Improve performance

How to adjust the self-hosted server to improve performance


Parameters that impact performance

The following parameters are described in Configuration and they influence in the performance of the Push Server:

  • batch_size
  • parallel_flushing_size

Note: The default value of batch_size has changed since v1.4.0, as the plugin now uses a different batching mechanism. It means that the old default value of 30 was changed to 1500, and it now has a totally different meaning.

Adjusting the batch_size parameter

If you increase this parameter, the memory used by wp-cron.php will increase. The estimated maximum used memory is:

Max Memory used = Max payload size * Batch size

Where the maximum payload size by default is MAX_COMPATIBILITY_PAYLOAD_LENGTH = 3052.

For example, the default value of batch_size = 1500 will use (1500 * 3052 / 1024) / 1024 = 4,36 Megabytes of RAM.

Execution time

The execution is split in multiple cron jobs in case the sending loop time reaches 80% of the maximum execution time limit defined by the php.ini.max_execution_time option. That parameter is described at ini.max-execution-time.

By splitting the job into multiple cron jobs, the execution time will be longer. This is because of how WP-Cron works. So if you want to improve performance and gain some more time, consider using a long execution time limit or make it unlimited for wp-cron.php.

Adjusting the parallel_flushing_size parameter

This is the batchSize parameter defined in the internal web-push-php library.

The default value in the plugin is a low number for maximum compatibility with different server configurations, however we recommend adjusting it to achieve maximum performance.

For example, a value of 50 means that there will be 50 asynchronous HTTP requests at a given time during the send of the notifications. This can create high spikes in memory and CPU usage.

You need to be cautious because increasing this number to a high number (bigger than 1.000), will mean potential Out Of Memory errors and weird exceptions in the downstream components. In any case, it's recommended to monitor the logs when adjusting this parameter.

When you use a value that is higher than what your server can handle, you'll see an increase in log entries like these:

[27-Aug-2021 19:41:36] WARNING: [pool website.com]
child 4593 said into stderr: "PHP message: ERROR |
Failed to send one notification, error:
cURL error 60: Issuer certificate is invalid.
(see https://curl.haxx.se/libcurl/c/libcurl-errors.html) 
for https://fcm.googleapis.com/fcm/send/XXXXXXXXXXXXXXXXXXXXXXX"

Example

In a DigitalOcean droplet with 2 Gb of RAM and 2vCPU, it can send around 13.801 notifications in 56 seconds, using the following parameters:

  • batch_size = 15.000
  • parallel_flushing_size = 200
  • memory_limit = 256M
  • max_execution_time = 256

When the parallel_flushing_size was increased to 300, lots of notifications failed with cURL error 60 because the server couldn't handle that amount of concurrent requests.

Edit on GitHub