From: | Josh Berkus <josh(at)agliodbs(dot)com> |
---|---|
To: | Adam Scott <adam(dot)c(dot)scott(at)gmail(dot)com>, pgsql-performance(at)postgresql(dot)org |
Subject: | Re: Queries Per Second (QPS) |
Date: | 2015-09-30 17:06:46 |
Message-ID: | 560C16A6.1000004@agliodbs.com |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
On 09/26/2015 09:24 AM, Adam Scott wrote:
> How do we measure queries per second (QPS), not transactions per second,
> in PostgreSQL without turning on full logging which has a performance
> penalty and can soak up lots of disk space?
Measure it from the client side. pgBench does this.
If you mean on your production workload, then I recommend using a
connection proxy which counts statements. A few exist for Postgres, for
example:
VividCortex:
https://www.vividcortex.com/blog/2015/05/13/announcing-vividcortex-network-analyzer-mysql-postgresql/
WireShark: https://github.com/dalibo/pgshark
You'd need to measure how much one of these tools affects your QPS, of
course, but that should be easily measurable on a test system.
Also, if the PostgresQL activity log is moved to a seperate SSD from the
database storage, I've found overhead in writing to it to be less than
3% ... depending on the nature of your query traffic. Pathological
situations are mainly databases which have a high volume of very long
queries or failed connection attempts.
>
> We are using 8.4, but I'm interested in any version as well.
You are aware that 8.4 is EOL, yes? Not to mention missing 5 years of
performance improvements ...
--
Josh Berkus
PostgreSQL Experts Inc.
http://pgexperts.com
From | Date | Subject | |
---|---|---|---|
Next Message | ajaykbs | 2015-10-03 08:39:33 | Multi processor server overloads occationally with system process while running postgresql-9.4 |
Previous Message | Michael Paquier | 2015-09-30 02:38:28 | Re: dump restoration performance |