Skip to contentSkip to conversation

coLinux benchmarks

DISCUSSION Dan Aloni 5 participants 8 messages Apr 5, 2004 – Apr 6, 2004

lore
Share & tools
lore

Dan Aloni

author
#0 lore raw

Hello,

Today I ran some dbench2 benchmarks in order to test coLinux's
(http://www.colinux.org) virtual disk I/O performance.

I'm cross-posting this message to the LKML, as I know that on that list there are some benchmarking experts or other people who may find this interesting.

This is the output from a coLinux 2.4.25 guest VM configured with 128MB RAM running on a Linux 2.6.3 (BK) host that has a total of 256MB RAM. The host machine has a Mobile Intel Celeron CPU (2.20GHz). All filesystems used are ext3.

    colinux:/home/dax# dbench 5 -s -S
    5 clients started
       0     62477  10.01 MB/sec
    Throughput 10.0026 MB/sec (sync open) (sync dirs) 5 procs
    colinux:/home/dax# dbench 5 -s -S
    5 clients started
       0     62477  10.43 MB/sec
    Throughput 10.4262 MB/sec (sync open) (sync dirs) 5 procs
    colinux:/home/dax# dbench 5 -s -S
    5 clients started
       0     62477  10.90 MB/sec
    Throughput 10.8926 MB/sec (sync open) (sync dirs) 5 procs

I then ran the same thing on the host itself, *without* the coLinux VM running in the background:

    hostile17:~/colinux# dbench 5 -s -S
    5 clients started
       0     62477  5.08 MB/sec
    Throughput 5.07573 MB/sec (sync open) (sync dirs) 5 procs
    hostile17:~/colinux# dbench 5 -s -S
    5 clients started
       0     62477  5.13 MB/sec
    Throughput 5.12705 MB/sec (sync open) (sync dirs) 5 procs

The VM shows better results than the host. What gives? Perhaps it is because of the combination of the host and guest's buffer cache? I'd like to know about more percise benchmarking methods for VMs.

Signature · 3 lines
Dan Aloni
Cooperative Linux, lead developer
[email protected]
#1 lore raw
Quoted 4 lines

The VM shows better results than the host. What gives? Perhaps it is because of the combination of the host and guest's buffer cache? I'd like to know about more percise benchmarking methods for VMs.

How are the virtual disks for the VM implemented? If you're doing direct I/O these numbers are indeed strange. If not OTOH that's expected because even synchronous I/O in the guest is actually async which makes it a lot faster.

author
#2 lore raw
Quoted 9 lines · Christoph Hellwig
Quoted 4 lines

The VM shows better results than the host. What gives? Perhaps it is because of the combination of the host and guest's buffer cache? I'd like to know about more percise benchmarking methods for VMs.

How are the virtual disks for the VM implemented? If you're doing direct I/O these numbers are indeed strange. If not OTOH that's expected because even synchronous I/O in the guest is actually async which makes it a lot faster.

The virtual block device driver in coLinux, named cobd, is synchronous with the host OS highest level read()/write() functions, which means e.g. for a READ block I/O request in the guest, filp->f_op->read() is called on an open 'struct file' in the host. If the call blocks, the entire guest VM blocks on it.

So, according to this, any type of I/O in the guest means synchronous I/O in the host unless the data is already in the guest's buffer cache.

It's not really the implementation I am planning to stick to, but it sure was very easy to implement.

BTW, the block device on the host side can be a file or any device that exposes read()/write() interfaces to userspace. In this benchmarking case it is a 3GB file that hosts an image of an ext3 filesystem.

Signature · 2 lines
Dan Aloni
[email protected]
#3 lore raw
In reply to Dan Aloni

Hey, I stumbled upon this when checking my mail :)

I think the reason may be that Windows is using the disks better and making access faster. Perhaps DMA acceleration or some other feature is turned off on the Linux host side, making disk access slower on the Linux side.

Then again, that could only explain the first run of the benchmark and not the cached runs. Maybe some bug in Linux's caching/buffering?

Quoted 57 lines · --- Dan Aloni

Hello,

Today I ran some dbench2 benchmarks in order to test coLinux's
(http://www.colinux.org) virtual disk I/O performance.

I'm cross-posting this message to the LKML, as I
know that on that
list there are some benchmarking experts or other
people who may
find this interesting.

