When a web client needs to periodically refresh data held on the server there are generally two approaches. Interval polling (“short polling”), which is most commonly used, where the client repeatedly re- connects to the server for updates, and a technique in which the HTTP connection is kept open (“long polling”). Although work exists investi- gating the possibilities of long polling few if any experiments have been performed using an Apache MySQL PHP (AMP) stack. To determine the potential effectiveness of long polling with this architecture an ex- periment was designed and conducted to compare update response times using both methods with a variety of polling intervals. Results clearly show a marked improvement in timings with long polling, with the mean response time down to 0.38s compared to a mean of just under the polling interval for short polling. Issues of the complexity and load implications of wide use of long polling are discussed but were outside the remit of this experiment owing to a lack of resources.

Notes

Note that this is a just basically a subset of work performed as part of work performed for my MSc dissertation.

Links

Available from Purplepixie Labs: http://labs.purplepixie.org/content/cutting-long-held-php-2015.pdf

Purplepixie Permalink: http://go.purplepixie.org/sleepjax-long-held-http-request