Opened 15 years ago
Closed 9 years ago
#3834 closed defect (obsolete)
OpenSolaris guest has high system time overhead — at Version 3
| Reported by: | Scott Fehrman | Owned by: | |
|---|---|---|---|
| Component: | other | Version: | VirtualBox 2.2.0 |
| Keywords: | opensolaris tsc_read gethrtime | Cc: | |
| Guest type: | Solaris | Host type: | other |
Description (last modified by )
We created an image using OpenSolaris 2008.11 (32-bit), MySQL, OpenDS, Glassfish to support applications. After the image boots, the apps have started and "steady state" is reached, the image has 30-35% system time constantly being consumed. We've seen this on vbox 2.0.x, 2.1.x, and 2.2.x.
We've seen this happen on the following Host OS platforms: OpenSolaris on Sun Ultra40 (AMD) Mac OS X on MacBookPro (Intel) Linux (Ubuntu 8.10) on Toshiba Laptop (Intel) Windows XP on Acer Laptop (AMD)
I "profiled the kernel" using a dtrace script and the kernel is calling: unixtsc_read / genunixgethrtime_unscaled A LOT.
I'll attach the output of the script and the dtrace script
This might be related to bug # 894
Change History (5)
by , 15 years ago
| Attachment: | vbox_kernel_dtrace_profile.txt added |
|---|
by , 15 years ago
| Attachment: | vbox_kernel_dtrace_profile.2.txt added |
|---|
comment:1 by , 15 years ago
comment:2 by , 13 years ago
I'm seeing the same when running Solaris 11express inside Virtualbox OSE 4.0.4 on Fedora 15
comment:3 by , 9 years ago
| Description: | modified (diff) |
|---|---|
| Resolution: | → obsolete |
| Status: | new → closed |


Here is the dtrace script ...
#!/usr/sbin/dtrace -s #pragma D option aggsize=256k #pragma D option bufsize=256k BEGIN { interval = 30; } profile-997 /arg0 && curthread->t_pri != -1/ { @[stack(8)] = count(); @f[func(arg0),func(caller)]=count(); total++; } tick-1sec /--interval < 0/ { trunc(@,100); printa(@); trunc(@f,100); printa(@f); printf("total samples == %d\n",total); exit(0); }