MyDigitalSSD is a private label MLC drive distributed by MyDigitalDiscount of Oswego, New York.
The basic design suggests two basic problems when this drive is used as a bare device without MFT. The first of these is that the MLC version has a very slow native random write speed: +/- 4 random writes per second. This is the same speed as is found on MLC USB sticks as well as CF Cards. This is troubling as linear write speed is a very good (for MLC) 56mb/sec. The other dramatic problem with the drive when used in native mode is that the device uses an 8mb erase block. This means that for each random write, 8mb of data have to be written. On the one hand, this explains the high random write rate. On the other, using an 8mb erase block, rather than the 2mb used by most other manufacturers, means that the write life of the drive is reduced by a factor of 4.
In short, we would strongly discourage use of this device without MFT except for specific applications. These would be any environments where the drive is to be either used almost exclusively read only (possibly, some web servers), or where the drive will be used together with large video or sound image on demand systems. Here, the drive's extremely low cost combined with its high linear speeds are desirable.
For general computing use, this product must be used with MFT. And what is exciting is that when it is used with MFT, its effective speed improves 183-fold (see Aggregate results) to a level somewhat above CF cards, and soundly in the middle-ground of performance. Just as importantly, MFT overcomes the wear problems of this device. While the expected write life of most flash devices improve somewhat more than 50-fold, when used with MFT this drive's life actually improves over 200-fold, because MFT negates the disadvantage of the 8mb erase block design choice. Even more interesting is the notion that if the manufacturer of this drive were able to speed up the very slow random access speed, performance of this device using MFT would probably accelerate a further 3- to 4-fold, making it a top performer in the MLC class and almost on par with fast SLC drives. Thus, it demonstrates how MFT can give designers freedom to make radical design choices.
Because of the very large erase block size of this device, we don't really recommend the use of this device in large arrays. For example, with a raid-5 array of 8 drives, this would create an "effective erase block size of 56 MB ( (8 drives - 1) * 8 MB ). While MFT does handle this situation, MFT's kernel memory footprint starts to get large and the array can get a bit "bursty" as reads are locked out during writes that can take many milliseconds. While these drives are unquestionably good solutions at up to 4 drives, larger arrays should be tested before any large scale deployment. |