high performance for a cluster ??

high performance for a cluster ??

Post by Ed Bra » Thu, 30 Nov 2000 04:00:00

Hellu there,

Can you please give me advise on a database choice that fulfils the
following needs :
- Very fast. Is able to do at least 150 simple transactions per second on a
simple allround machne, like a PIII 800 with 256 MB RAM.
- Is able to function in a cluster
- Doesn't have that license shit as Oracle.

Ed Bras


1. SQL 6.5 Confilicting Performance: Memory-Page Faults: High AND Cache-Avg Free Page Scan High

My SQL 6.5 has some performance monitor object measurements that appear to
have confilicting solutions.
My server is a HP NetServer with dual P200 processors, and 416 MB RAM
running NT 4.0 w/SP6a and SQL 6.5 w/Post SP5a
The SQL Server is allocated:  256 MB RAM

The Memory-Page Faults/sec on the Server are: AVG 400+
The SQLServer-Cache Hit Ratio is: 99
The SQLServer-Cache - Avg. Free Page Scan is: 159
The SQLServer-Cache - Max Free Page Scan is: 52

My research on SQL performance says:
1. The Memory-Page Faults/sec are too high and more RAM should be given to
the System.
2. The Cache Hit Ratio is good, virutually all hits are found in cache, an
indication that SQL has more than enough RAM?!
3. The Cache Free Page Scans (Avg. and Max) are way too high, they are
supposed to be around 10 or less.  The recommended solution is to increase
the Data Cache size or frequency of checkpoints.

I am confused because the Page Faults/sec indicate the system does not have
enough memory allocated to it, but the Free Page Scans indicate that the
Data Cache is not big enough.  It appears that these performance monitor
objects are asking for the oppisite resolutions.  More system memory to
reduce Page Faults/sec will give the Data Cache less memory to divide
between it and the Procedure Cache thereby increasing the Free Page Scans.

The SQL Server has 256/416 = 0.61 or 61% of the RAM, the System has 160 MB
RAM or 39%, from what I am reading this should be more than enough RAM for
the System yet the Page Faults/sec are high.

Could the system need more RAM, and the percentage of RAM for the Procedure
Cache be reduced to keep the Data Cache amount of RAM higher??

Does anybody have any suggestions?



2. Borland quarterly/yearly results?

3. TechTips: High-volume / High-performance Paradox

4. Optimising size requirements of replicated SQL CE database.

5. DEC Rdb:Highest TPC-A Performance and Lowest Price/Performance Records

6. Script FK

7. Higher than normal CPU Utilisation on CLUSTER

8. INFORMIX Report Writer

9. High Availability and Clustering Newbie question!!!

10. clustered index scan cost is high

11. Clustering high-availability

12. Clustering High availability

13. SUN High Availibility clusters and Ingres 6.4