nagy%43198.hepnet@LBL.ARPA (04/14/87)
Some points for/against volume sets: 1. Too many volumes in a volume set is probably hazardous to a system manager's health. Especially in a software developement or normal timesharing environment where there are lots of little files on the disks... 2. Free space on members of a volume set are used in parallel by writing a new file on the member with the greatest amount of free space instead of completely filling member #1 then filling member #2, etc. This has promotes point #3: 3. Volume sets to improve performance by bring more seek arms to play on the file system. 4. In my past life as a system manager, I'd only played with dual-disk volume sets: 2 pairs of RA81s and a pair of RM80s as 3 separate volume sets. No problems beyond the usual. Online backups proceeded as before and did NOT affect the users any more adversely than online backups did before I installed the volume sets. The "great" argument for volume sets and for making ALL your file system one giant volume set is make it easy for users to locate the files of other users since they don't have to remember which disk user X is on. This argument is actually rather specious under VMS V4.x: the solution to the "problem" is to define a logical name which is a search list of all the mounted disk volumes in the public file structures: $ define usr$disks usr$disk1:,usr$disk2:,usr$disk3:,usr$disk4: Thus anyone can locate anyone elses file by using the usr$disks logical name as the "device" specification. This assumes that a person only has a single top-level directory on the set of disks listed in usr$disks; this does not seem to be much of a restriction.
9070NATS@MUCSD.BITNET (Keith Natschke) (02/02/88)
In light of all the discussion about volume sets, I would like to describe our experience with them. Although I don't recommend them to everybody, we have had some success increasing RMS throughput by utilizing a volume set. We have one volume set consisting of 2 RA81 disk drives dual ported to 2 HCS50's. The set is shared by 3 different applications, student financial aid, student fee accounting and library serials. The financial aid and fee accounting use several large indexed files. We took advantage of the volume sets by breaking the files into multiple areas, one area for the data and one for each index. We then placed the data area on one volume and the index area(s) on the other volume. This resulted in a dramatic decrease in direct I/O's and connect time when processing these files. ------------------------------------------------------------------------------ MM MM UU UU Keith S. Natschke Analyst/Programmer MMM MMM UU UU Marquette University MMMM MMMM UU UU Computer Services Division MM MMMM MM UU UU Systems and Programming MM MM UU UU 517 N. 14th Street MM MM UUU UUU Milwaukee, WI 53233 MM MM UUUUUU Phone: (414) 224-3765 Marquette University BITNET: 9070NATS@MUCSD INTERNET: 9070NATS%MUCSD.BITNET@CUNYVM.CUNY.EDU -------