Advice please - on Cloud API performance techniques

[chum63]

Hi - me again, but testing the local WURFL vs the Cloud - sorry

I have set up our system so I can switch from it using the local WURFL and WURFL in the cloud and am doing some comparisons. The service we have keeps track of how long it takes to process in milliseconds and this gets logged with each transactions. Obviously the service does more than just getting the device information, but I found that when using the local version my range of millisecs was from 18 to 27 millisecs - so pretty static. But when I switch to the cloud it ranges from 7 to 381 - obviously the 7 is nice, but I assume this is due to using cache rather than making a service request to you. My instinct is that when the service gets called after a while of downtime it takes longer to make the first request, but after that it gets quicker. I have been using about 3 different mobiles for the testing = 2 x android and 1 x windows.

I have attached snapshot picture from our db and you will see in the range of millisecs in the first column and 1 in the second column when the cloud is being used and a 0 when the local version is being used.

Do you have any advice or suggestions/references on server/coding that will remove the fluctuation in time taken for the cloud? Anything else I can try or do - to stop the 300 millisecs?

Cheers, Chris.

[kamermans]

Hi Chris,

Your assumptions are correct. The cloud can be faster than the OnSite API due to it's lightweight caching, but most requests will be longer. If your servers are in the US or Europe on a decent backbone, you should be averaging <100ms round trip for requests, however, some of your lookups will require a round-trip DNS request, this could effectively double the delay for some requests. Finally, there is always the issue of normal network issues since your requests are going across the public internet.

[chum63]

OK - thank you for that - appreciated.

Cheers, Chris.