investigating spinlock contention using HS

investigating spinlock contention using HS

Post by Ashi » Fri, 08 Mar 2002 09:50:42

We had perf probs on a server(ASE 11.5.1, solaris 2.6).
So, we added another CPU and some RAM.

Now, I want to measure the spinlock contention.
I am collecting stats using Historical Server.
The manual says that there is one data item - "cache spinlock contention".

using sp_sysmon,  we can find spinlock contention for
. object
. index
. hash
. cache

Is it possible to see all of these from the Historical Server ?



1. Spinlock Contention

I have a named cache bound to tempdb. The tempdb size is
3.5GB and the named cache size is 650MB. I have 2K, 4K, and
16K buffer pools in my named cache.

When I ran sp_sysmon, I was really alarmed by the spin lock
contention on this named cache. The spin lock contention was 602%
and the utilization was 11%.

Can someone please explain what this number (spinlock contention 602%) is
and whether I should be worried about the effectiveness using named cache
for tempdb

Your response is highly appreciated.


2. Setting up a Novell 4.1x server with FoxPro DBF Tables to link with SQL Server 7 on Windows NT Server.

3. spinlock contention on data cache in 10.0.3

4. Faxing

5. spinlock contention om tempd

6. 3rd Generation Database Manifesto

7. Spinlock contention on tempdb

8. Update This

9. investigating deadlocks without using trace flag 1204 ?

10. Multi user contention - using BEGINTRANS and COMMITTRANS

11. Bus. hs logic - ignore weekends

12. HS / Transparent Gateways / Linux Redhat 6.2 / Oracle 8i (8.1.6)

13. HS ODBC or OLEDB connections