How do Google's data centres work?

Different data on different centres

[ToyTalk]

I have seen a lot of talk about Google's data centres on this board, plus some lists.

When I visit these data centres I see wildly differing numbers of indexed pages for my site - usually many, many more.

It's frustrating to see a data centre with 64, or 96 or 101 pages from my site indexed, only to see 4, or 8 or 7 when I simply tap in www.google.com and do a site: search.

Can anyone shed some light on why data centres would carry such differing data?

I'm still awaiting full recovery from the August turmoil on Google, when I went from about 164 indexed pages to one. My indexed pages fluctuate most days, between 4 and 7, and it's always the same, old pages coming and going...

I find it confusing that one day a page is considered worthy of inclusion, and another day it is not.

[reseller]

ToyTalk

Can anyone shed some light on why data centres would carry such differing data?

The reason is different datacenters get different data at different times. Sometimes even different filters applied to some datacenters.

[g1smd]

Across their 46 datacentres (the 46 active Class-C IP blocks) Google usually has at least 2 or 3 different sets of results in play, and often there is some sort of "experiment" running at a single datacentre from time to time too. Google compares what visitors do with the different result sets, and makes changes all the time.

When a big change is coming they often put it on a single datacentre and play around with it for several weeks or more, before eventually copying it to many other (but rarely all) datacentres.

As for the rest; ask Matt Cutts...