Forum Moderators: Robert Charlton & goodroi
Stop looking at [64.233.187.104...] aka [gfe-jc.google.com...] when you should be looking at [gfe-eh.google.com...] aka [72.14.207.104...] instead...
On the latter DC there is a big cleanup of older Supplemental Results in progress.
[edited by: tedster at 3:53 am (utc) on Aug. 18, 2006]
These were active [webmasterworld.com] back in June and are still active now.
I believe gfe-eh to be sort-of where we are heading, and gfe-gv to be where we have been.
cheers,
russell
Let me put some detail in first...
Several years ago, some pages of content were amended to remove a particular word from the page.
Google reindexed the pages and they continued to rank for various terms as normal results. There were also some duplicate content issues where the "exact" same pages were published on other domains, as well as there being several ways to get to some of the pages using slightly different parameters. Those all got amended too. The word was removed from all of them. Some of those URLs were marked up as Supplemental Results, and some were not.
After a while, most of the duplicate URLs vanished from the SERPs, but some stayed around.
If you searched for current content, you got a normal result, but if you searched for the "removed" word, you got the same URLs back, but this time as Supplemental Results. The "removed" word also appeared in the snippet.
After a year, Google dropped all those Supplemental Results and you could no longer find any evidence of the "removed" word on that site, or related sites.
That was more than a year ago.
Yesterday, some of those URLs are back in the index when you search for the "removed" word. The "removed" word shows in the snippet too. The "removed" word is NOT on the real site, and has not been there for about two years.
This is the kicker: The URLs are NOT marked as being Supplemental Results.
What is going on?
Most have not. The majority remain "gone".
If it is rollback, then it is a very long way back. I have never seen anything go that far back before. Nowhere near.
However, it's the fact that the very old data is not tagged as Supplemental that interests me the most.
Also, now the page numbers are accurate, despite about half being supplementals.
I also predict at some point a google version of the waybackmachine.
For may webmasters, google is already the "waybackmachine"
I would say that the DCs are calm and somewhat relatively quite.
Maybe we have already witnessed a modern Google update a la "Smooth Operator" :-)
Well.. we shouldn't be surprised at all if that was the case. Uncle Matt The Cutts has indicated that we shouldn't expect ton of movements on the DCs ;-)
The sad side of all that is we aren't going to have in future updates where our Brett gives them specific names , unfortunately. I will for sure miss that.
209.85.129.104 is serving UK results at the moment - is this a new DC?
Nothing new under the sun of mighty GOOG :-)
209.85.129.104 is showing something like the results of [72.14.207.104...] at the moment, IMO.
Its Google birthday today!
Don't know whether we can also consider the same date as birthday of all Google's datacenters. I assume so.
Happy Birthday To ALL Google's GREAT FANTASTIC datacenters including [72.14.207.104...] of course :-)
And of course a special reseller Google birthday song.
Oh Google.. You Are Simply The Best.. Better Than All The Rest!
It seems that we have to live with the serps of 72.14.207.*** for the near future at least. Because I see the new infrastructure DCs have been relatively quite during the last few days or so. The remaining movements mightbe only to see propagation of the the serps of 72.14.207.*** to the rest of the DCs, or most of them.
To those of you kind fellow members who are doing well on the current 72.14.207.*** ... congrats.
To the rest... lets wait and see what will happen at the end of October. Maybe another White Revolution, who knows ;-)
They wouldn't put old data on brand new machines would they?
They wouldn't put old data on brand new machines would they?
Of course not.
And now we may as well start the new period of very interesting datacenters watching :-)
And thanks for your great generous contributions to the thread, g1smd!
With the possible complexities involved with infrastructure design, IP addresses mean very little, any external facing ip address can be routed to any internal ip address, anywere, and you would never know which boxes really return a response to a query.
Are you watching 50 plus datacenters, serving up 3-4 sets of results, or are you watching 50 ip address ranges for 10 load balancers that distribute to 4 different datacenters with different results, or are you watching 50+ ip address ranges for 20 load balancers with 100 datacenters throughout the world with a mix of 3-4 sets of results, and on and on, because any of these are possible, and can be rearanged on the fly, and you would never know.
Once again, this is IMHO
Back to watching.
WW_Watcher
[edited by: theBear at 4:59 pm (utc) on Sep. 28, 2006]
"WW_Watcher would you say, the only thing out there that might be static would be some MAC addresses? Even that is iffy if you know what I mean? ;). "
Yes I would agree with that, (not normally changed for possible confict reasons), but anytime a nic failed, the IP address might not change, but the MAC would. Anytime a load balancer was replaced (with multiple nics) the IP address might not change, but the MAC would.
BUT back to the original issue, these are ip addresses on load balancers, probely not actual datacenters, I would expect every true datacenter (building full of boxes) would have many different IP ranges for load balancers facing the internet for us to see. With DNS smoke and mirrors finding the closest (fewest number of hops) to the datacenter nearest us, adding and removing ip addresses for load balancers at will. After your request has reached a load balancer, your request could be routed to a pool of more load balancers, picking the least busy one (based on number of connections, or traffic load, or CPU utilization) of which the least busy would then route your request to the least busy server in the pool. Each server in that pool, can also be a swimmer in another pool serving up responses out another load balancer. At any point during the routing between load balancers, they could choose to send your original request over to another totally different Datacenter(building full of boxes), within their network and you would never know.
To confuse it further, I would expect that each of the true boxes serving up responses to queries are load balanced possibly to other boxes that actually service the databases. So a single box that responds to a query could be asking that query to multiple databases, possibly with different data, or algo. The possiblities are endless, and only G knows for sure are they are setup at any given moment.
Now I am lost, confused and forgot what my original point was, I asked, but was given 3 different responses....
I'm better off watching (not the datacenters, time is better spent lately revewing what others have discovered about the current serps, and the making the changes reqired to properly instruct search engines in general how to view and index your site. As the real professionals have stated over and over(not me, I normally do not post anything, and as a "webmaster" I consider myself almost clueless, successful, but almost clueless))
WW_Watcher
I suspect that there may be another update/shuffle/everflux coming this weekend - don't know why but I do...
What are the current data ceneters G is using and is "Big Daddy" still producing results?
Thanks in advance!
ARC
