| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 835.1 | Check the ACMS notes conference note 1636 | IJSAPL::OLTHOF | Henny Olthof @UTO 838-2021 | Thu Dec 27 1990 13:50 | 6 |
| Hi,
Note 1636 in the ACMS notes conference (CLUSTA::ACMS) might give you
some answers.
Henny Olthof, TP-DB Netherlands
|
| 835.2 | A couple of examples | COOKIE::C_DAVIS | Bugchecks R Us | Thu Dec 27 1990 18:38 | 38 |
| The answers as I remember them from on V3.1B site running on a 6000 machine
using disk backups (bound volume sets) and TA90 (2 hoppers):
>
>What kind of RMU/BACKUP performance can be expected when backing up to disk
>(using stripe sets to back up to multiple disks if needed)?
Poor performance when compared to TA90. 8 million block RBF file in 10 hours
from a 6 gig multi-file database.
>
>What about backing up to a single TA90 drive?
Excellent performance. The same backup as above took 2 hours to backup (20
TA90's worth) and about five hours to restore.
>
>How about multiple TA90 drives on multiple HSCs (let's say three) to do a
>multi-threaded backup?
No multi-threaded tests where done, just single thread using two hoppers.
>
>Any clues as to the performance difference between online vs. offline backup
>(assuming that there were no competing jobs just to keep the playing field
>level)?
>
Not alot of difference.
Also, similar tests where repeated in the UK where the customer had to rethink
their backup strategy after they found that the TA90 backups where many times
faster than disk backups. The customer was testing against a 12 million block
database. The database was fairly empty (2 million blks of data; 4 TA90s) and
I was able to restore in my tests in about two hours on a 6000 machine.
RMU is optimized for tape backups.
Chris Davis
CSSE
|
| 835.3 | Thanks for your help (and a pointer) | LABC::WALLIS | Barry L. Wallis | U.S. DECtp Resource Center | DTN 535-4313 | Wed Jan 09 1991 21:59 | 7 |
| Thanks for your help. We should know by next week whether we've won the
project. If we do, I will post more details here.
If you are looking for more information on this subject, try the note I
cross-posted to RDB_31 (#1511).
- Barry
|
| 835.4 | TF800 series | AKOCOA::HAGGERTY | GIA EIS/SWS, Acton MA. | Wed Mar 06 1991 23:14 | 8 |
| Does anyone have any statistics on the new TF837 tape subsystem, or
is it still too new ?
thanx
Kevin
|
| 835.5 | | NOVA::DIMINO | | Thu Mar 07 1991 00:51 | 15 |
| > Does anyone have any statistics on the new TF837 tape subsystem, or
> is it still too new ?
Yes, still too new. The TF837 subsystem is low-medium performance
high capacity. 2 Gbyte capacity but IO of about 100 Kbyte/second
(.3 Gbytes/hour). But I doubt if there have been any realistic
performance mesurements performed yet.
Sounds great if your database isn't very large but you want
unattended backup.
-lou
|
| 835.6 | Try the TF857 | KYOA::KOCH | It never hurts to ask... | Wed Apr 17 1991 23:59 | 3 |
| Try the new TF857. It is 2.6GB/tape, 18.2GB unattended.
The I/O is 800kb/sec, with 2 supported on each KFMSA.
|