Path: utzoo!utgpu!jarvis.csri.toronto.edu!mailrus!cs.utexas.edu!usc!snorkelwacker!think!linus!eachus From: eachus@aries.mitre.org (Robert I. Eachus) Newsgroups: comp.sys.amiga Subject: Re: 2000s / 500s and hard drives -- reliability Message-ID: Date: 23 Jan 90 15:45:18 GMT References: <7196@dayton.UUCP> Sender: news@linus.UUCP Distribution: usa Organization: The Mitre Corporation, Bedford, MA Lines: 59 In-reply-to: joe@dayton.UUCP's message of 22 Jan 90 22:13:09 GMT If you truely mean that reliabiltiy of the disk is absolute, you are going to have to have some scheme for fault tolerance. From experience, cheap disks or expensive disks, they can all be killed by a major power failure. (Surge protectors and such are nice, but when the whole city goes out, power lines do very strange things, and these up and down surges often cause head crashes.) Thus the best strategy for protecting data is either 1.) Write everything to two hard disks. or 2.) Keep the days transactions in battery backed RAM (and on hard disk) and back that up using floppies or some other media. Approach number two is very inexpensive on the Amiga. Several manufacturers of RAM boards sell a battery backup option, and the Amiga kernel requires no modification to support it. With careful system design, you can have a fault tolerant Amiga application for less than $1000 extra in hardware costs, and no modifications to any commercial applications you may be running (assuming they allow multitasking). All you need is a background task which shadows the key files,or you can modify the device driver for the hard disk to write to two places. Having said all that, the choice should be obvious. You can either rely on high-cost high-reliability disk drives (there are very reliable SCSI disks available in the $4000 range), and do regular backups (and hope they are recent enough when something happens) or set up your system to assume disks will fail, and keep a few spare disks in stock. If you plan to do good backups (or should I say if you will do good backups), then even one failure per month should be no problem. Some quick relability calculations: 24*365 = 8760 hours/year. Above 40000 hours guarenteed MBTF, disk drives get VERY expensive. Forty machines at 40000 hours MBTF is eight to nine failures per year. Recommendation: buy 25000 hour MBTF drives, and set things up so that a disk failure results in less than an hour of lost time. (If that means external drives so be it.) BB-RAM is worth the price whatever you do. Another alternative is to buy Ethernet cards (or perhaps use DNet) and maintain all records on a single central fileserver. Hard disks in the Amigas will be unnecessary, and your only problem will be making damn sure the file system on the fileserver is reliable. When the inevitable crash occurs three or five years down the road should not be the first time you try to restore things from backups. My experience with the Ameristar card can be summed up as I'm not going to tear my hard disk out of the Amiga, but if it fails I may not replace it. (Acutally I'll probably replace it with a Bernoulli or some such.) The convenience of having to do one backup instead of 40 may make this the best approach for you. -- Robert I. Eachus with STANDARD_DISCLAIMER; use STANDARD_DISCLAIMER; function MESSAGE (TEXT: in CLEVER_IDEAS) return BETTER_IDEAS is...