| T.R | Title | User | Personal
Name | Date | Lines |
|---|
| 264.1 | Here are Three | BROKE::BOOTH | What am I?...An Oracle? | Fri Dec 09 1988 04:14 | 10 |
| Try using a MicroVAX for the performance test. There is not enough
memory available to set up the huge caches that O & S need to enhance
performance.
Also, a VAXcluster system with database access from more than one
node will eliminate Sybase and cripple Oracle V6.
Use direct access to RMS files. O & S can't do it.
---- Michael Booth
|
| 264.2 | design against the global cache | DEBIT::DREYFUS | | Mon Dec 12 1988 21:17 | 19 |
| > Try using a MicroVAX for the performance test. There is not enough
> memory available to set up the huge caches that O & S need to enhance
> performance.
If all database users are going after the same data, doing the same thing,
this may not help. Global caching does work on microVAXes. However,
what you can do is construct an application design that causes the global
cache to thrash. While I have not tested this approach, it may work.
If each user goes after a different data set (part of the database), then
the approach that we use allows each user to maintain access to their
own buffer space. In a shared buffer-space environment using LRU page
management we may find that Oracle and Sybase constantly flush each users
page out of memory just before they need it.
In the worst case, this type of application configuration doesn't give
Oracle or Sybase an edge with their global caches and fast-commit algorithms.
--david
|
| 264.3 | Another thought | MDVAX1::DUNCANG | Gerry Duncan @KCO | Tue Dec 13 1988 21:11 | 12 |
| For Oracle V5, have them do a prototype which updates different
rows in the same table from more than 1 program. That will put
O out to lunch. Then force a cluster, even the "hot" V6/TPS will
choke.
Re .2 David is correct. In our recent benchmark, the global caching
doesn't seem to help our throughput. We ran almost identical times
on 8820 and 6240 under V6/TPS using approx. 5mb SGA. If Oracle
was right about the 6xxx bus vs 8xxx, I would have expected to see
the 6240 run slightly ahead of the 8820.
--gerry
|
| 264.4 | >>> Further Infor to .0 and ???? <<< | TROA02::WILLMS | | Wed Feb 01 1989 20:07 | 29 |
| Current Config. is a LAVC consisting of :
5 Diskless VS2000 with 14M memory each.
( possibly expanding to 10-15 workstations)
VAXserver 3400 acts as the BOOT node for the VS and a File server.
(1 TK70 Tape Drive, 3 RF30 Disk Drives)
? - Customer (and myself included) will be looking for help/info
on how to monitor performace of both RDB and SYBASE and acheive
Price Performance info. (Any tricks to be aware of.)
Note 1 - Customer is rewritting code from IBM onto the VAX and they
are new to the VMS enviornment.
Note 2 - I am new to Digital (1 Month) and I am also new to the
LAVC Environment. Before coming to DEC I worked in a VAX Cluster
Environment and have a basic understanding of how things work but not
in great detail. I have also had some experience with DEC DBMS but
not RDB.
Any answers on how to moniter etc.. or where to find it would be
appreciated.
P.S. I have been reading some of the other NOTES conferences to get
more infor on RDB/RALLY/SQL etc... and 3rd party DB.
P.S.S. Does anybody know how much disk storage each additional Node
in the LAVC will take when added.
|