This is the output from a coLinux 2.4.25 guest VM
configured with
128MB RAM running on a Linux 2.6.3 (BK) host that
has a total of
256MB RAM. The host machine has a Mobile Intel
Celeron CPU (2.20GHz).
All filesystems used are ext3.

    colinux:/home/dax# dbench 5 -s -S
    5 clients started
       0     62477  10.01 MB/sec
    Throughput 10.0026 MB/sec (sync open) (sync
dirs) 5 procs
    colinux:/home/dax# dbench 5 -s -S
    5 clients started
       0     62477  10.43 MB/sec
    Throughput 10.4262 MB/sec (sync open) (sync
dirs) 5 procs
    colinux:/home/dax# dbench 5 -s -S
    5 clients started
       0     62477  10.90 MB/sec
    Throughput 10.8926 MB/sec (sync open) (sync
dirs) 5 procs

I then ran the same thing on the host itself,
*without* the
coLinux VM running in the background:

    hostile17:~/colinux# dbench 5 -s -S
    5 clients started
       0     62477  5.08 MB/sec
    Throughput 5.07573 MB/sec (sync open) (sync
dirs) 5 procs
    hostile17:~/colinux# dbench 5 -s -S
    5 clients started
       0     62477  5.13 MB/sec
    Throughput 5.12705 MB/sec (sync open) (sync
dirs) 5 procs
Quoted 15 lines

The VM shows better results than the host. What
gives? Perhaps
it is because of the combination of the host and
guest's buffer
cache? I'd like to know about more percise
benchmarking methods
for VMs.

Signature · 3 lines
Dan Aloni
Cooperative Linux, lead developer
[email protected]

-------------------------------------------------------

Quoted 7 lines

This SF.Net email is sponsored by: IBM Linux
Tutorials
Free Linux tutorial presented by Daniel Robbins,
President and CEO of
GenToo technologies. Learn everything from
fundamentals to system

administration.http://ads.osdn.com/?ad_id=1470&alloc_id=3638&op=click

Quoted 4 lines

_______________________________________________
coLinux-devel mailing list
[email protected]

https://lists.sourceforge.net/lists/listinfo/colinux-devel

__________________________________
Do you Yahoo!?
Yahoo! Small Business $15K Web Design Giveaway
http://promotions.yahoo.com/design_giveaway/

author
#4 lore raw
Quoted 5 lines · Eyal Lotem

I think the reason may be that Windows is using the disks better and making access faster. Perhaps DMA acceleration or some other feature is turned off on the Linux host side, making disk access slower on the Linux side.

No Windows was involved with these benchmarks in any way. I ran coLinux on Linux.

Signature · 2 lines
Dan Aloni
[email protected]
#5 lore raw
Quoted 10 lines · Dan Aloni
Quoted 5 lines · Eyal Lotem

I think the reason may be that Windows is using the disks better and making access faster. Perhaps DMA acceleration or some other feature is turned off on the Linux host side, making disk access slower on the Linux side.

No Windows was involved with these benchmarks in any way. I ran coLinux on Linux.

You ran coLinux on a Linux host? Perhaps I've missed something on the list.. is there a native Linux kernel port now? An alternative to User Mode Linux is a rather big thing for me.

- Ian C. Blenke <[email protected]>

author
#6 lore raw
Quoted 9 lines · Ian C. Blenke
Quoted 8 lines · Dan Aloni
Quoted 5 lines · Eyal Lotem

I think the reason may be that Windows is using the disks better and making access faster. Perhaps DMA acceleration or some other feature is turned off on the Linux host side, making disk access slower on the Linux side.

Quoted 6 lines
Quoted 2 lines

No Windows was involved with these benchmarks in any way. I ran coLinux on Linux.

You ran coLinux on a Linux host? Perhaps I've missed something on the list.. is there a native Linux kernel port now? An alternative to User Mode Linux is a rather big thing for me.

Yes, it's an alternative to User Mode Linux, thought it's a bit early and doesn't have all the wide range of support tools and nifty stuff that UML has.

Signature · 2 lines
Dan Aloni
[email protected]
#7 lore raw

ReactOS and CoLinux status

In reply to Dan Aloni

Hello All,

If you want to run Linux and have your Windows device drivers then this message is for you.

I just wanted to give all of you a update on our work to get CoLinux working on ReactOS. Where it stands right now it looks like ReactOS will be able to run CoLinux quite soon. We still have some issues with running Cygwin applications so we need some help either fixing that or we need to see if the CoLinux deamon can be built linking to msvcrt rather than cygwin. Other than this the only show stopper we have atm is our TCP/IP stack is still under heavy development and will not be ready to run the Xserver and Linux network applications for another few months.

Thanks
Steven

__________________________________
Do you Yahoo!?
Yahoo! Small Business $15K Web Design Giveaway
http://promotions.yahoo.com/design_giveaway/