← Back to the journal
How to

The May 2019 Platform Upgrade and How to Check a Hosting Change

Systm Web doubled available memory and CPU in May 2019. Use comparable cached and uncached tests to see what changes for readers.

On May 3, 2019, Systm Web announced that it had doubled the memory and CPU available to customers on its hosting platform. The notice expected faster page loads. It supplied no resource units, benchmark or per-site results. The archived announcement records what the provider said about the earlier service.

Test the type of request that used to wait

More available memory and CPU can help if work was waiting for those resources. A single visit to an already cached homepage will reveal little about a busy uncached route. Page generation, database waits and an external call can all contribute to the time a visitor sees. The WordPress slowdown guide explains how to look at these separately.

For a simple before-and-after check, choose a public page that normally uses a cache and an uncached action you can safely repeat on a staging site. Run a small set of comparable requests before the change and again afterward, with the same page data and cache state for each test. Keep the location and time of day comparable where possible. Record response time, errors and any host-reported signs of exhaustion. Saving both typical times and occasional slow requests can reveal a difference that a single average hides.

Work through the result

  • Uncached responses improve. Review application or host timings to see whether request processing spent less time waiting. Check that the result persists over ordinary activity.
  • Only the cached page improves. Check whether the first round and the second round actually used the same cache state. Record cache status beside each response time.
  • A slow action still waits on a remote service. Keep that dependency visible in your notes. Compare time spent on the remote call with the total duration before drawing a conclusion about hosting.

Suppose a cached page goes from 170 ms to 165 ms and an uncached page goes from 900 ms to 560 ms in repeated tests. Those are illustrative numbers. They point to the uncached transaction as the place to investigate and show why both paths belong in your test record.

Two hosting changes in the historical Upgrade listing

The archived Upgrade tag listed this May 2019 resource announcement and a March 2018 platform migration. The March notice involved domain changes for an earlier service. The separate January 2018 maintenance notice described a window when sites were expected to be unreachable.