Guest disk Speed on 10.7.3

[oxoocoffee]

I am running 10.7.3 with VB 4.1.8. My guest is Win7 x32 with following guest configuration

2048 RAM, 64M VRAM, 2 Cores, 30G HD (7.7 G free), SATA controller, Type AHCI, Port count 1, VDI file is checked as "Solid State Drive"

My physical HD is OCZ Vertex 3 240 SSD wit latest firmware available today. I am using NovaBench program (Mac and Win available) to test read and write speeds
from guest and host. On Host (mac) I am getting write speed of Write 300Mb/s, Read 495Mb/s. On Guest I am getting Write speed 82Mb. My question is why is it so low? Is this artificially limited? Is there way to increase it? I am only using Visual Studio to work on my project nothing else. All tests were done when no other programs were running repeated 3 times and avg read taken. No HD test were done at the same time (guest and Host). Is there a way to increase this speed?

Also H"Use host I/O cache" is off on SATA controller.

Disk Speed Test 2.1 shows the same values as NovaBench (both on Mac since Disk Speed Test is mac only) just to show that this is not isolated to specific implementation.


Here is info from my MBP. I also wan to point out this is on a fresh installation (about week old). TRIM is not supported but OZC firmware has something similar from what I was told my OZC tech support to compensate for it.

Also I want to point out that this drive was used with previous version of VBbox and it was quite fast. I have to revert to original SSD for other reasons. So now since 10.7.3 is out I decided to switch back to OZC SSD but with latest VBox. Host is very fast. Guest not so much...

Intel 6 Series Chipset:

Vendor: Intel
Product: 6 Series Chipset
Link Speed: 6 Gigabit
Negotiated Link Speed: 6 Gigabit
Description: AHCI Version 1.30 Supported

OCZ-VERTEX3:

Capacity: 240.06 GB (240,057,409,536 bytes)
Model: OCZ-VERTEX3
Revision: 2.150000
Serial Number: OCZ-WT6G4I4IV7IFYJ49
Native Command Queuing: Yes
Queue Depth: 32
Removable Media: No
Detachable Drive: No
BSD Name: disk0
Medium Type: Solid State
TRIM Support: No
Partition Map Type: GPT (GUID Partition Table)
S.M.A.R.T. status: Verified
Volumes:
disk0s1:
Capacity: 209.7 MB (209,715,200 bytes)
BSD Name: disk0s1
Content: EFI
Lion:
Capacity: 239.2 GB (239,197,650,944 bytes)
Available: 182.43 GB (182,433,857,536 bytes)
Writable: Yes
File System: Journaled HFS+
BSD Name: disk0s2
Mount Point: /
Content: Apple_HFS
Recovery HD:
Capacity: 650 MB (650,002,432 bytes)
BSD Name: disk0s3
Content: Apple_Boot

[oxoocoffee]

Here are my log files

[oxoocoffee]

Here ozone speed test executed under cents 6.2 (Guest) (2.6.32-220.4.2.el6.x86_64 #1 SMP Tue Feb 14 04:00:16 GMT 2012 x86_64 x86_64 x86_64 GNU/Linux)

# ./iozone -t 2 -s 4G -r 4k -H 32 -T -I
Iozone: Performance Test of File I/O
Version $Revision: 3.398 $
Compiled for 64 bit mode.
Build: linux-ia64

Contributors:William Norcott, Don Capps, Isom Crawford, Kirby Collins
Al Slater, Scott Rhine, Mike Wisner, Ken Goss
Steve Landherr, Brad Smith, Mark Kelly, Dr. Alain CYR,
Randy Dunlap, Mark Montague, Dan Million, Gavin Brebner,
Jean-Marc Zucconi, Jeff Blomberg, Benny Halevy, Dave Boone,
Erik Habbinga, Kris Strecker, Walter Wong, Joshua Root,
Fabrice Bacchella, Zhenghua Xue, Qin Li, Darren Sawyer.
Ben England.

Run began: Wed Feb 15 10:22:49 2012

File size set to 4194304 KB
Record Size 4 KB
POSIX async I/O (with bcopy). Depth 32
O_DIRECT feature enabled
Command line used: ./iozone -t 2 -s 4G -r 4k -H 32 -T -I
Output is in Kbytes/sec
Time Resolution = 0.000005 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
Throughput test with 2 threads
Each thread writes a 4194304 Kbyte file in 4 Kbyte records
aio_error 5: ret 28 0
fd 4 offset 1649160192 size 4096

[mpack]

Benchmarking inside a guest can be fraught with problems, given that the time references are virtual. People have done benchmarks and come up with totally ridiculous numbers, e.g. performance figures far above or far below what the host is capable of.

