Why are results within same datacenters are different?

[AthlonInside]

There are a few variation of SERPs from different datacenters such as EX vs FI vs CW ...

What come out my mind is that they (The GooglePlex Guys) are testing different algo on different datacenters so they can evaluate a few different algo at the same time.

You might have other comments. I am interested in hearing all of them!

[Brett_Tabke]

You are not connecting to "center" - you are connecting to one of 50,000+ machines.

[Powdork]

You are not connecting to "center" - you are connecting to one of 50,000+ machines.

That would explain differences in serps while trying the same search consecutively at one 'center'. But there are clearly different sets of results that are dispalyed consistently at individual 'centers' for a period of time. Whether the differences in serps among centers represents a different algo, filter, index, or backlink set is up in the air, but I suspect we see all at one time or another.

[Brett_Tabke]

> But there are clearly different sets of results

We don't know how googles index is distributed - let alone how their load sharing system works. It could be sequential, round robin, random, or fixed to serve a particular ip range. eg: just because you get the same index a few times...

[kaled]

There is no doubt that individual data centers deliver results sometimes that can only be explained by the use of different algos. As for results varying within a data center, I seem to recall (but I could be wrong) that each data center is divided into two banks, so a single data center could offer variable results from time to time.

Kaled.

[Powdork]

What I am referring to can be seen using the dance tool viewing 10 datacenters. You will see different sets of results (at least during updates) that 'apparently' are sets. Usually, when its dancing, I will see three different sets and I typically break these down into; new index, new index (no filters), and old index. These names are just names rather than truthful descriptions, but...

[Powdork]

And now we've reached the part of the dance (fat lady sang) when all eleven datacenters are showing the same set of results.

[Gus_R]

Hi, for my main phrase (two keywords) I've differences between 2 and 6 million results.
The lower number is 6, higher: 11.5
Really weird.

[Wired Suzanne]

Results are going like a jojo now. www-in changed very much in the past hour or so and now I see these results in every 'center'.

Very irrelevant results, I have to say...

(The results are 'dancing'. Yes! That is the correct word: Dancing!)

What's going on?

[eraldemukian]

with the number of commodity hardware nodes that are involved in the solution for a given query, it is not surprising that results vary at times.
There will always be machines gone. GFS seems to deal with this nicely, but it could have an impact on results. (Fresh data beeing replaced with older one, until the node recovered etc).

Google is not as monolithic as a single computer with a single data set, since its a little bit more than that.

[sidyadav]

Same problem here, well, not a problem but a find.

As I remember for a query about 4 months ago, it returned with 13 million results, after somedays , got upto 15 , 16 , 18 , 20 , 22 , 23 , 25 and now guess what? its back to 12 million!
How can this happen?

(How I know when the query changed? The query I tried/still try to check a search engine's revelancy was/is my first name!)

Sid

[claus]

just an hour ago or so i did a search for "webmaster" to find around 86M results. Some minutes later it was showing 146 million.

Right this minute it's 190 million: [google.com...]

I guess an update is going on.

/claus

[AthlonInside]

Seeing the GOOD side ...

Sometimes it fears me when I found my listing drop so much, but later I found out it is just one or few of the datacenters. So I still have listing that rank high in other datacenters. :)

This really make me feel a lot better.