We will share some new test results and repeat our old test again with a different method. Here's the results:

Samsung MLC SSD (16 GB)

ad6: 15272MB <SAMSUNG MMBRE16GSMPP-MVA VAM1101Q> at ata3-master SATA150
# diskinfo -c -t /dev/ad6s1d
/dev/ad6s1d
        512             # sectorsize
        16013910528     # mediasize in bytes (15G)
        31277169        # mediasize in sectors
        31028           # Cylinders according to firmware.
        16              # Heads according to firmware.
        63              # Sectors according to firmware.
        ad:SE932A6474s0s3       # Disk ident.

I/O command overhead:
        time to read 10MB block      0.126062 sec       =    0.006 msec/sector
        time to read 20480 sectors   2.174676 sec       =    0.106 msec/sector
        calculated command overhead                     =    0.100 msec/sector

Seek times:
        Full stroke:      250 iter in   0.044461 sec =    0.178 msec
        Half stroke:      250 iter in   0.037622 sec =    0.150 msec
        Quarter stroke:   500 iter in   0.081761 sec =    0.164 msec
        Short forward:    400 iter in   0.059677 sec =    0.149 msec
        Short backward:   400 iter in   0.065554 sec =    0.164 msec
        Seq outer:       2048 iter in   0.256953 sec =    0.125 msec
        Seq inner:       2048 iter in   0.251158 sec =    0.123 msec
Transfer rates:
        outside:       102400 kbytes in   1.389104 sec =    73717 kbytes/sec
        middle:        102400 kbytes in   1.038866 sec =    98569 kbytes/sec
        inside:        102400 kbytes in   1.582221 sec =    64719 kbytes/sec

Seagate 320 GB 7200 RPM HDD

ad4: 305245MB <Seagate ST3320613AS SD22> at ata2-master SATA150
# diskinfo -c -t /dev/ad4s1h
/dev/ad4s1h
        512             # sectorsize
        83310216192     # mediasize in bytes (78G)
        162715266       # mediasize in sectors
        161423          # Cylinders according to firmware.
        16              # Heads according to firmware.
        63              # Sectors according to firmware.
        ad:6SZ08FFZs0s7 # Disk ident.

I/O command overhead:
        time to read 10MB block      0.182695 sec       =    0.009 msec/sector
        time to read 20480 sectors   2.006781 sec       =    0.098 msec/sector
        calculated command overhead                     =    0.089 msec/sector

Seek times:
        Full stroke:      250 iter in   5.686091 sec =   22.744 msec
        Half stroke:      250 iter in   4.090876 sec =   16.364 msec
        Quarter stroke:   500 iter in   7.218680 sec =   14.437 msec
        Short forward:    400 iter in   3.434101 sec =    8.585 msec
        Short backward:   400 iter in   3.063467 sec =    7.659 msec
        Seq outer:       2048 iter in   0.207484 sec =    0.101 msec
        Seq inner:       2048 iter in   0.898884 sec =    0.439 msec
Transfer rates:
        outside:       102400 kbytes in   1.349878 sec =    75859 kbytes/sec
        middle:        102400 kbytes in   1.347827 sec =    75974 kbytes/sec
        inside:        102400 kbytes in   1.919180 sec =    53356 kbytes/sec

----------
Note: The test is done on a working FreeBSD server with both a SSD and HDD on SATA-150.

We'd like to share an experience about PCI Sata controllers with VIA VT6421 chipset and solid state drives. This PCI cards support software raid. You can set raid with jumpers on the card. We have not tested this system with a Windows or another operating system. The test consist on BSD Unix.

We've tested this card on both FreeBSD 8.0 and Slackware Linux 13. To be honest, the card alone is awfull. It gives so much read/write errors on both BSD Unix and Linux.

The card detected both the HDD and SSD. The dmesg output includes the drive information. However, operating system don't see slices. Yes, you see the main drive on /dev, but you can't see the slices for even fdisk the drive. At least, we couldn't get succeed.

We don't hear but of couse somebody might use this system without any problem before, however we couldn't.

We do not suggest this VIA sata controller to Unix and Linux users. If you have an SSD, do not get close.

We will share some MySQL results of our BigBoy SSD Slim. The hardware is the same as our previous post. This time the test is based on MySQL queries.

We did a simple sql select query like that:
SELECT * FROM `table` WHERE `field` LIKE '%keyword%';

Executed this query on the same table for keywords at different long. The table contains about 100.000 rows and it's size is about 500 MB.

Test	SSD	HDD
Query 1	0.0261 sec	0.0410 sec
Query 2	16.2090 sec	28.2846 sec
Query 3	21.1690 sec	34.0826 sec

Conclusion:
For the hardware and software we use, our SSD is better about 60% of the HDD. Please not that, our SSD is designed for personal use like laptops/desktops. Although that, it's also good for average servers. It's performance is better than HDD. You can see the difference with various tests.

The bad thing about this SSDs is cost/performance rate. Today the cost of this 16 GB SSD equals a 1 TB HDD. The cost is very expensive, if your priority is storage.

But if you want to speed up your operating system or database server, and your database is not bigger than 16 GB, you can buy a SSD for your workstation. Because our priority is performance rather than storage.

If you want to raise your performance more, you can connect your SSDs in raid. This will give you more performance.

We have bought 16 GB BigBoy (Samsung Chips) Slim MLC SSD last week. In physical size, it's really so slim, thin, small. It's about 1 inch in size. The size advantage of solid state drives really matters.

What about the performance?
First of all, we didn't test the drive with for various factors. We only did simple copy and read tests.

The platform is FreeBSD unix. We formatted the SSD as UFS file system. The motherboard supports Sata 1. So the results are based on Sata 1 only.

We wrote a shell script, that copies a 35 byte file 10.000 times and then randomly reads this files.

We use a Seagate 320 GB 7200 RPM HDD to compare with our SSD.

The results:
SSD:
Write: 44.8106 seconds
Read: 1.9424 seconds

HDD:
Write: 49.6177 seconds
Read: 2.7077 seconds

We see that, this SSD is a little better than the HDD we use. There is no big difference for this models. However, we have to say that, we have tested the drives with a simple script which we coded ourselve, and the connection was Sata 1. Of course you can get different result with different platforms.

We will use this SSD in one of our servers for operating system or database (e.g. mysql). We didn't decide yet. We will do some more tests.

In near future, we will buy and test other SSD models and share the results in SSDAge.com

Solid state drives write the data into memory cells. This cells are made of floating-gate transistors.

Single Level Cell (SLC)
SLC flash memories store only one bit of data in a single cell. Each cell stores one bit. So SLC memories are capable of faster write. SLC also has the advantage of higher cell endurance and lower power consumption. Although SLC technology has so many advantages, it costs more than MLC. SLC technology is chosed for servers and heavy applications by companies.

Multi Level Cell (MLC)
MLC flash memories store generally three or more bits in a single cell. Because of the high cost of manufacturing SLC memories, MLC technology is designed. MLC memories are generally cheaper than SLC memories. However, they have higher power consumption, lower cell endurance and slower transfer speed rates. SLC technology is chosed for desktop and laptop computers by home users.

• 1

• 2