After reading Eric Sloof’s recent post about the performance he received from his Intel X25-M solid state drive (SSD) attached to his VMware ESX lab server I just had to try out my newly purchased Samsung MMCRE64G5MXP-0VB SSD in my own ESXi lab server to see how it compared with his Intel SSD’s benchmark results. So rather than install Windows 7 on my SSD for my main home PC for which it was intended I installed it into one of my ESX lab servers, an HP Proliant ML110 G5. 

If you haven’t read Eric’s post yet I recommend you check it out as it’s an interesting read. 

As mentioned the SSD I purchased is a Samsung MMCRE64G5MXP-0VB which is a second generation SSD and offers 64GB of SATA II (3Gb/s) storage.  My reason for purchasing this particular make and model of SSD came down to price and the fact that, so I have been told, Samsung pretty much produces all the SSD’s for all other branded SSDs currently on the market.  I appreciate that there are faster SSDs from the likes of Corsair and OCZ but the £145 I spent was as far as my budget would stretch.

Here is the specification of the Samsung MMCRE64G5MXP-0VB SSD:

Capacity: 64GB
Read: Up to 220MB/sec
Write: Up to 120MB/sec
NAND Flash: Multi-Level Cell (MLC)
Interface: SATA-II
Low Power Consumption
Lightweight (80g)
Shock Resistant
Warranty: 2 Years

Comparing the technical specifications of the SSDs my Samsung SSD is slower for reads (>220MB/sec vs >250MB/sec) though on paper is slightly quicker than the Intel X25-M for writes (>120MB/sec vs >70MB/sec) so it was going to be interesting to see how the two SSD’s would perform when presented through ESXi as vmfs storage to a VM.

In the ML110 G5 ESXi 4.0 lab server I was going to use an HP E200 Smart Array controller which I had been using for a series of performance tests for an upcoming blog post though the Initial tests from the HP E200 controller produced shockingly slow results (blog post to follow on this) so I attached it to the onboard ICH9 SATA controller.  Another reason being that Eric used an ICH9 SATA controller for his testing so it was good to match test environments like for like as much as possible.  The process of attaching the SSD drive was no different than that of attaching a local SATA, SCSI or SAS disk and looks exactly the same from within the vSphere Client (see below).

The Test:

Following the same procedure and configuration as Eric I presented a W2K3 VM with a separate 8GB drive from my locally attached SSD for testing.  I also used HD Tune Pro (trial) to perform the read and write benchmarking tests.

The Results:

Onboard ICH9 SATA Controller:  Here are the results when presenting a blank 8GB SSD (connected via the onboard ICH9 Controller) based (thick) drive through to my W2K3 VM.

Sequential (64KB Block) – Read:	Sequential (64KB Block) – Write:
Random Access – Read:	Random Access – Write:
Extra Tests – Read:	Extra Tests – Write (with Cache):
	Extra Tests – Write (without Cache):

From the results given by HD Tune Pro the following is a comparison of the sequential and random read/write benchmark results. As you can see the Intel X25-M is quicker than the Samsung SSD in almost all read based tests.  Thanks to Eric for allowing me to use the results from his testing to generate these comparisons.

Summary

The Intel X25-M SSD is consistently quicker for reads than the Samsung MMCRE64G5MXP-0VB in almost all areas with the only exception being the sequential read burst rate.  Technically the Samsung SSD should be slightly quicker, according the to the specification, for write based transactions – I will check with Eric to see if he has any write based results for the X25-M so we can compare.

As mentioned earlier I found that there was a massive difference in both read and write results when performing these benchmarking tests with the Samsung SSD connected to the E200 (located in the PCIe x8 bay) Smart Array controller compared to when performing the same tests with it connected to the onboard ICH9 based SATA controller.  I will posting another article that covers this as it would appear that the E200 controller really offers sub-standard (to put it politely) performance results with SATA based storage which is consistent with the findings of Lukas Beeler in his post here.

For me both of these SSD’s (Intel and Samsung) offered decent results with the Intel definitely having the edge with the Reads.  In all I have been very impressed with my Samsung SSD purchase and also the results from Eric’s Intel X25-M which I would also happily purchase.  I look forward to the prices of SSDs dropping and the capacity increasing in the not so distant future so I can start using SSDs more in my home environment.  Even more exciting will be the inevitable emergence of wide spread use of SSDs in enterprise storage.