Comments on: Intel, Hitachi to develop solid-state drives

Intel and Hitachi Global Storage Technologies announce plans to jointly develop and deliver solid-state drives for servers, workstations, and storage systems.

[hmdz105]

What will happen to Perfect Disk and Deekeeper and all other defragmentation companies then?

[NWLB]

Depends on if data can be more efficiently arranged on an SSD as it can a HDD. Plus, these drives aren't big enough to replace bulk storage on HDDs for now. So they have a market for a long time to come.

[Thomas, David]

Defragmentation has more to do with the operating system, and it's file system, then any type of hard drive. Correct me if I'm wrong, but I don't think it has anything to do with the hard-drive at all, except where the file system has to manage "dead spots".

[dwinks]

Fragmentation is purely a function of the OS and File-system. HDDs have moving parts which take time to seek (move), and when a file is fragmented, it requires more than 1 seek to read, which slows down speed, and on crappy filesystems that get fragmented badly (FAT32/NTFS mostly), it could mean hundreds or thousands of seeks per file. Since SSDs don't have moving parts, and their seeks (which they DO have, though non-mechanical) take around 0.001 milliseconds, rather than 10-25 ms like HDDs. Such fast seeks make defragmenting useless.

Not only that, but almost all SSDs implement "wear-leveling", which sends data to random parts of the disk, rather than contiguous areas, even if the OS/Filesystem tries to send a contiguous write, and consquently, even if you "defragment" a SSD, each file is stored in many many parts on the disk and the OS is unaware if this, which means that defragmenting a SSD does ONLY ONE thing, which is unnecessarily wear down the drive. However, that really isn't a concern, because contrary to what some people would try to tell you, MTBF (mean time between failure) on a SSD is measured in millions of hours, which under fairly heavy use, 24/7/365 works out to something like 30 years, which is far beyond the actual useful life of a SSD.

As for the "bulk storage", magnetic media is getting VERY close to the physical limits of data density per platter, and I have read that a 3.5" HDD will top out at around 5-6TB. The physical limits of density were already reached with parallel recording, for the most part, and only by switching to perpendicular recording have they managed to squeeze more onto a disk, but that is rapidly hitting it's limit too, with no new way to improve it. SSDs on the other hand are estimated to be able to eventually hold thousands of TB in a 3.5" form factor.

DNA is most likely the close to, if not the highest data per gram/volume, and it's said to be something like 2.25 ZETTABYTES per gram. There isn't anything physically preventing SSDs from getting at least 10% that data density, which would likely be around 100-200 zettabytes per disk. HDDs can't pack too much more data per platter before random bit-flipping starts to occur. It should only be at most 5-10 years before HDDs are completely replaced.

Also, due to the nature of how SSDs are made, it's possible to put each memory chip in parallel in what amounts to a raid-0 array, and there are disks already in the lab being tested that exceed 1gbps transfer rate with effectively 0 seek time and 100x the reliability of HDDs. It's expected that SATA's 3.0gbps transfer speed will bottleneck SSDs in only a couple of years!

[quirK]

Not to pour cold water on your analysis, which is highly interesting, but DNA in its highest density state is packed tightly and unable to be accessed for information (wrapped into chromosomes for cell division). Nevertheless, for anything man-made to approach DNA's density, even uncoiled; that is a feat.