home *** CD-ROM | disk | FTP | other *** search
- Newsgroups: comp.os.vms
- Path: sparky!uunet!zaphod.mps.ohio-state.edu!sdd.hp.com!ux1.cso.uiuc.edu!news.cso.uiuc.edu!jsue
- From: jsue@ncsa.uiuc.edu (Jeffrey L. Sue)
- Subject: Re: Bound volume sets: are they a bad idea?
- References: <1992Nov18.122133.1@rhodes.aero.org> <1egbrqINNiqd@post.its.mcw.edu> <102616@bu.edu>
- Message-ID: <1992Nov20.150327.11260@ncsa.uiuc.edu>
- Originator: jsue@troon.ncsa.uiuc.edu
- Sender: usenet@news.cso.uiuc.edu (Net Noise owner)
- Organization: The Dow Chemical Company
- Date: Fri, 20 Nov 1992 15:03:27 GMT
- Lines: 71
-
- In article <102616@bu.edu> TEWKSBURY@BUDSGA.BU.EDU (Chip Tewksbury) writes:
- >In <1egbrqINNiqd@post.its.mcw.edu> kcb@post.its.mcw.edu writes:
- >
- >> singer@rhodes.aero.org wrote:
- >>
- >> : >>As far as I can reason it out, the pro is some degree of load balancing
- >> : >>between the disks, and no management hassles sorting out what to put where.
- >> : >>The con is a halved MTBF on the bound set. Backup shouldn't be a problem,
- >> : >>both together will fit on one DAT.
- >>
- >> Wrong. Volume binding does NOT do much load balancing at all. At the
- >> very least, you have absolutely no control over it. My best example
- >> is our own shop.... we have humungous MUMPS.DAT files (on the
- >> order of several million blocks large), and we used to have bound
- >> sets. Most of the time, the first (n-1) disks were totally idle,
- >> while the last disk in the set got absolutely pounded. Your
- >> mileage will vary, of course.
- >>...
- >
- > Kent, Your response is partially correct. Volume sets automatically
- > create new files on the volume in the set which has the most free space,
- > unless the creation is done with explicit placement. If a system's file
- > sizes are relatively small and comparable, then excellent load balancing
- > at the file level will be realized from a volume set. Striping won't
- > add much, if anything, to the load balancing performance if you have
- > smaller files of comparable size.
- >
- > But as you said about your sight, if you primarily have "humungous"
- > files then the greatest benefit comes from striping.
- >
- > "Your mileage will vary, of course." And each site needs to identify
- > their system's file population (and propagation) before jumping on the
- > single physical volume, multi-volume volume set, or multi-volume stripe
- > set bandwagon.
- > Chip Tewksbury.
-
- Has anyone ever questioned DEC on their rather lackadaisical support of
- bound volume sets in VMS? I always found them extremely difficult to
- work with when doing "operation" types of functions. For example, try
- doing a BACKUP/IMAGE from one volume set to another, with the output set
- being a different number of volumes... good luck! I doesn't work.
- The only choice is to BACKUP [*...] from one set to the other, and this
- took 16 hours to complete on a 3 volume RA82 disk (VMS 5.0 days). I didn't
- like that kind of downtime, and neither did my customers.
-
- As to the I/O load-balancing: If all of your files are fairly small, then
- there is some load-balancing of the I/O due to files being placed on different
- volumes. However, if you have one or more files that are very active, then
- I/O load-balancing is practically non-existent. In this case, Striping has
- been the only saving grace for us.
-
- Basically, Striping gave us much better support in VMS for systems/operations
- functions, and also much better I/O load-balancing. As stated before, the
- downside is that you can't *tune* the stripset without creating a new one
- and copying the old one over to it (either disk-to-disk, or disk-tape-disk).
- However, the overall I/O loading capacity is just about n*"disk saturation
- point". In other words, with 3 RA82s, which could handle 30 I/Os/second,
- we could get up to 80-85 I/Os/second fairly easily before we started getting
- queued requests. Today's faster disks offer even more requests/second.
- In fact, we saw the I/O bus become more of a bottleneck - so we switched
- to the CIXCD controllers (over the CIBCA); and HSC70s switched from HSC50s.
-
- Anyway, this has been my experience in the volumeset vs stripeset
- controversy.
-
-
-
- --
- -----
- Jeff Sue
- - All opinions are mine - (and you can't have any, nya nya nya)
-