ADSM-L

Re: Attn IBM: Pre-staging of ADSM tape mounts

1995-08-30 12:44:35
Subject: Re: Attn IBM: Pre-staging of ADSM tape mounts
From: Francis Dequenne <syf AT ECMWF DOT INT>
Date: Wed, 30 Aug 1995 17:44:35 +0100
On Aug 30, 12:24pm, Rob Creecy wrote:
<=> Subject: Re: Attn IBM: Pre-staging of ADSM tape mounts
<=>Andy Raibeck wrote:
<=>>Here's a feature that I'd *love* to see put into ADSM: the ability for ADSM
<=>>to pre-stage tape mounts, especially for restore. Here's how it would work:
<=>>
<=>>1) A restore that will require multiple tape mounts is started.
<=>>
<=>>2) ADSM calls for the first n tapes to be mounted. 'n' is defined by an
<=>>   ADSM server option called PRESTAGE (I'm making this up). The default
<=>>   value for 'n' is '1', which is what we have today. 'n' is limited only
<=>>   by the number of available tape drives.
<=>>
<=>>3) As each tape is mounted, ADSM positions it to the first location from
<=>>   where data will be read. While the first tape is actually being read,
<=>>   the second, third, and so on are mounted, positioned, and ready to be
<=>>   read.
<=>>
<=>>4) After the first tape has been read, ADSM can immediately start reading
<=>>   the second tape. The first tape is dismounted, and the nth + 1 tape is
<=>>   called for.
<=>>
<=>Why would ADSM have to wait until the first tape is read before reading
<=>tapes 2,3 etc.? Theoretically it should be possible to read all tapes
<=>simultaneously, although the implementation may be more difficult. Is there
<=>something I am overlooking?
<=>
<=>>This should work because ADSM already knows what tapes will be called for
<=>>(this must be true because ADSM sorts the list of versions to be restored
<=>>by volume).
<=>>
<=>>Such an implementation would speed recovery by almost eliminating the lag
<=>>time between tape mount and locate. This would provide the restore
<=>>performance benefits of collocation+ without having to incur the overhead
<=>>of collocating the data. The actual number of tape mounts would not be
<=>>reduced, just the lag time between one tape and the next.
<=>>
<=>>Any comments/thoughts?
<=>>
<=>>Andy Raibeck
<=>>Connecticut Mutual
<=>>203-987-3521
<=>>
<=>>
<=>Rob Creecy
<=>Census Bureau
<=>DIR IT, 1029-3
<=>Washington, D.C. 20233
<=>
<=>(301)-457-1774    rcreecy AT census DOT gov
<=> End of excerpt from Rob Creecy

It would also be nice for a user to be able to pre-stage in a disk storage-pool
files that he is likely to use in the next future (e.g. if a client restarts
work on an experiment whose data has been archived in the past, he might want
to be able to pull this data in a short period of time, without having to
incure the tape related overhead).

If I am right, nothing is provided to allow this kind of action, today.

--
Regards
Regards

+-------------------+----------------------------------+---------------------+
| Francis Dequenne  | Systems Section                  |      /~~\  /~~\     |
| ECMWF             | e-mail: fdequenne AT ecmwf DOT int      |     /    \/    
\    |
| Shinfield Park    | Tel:    (+44 1734) 499361        |   ECMWF             |
| Reading           | Fax:    (+44 1734) 869450        |   ECMWF             |
| Berkshire RG2 9AX | Telex:  (+44 1734) 847908        |     \    /\    /    |
| United Kingdom    |                                  |      \__/  \__/     |
+-------------------+----------------------------------+---------------------+
<Prev in Thread] Current Thread [Next in Thread>