jhs@MITRE-BEDFORD.ARPA (07/25/87)
Glen Stone's idea of a 16Mbyte RAMdisk at first sounded to me like a solution to which there is no corresponding problem. The problem I see is that the main point of a hard disk is that it is nonvolatile. In fact the main problem with hard disks is that one accumulates a large amount of data on them and then is vulnerable to catastrophic consequences if it all evaporates. I would think that a RAMdisk version would be all the more vulnerable to this sort of problem. However, there is ONE problem for which Mr. Stone's idea would be a darn good solution: speeding up the access time in a hard disk. That is, Mr. Stone's gadget should perhaps be integrated with a normal, in fact fairly cheap, hard disk, say one with really snail-like access times, 100 milliseconds or more. Perhaps a 16 Mbyte hard disk with such access times could be made really inexpensive. If so, perhaps some enterprising hard disk manufacturer should integrate one of their cheap, slow disks with Mr. Stone's Supercharger. When you power the supercharged disk drive on, it copies the magnetic storage contents over to RAMdisk. This might take awhile, but perhaps only several seconds if it was well designed with a very high speed internal path. And with a sensible sequence which involved only the 1-track step access time rather than the "average" time. Also, at least READ access could be allowed from the magnetic medium at boot-up while the transfer was going on. Or the RAM could be SRAM with battery backup, so most of the time there would be no problem. Anyway, once the transfer to RAM was complete, the RAMdisk would take over realtime processing, giving disk access times in the microseconds instead of milliseconds. While we are at it, the controller for this device could keep track of write accesses and back them up to nonvolatile magnetic disk storage at its leisure -- basically as fast as it could get around to it. The BACKUP process would lag realtime by a few tens of milliseconds to a few seconds, perhaps. When power was again shut down, the system would grab the power switch away from you and keep power on long enough to finish any backup operation in progress before actually shutting down. Now THIS device would be a cheap, blindingly fast hard disk, and might well have a lot of eager customers! You wouldn't even have to mention the fact that the 100-usec average access time was done with a RAMdisk -- just let the competing disk manufacturers wonder how the heck you did it! Or more likely, let them think it was a misprint in the spec sheet! Any comments from the net? -John Sangster / jhs@mitre-bedford.arpa
geoffs@brl-smoke.ARPA (Geoffrey Sauerborn ) (07/27/87)
Why not have this "16 Mbyte RAMdisk" battery backed-up? (With a built-in clock/calendar to boot!) If it is battery backed-up and the battery will last 5+years and/or is rechargeable, then who needs a hard disk? -- ---> geoffs@brl.arpa --
ljdickey@water.UUCP (07/29/87)
In article <8707250333.AA08699@mitre-bedford.ARPA> jhs@MITRE-BEDFORD.ARPA writes: > ... there is ONE problem for which Mr. Stone's idea would be a darn good >solution: speeding up the access time in a hard disk. ... > ... once the transfer to RAM was complete, the RAMdisk would >take over realtime processing, giving disk access times in the microseconds >instead of milliseconds. .. >The BACKUP process would lag realtime by a few tens of milliseconds to a few >seconds, perhaps. Great idea. Fantastic. This sounds a lot like the idea of CACHE memory to me. I want it. -- L. J. Dickey, Faculty of Mathematics, University of Waterloo. ljdickey@water.UUCP ljdickey%water@waterloo.CSNET ljdickey%water%waterloo.csnet@csnet-relay.ARPA ljdickey@watdcs.BITNET UUCP: ...!watmath!water!ljdickey
brinsmead@calgary.UUCP (Mark Brinsmead) (07/29/87)
In article <6181@brl-smoke.ARPA>, geoffs@brl-smoke.ARPA (Geoffrey Sauerborn ) writes: > Why not have this "16 Mbyte RAMdisk" battery backed-up? (With a built-in > clock/calendar to boot!) If it is battery backed-up and the battery > will last 5+years and/or is rechargeable, then who needs a hard disk? > I think Glen already answered this question on the net awhile back... his argument is a) that to provide a reasnable battery backup (say 24 to 48 hrs) for 16 meg of dynamic RAM would take a REAL big battery and b) that for every megabit of memory, you can expect to have one bit randomly toggled every so many days (2? 10?). A Ramdisk is just not SAFE for long term storage, but with a 20 minute battery-backup to protect against most power failures, I for one would be willing to risk it (as long as I have backups). Of course, this means that if you are occasionally prepared to backup and/or rebuild your ramdisk, you could survive quite nicely without a hard disk for quite some time. Mark Brinsmead @ University of Calgary ------------------------------------------------ Disclaimer: What do I know about this? I'm a software type.