VMs may be assumed to be somewhat slower than native. There is no need to test to prove that. That's the obvious price you pay for sharing one set of physical hardware devices among two or more PCs. In order for your performance tests to be taken seriously you would have to show something extraordinary beyond that, done with a credible timing reference and credible tools. And prove that it isn't just because your host is busy with other work.

[oxoocoffee]

Run began: Wed Feb 15 11:08:07 2012

File size set to 4194304 KB
Record Size 4 KB
Command line used: iozone -t 2 -s 4G -r 4k -T
Output is in Kbytes/sec
Time Resolution = 0.000005 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
Throughput test with 2 threads
Each thread writes a 4194304 Kbyte file in 4 Kbyte records

Children see throughput for 2 initial writers = 232333.84 KB/sec
Parent sees throughput for 2 initial writers = 228264.79 KB/sec
Min throughput per thread = 115679.57 KB/sec
Max throughput per thread = 116654.27 KB/sec
Avg throughput per thread = 116166.92 KB/sec
Min xfer = 4143076.00 KB

Children see throughput for 2 rewriters = 299431.58 KB/sec
Parent sees throughput for 2 rewriters = 296086.24 KB/sec
Min throughput per thread = 148221.53 KB/sec
Max throughput per thread = 151210.05 KB/sec
Avg throughput per thread = 149715.79 KB/sec
Min xfer = 4101500.00 KB

Children see throughput for 2 readers = 247631.22 KB/sec
Parent sees throughput for 2 readers = 247564.21 KB/sec
Min throughput per thread = 123783.64 KB/sec
Max throughput per thread = 123847.58 KB/sec
Avg throughput per thread = 123815.61 KB/sec
Min xfer = 4192064.00 KB

Children see throughput for 2 re-readers = 249799.02 KB/sec
Parent sees throughput for 2 re-readers = 249727.69 KB/sec
Min throughput per thread = 124853.98 KB/sec
Max throughput per thread = 124945.04 KB/sec
Avg throughput per thread = 124899.51 KB/sec
Min xfer = 4191168.00 KB

Children see throughput for 2 reverse readers = 25662.00 KB/sec
Parent sees throughput for 2 reverse readers = 25661.17 KB/sec
Min throughput per thread = 12805.26 KB/sec
Max throughput per thread = 12856.73 KB/sec
Avg throughput per thread = 12831.00 KB/sec
Min xfer = 4177512.00 KB

[oxoocoffee]

Before successful test was executed with sudo sysctl -w kern.aioprocmax=50 executed on Host

This time this test was repeated with default sudo sysctl -w kern.aioprocmax=16

Run began: Wed Feb 15 11:30:47 2012

File size set to 4194304 KB
Record Size 4 KB
Command line used: iozone -t 2 -s 4G -r 4k -T
Output is in Kbytes/sec
Time Resolution = 0.000005 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
Throughput test with 2 threads
Each thread writes a 4194304 Kbyte file in 4 Kbyte records

Children see throughput for 2 initial writers = 214407.88 KB/sec
Parent sees throughput for 2 initial writers = 197994.83 KB/sec
Min throughput per thread = 103453.63 KB/sec
Max throughput per thread = 110954.24 KB/sec
Avg throughput per thread = 107203.94 KB/sec
Min xfer = 3912960.00 KB

Children see throughput for 2 rewriters = 280899.52 KB/sec
Parent sees throughput for 2 rewriters = 278974.36 KB/sec
Min throughput per thread = 140374.17 KB/sec
Max throughput per thread = 140525.34 KB/sec
Avg throughput per thread = 140449.76 KB/sec
Min xfer = 4190908.00 KB

Children see throughput for 2 readers = 265846.28 KB/sec
Parent sees throughput for 2 readers = 265778.56 KB/sec
Min throughput per thread = 132728.98 KB/sec
Max throughput per thread = 133117.30 KB/sec
Avg throughput per thread = 132923.14 KB/sec
Min xfer = 4182080.00 KB

Children see throughput for 2 re-readers = 261654.06 KB/sec
Parent sees throughput for 2 re-readers = 261554.44 KB/sec
Min throughput per thread = 130773.38 KB/sec
Max throughput per thread = 130880.69 KB/sec
Avg throughput per thread = 130827.03 KB/sec
Min xfer = 4190912.00 KB

[oxoocoffee]

I guess some benchmarks are not as good as iozone.

thank you for your help aeichner