« EPIA System Init | Main | tivo HME SDK »
May 23, 2005
The R-Drive
Lately all the rage has been solid state storage systems. This is all well and good, but there are some limitations that people probably won't worry about. That is until it bites them. Most flash-based storage chips/systems/etc are only good for a certain number of writes, actually everything is only good for a certain number of writes, but flash-based are a lot less than hard drives. So, if there are some intesive writers out there then there is the possibility that their flash-based storage systems will eventually crap out.
Don't get me wrong, normal usage of USB thumb drives and their ilk are more than acceptable because there are a limited number of writes. They're usually only used as a personal storage (i.e. 3.5" floppy) replacement, which means that someone will usually write a document to it and then just read or make a few modifications per week. Compare this to a swap device that makes writes as fast as the device allows, especially on high i/o servers. Where is all this leading you ask?
Earlier this week Samsung released a 16GB Flash Laptop hard drive and M-Systems released a 176GB Flash Drive. Cool, great, but will they have shelf lives over 5 years?
I remember in college I had this idea for a hard drive using RAM instead of platters. Of course this has been done in the past through software RAM drives and specialty RAM boards, but my idea was to create a persistent RAM-based device. But how can this be accomplished without Magnetic RAM (MRAM)? By using some sort of backing store. This would make the system, which I dubbed the R-Drive, appear more as a super-cache, but when I designed it I took lots of little things into consideration. Using different techniques it could provide the benefits of fast data access, both read and write, and the long-term storage of data on a backing store. How fast? The theoretical limit of the Southbridge and interface (ATA/SCSI/SATA). This generation of SATA tops at 150Mbs to 300 Mbs.
Maybe one day I'll hit the Lotto and will have time to work for myself and the funding to accomplish such feats. Until then we have Flash-based systems.
Posted by Guy at May 23, 2005 07:27 PM