applying a patch for bugs 4408508
     1078          and 4414557 (on an unaffected system).
     1079      </para>
     1080
     1081    </sect2>
    9971082  </sect1>
    9981083
    999   <sect1>
     1084  <sect1 id="ts_fbsd-guests">
     1085    <title>FreeBSD guests</title>
     1086
     1087    <sect2>
     1088        <title>FreeBSD 10.0 may hang with xHCI</title>
     1089
     1090        <para>
     1091            If xHCI (USB 3.0) emulation is enabled for FreeBSD 10.0 guests, the guest
     1092            OS will hang. This is caused by the guest OS incorrectly handling systems
     1093            where MSIs (Message Signaled Interrupts) are not used with the xHCI device.
     1094        </para>
     1095        <para>
     1096            The problem does not exist in earlier FreeBSD releases and was fixed in
     1097            FreeBSD 10.1.
     1098        </para>
     1099    </sect2>
     1100  </sect1>
     1101
     1102  <sect1 id="ts_win-hosts">
    10001103    <title>Windows hosts</title>
    10011104
     
    11481251  </sect1>
    11491252
    1150   <sect1>
     1253  <sect1 id="ts_lin-hosts">
    11511254    <title>Linux hosts</title>
    11521255
     
    13801483  </sect1>
    13811484
    1382   <sect1>
     1485  <sect1 id="ts_sol-hosts">
    13831486    <title>Solaris hosts</title>
    13841487
Note: See TracChangeset for help on using the changeset viewer.

© 2024 Oracle Support Privacy / Do Not Sell My Info Terms of Use Trademark Policy Automated Access Etiquette