Performance – Oracle Pool with PHP

By Juan Pablo Bosnjak & Javier Neyra

Introduction

Almost all applications use a database, furthermore, almost all applications with a database find that their main performance bottleneck is in their own database. Databases are too slow compared to code processing, but we need them.

We spend a lot of time improving queries, avoiding full scans, but we forget that to open, establish and close a connection to the database often takes more time than the query itself.

I am new in the PHP world (I come from J2EE), so imagine my surprise when I found out that most PHP productive applications do not use connection pooling.

After some research, we have found that PHP does not provide native connection pool support. Fortunately, I came across a new feature in Oracle 11g: Database Resident Connection Pooling (DRCP).

Basically, DRCP is a feature of Oracle Database 11g: it has been designed to address scalability requirements in environments that require a large number of connections with minimal database resource usage. DRCP pools a set of dedicated database server processes (pooled servers), which can be shared across multiple applications running on the same or several hosts.

For additional information, check:

https://docs.oracle.com/cd/B28359_01/server.111/b28310/manproc002.htm#ADMIN12483

Let’s give it a try.

Test Case – Simple PHP web Page

PHP test page:

Now, let’s create the pool in our Oracle 11g+ database:

Note: pool parameters should be defined according to your application needs.

Performance Test Definition

Let’s create 2 simple scenarios with JMeter.

10 clients made concurrent HTTP request connections for 30 seconds with think time at 0 milliseconds.

Environment

• 3 desktop machines, one for Oracle, one for Apache, one for Jmeter.

• Standard LAN megabit layout.

• Apache configuration:

  We used Apache 2.4 with the Event module as MPM with the following configuration parameters:

  • StartServers: 1

  • MinSpareThreads: 5

  • MaxSpareThreads: 10

  • ThreadLimit : 10

  • ThreadsPerChild: 10

  • ServerLimit: 1

  • MaxRequestWorkers: 10

  For PHP we used php-fpm with PHP5.6 with the following settings:

  • pm: dynamic

  • pm.max_children: 50

  • pm.start_servers : 8

  • pm.min_spare_servers: 4

  • pm.max_spare_servers: 16

  • pm.max_requests: 500

Results

With Pool Connections:

Without Pool Connections:

Result Summary

Conclusions

Results show a huge difference using pooled connections. Your application will be able to handle ten times more throughput with pooled connections.

Also you should consider that DBAs will thank you a lot:

• Database Memory usage will decrease a lot (at least 4 times less memory is used).

• DBAs know how many connections will be established from your application.

• High concurrent peaks of requests won’t crash the database, plus other applications using the same database should not be affected.

The best part is, we only need to make a minor setting change in the connection string; the API, and the code workflow remain exactly the same. You will get a huge performance boost with no effort.

References

http://www.oracle.com/technetwork/topics/php/php-scalability-ha-twp-128842.pdf

https://docs.oracle.com/cd/B28359_01/server.111/b28310/manproc002.htm#ADMIN12483