Google SERPs for
Nigritude Ultramarine are up and down, up and down... When I hit refresh button some of the sites change their place by as much as 10 positions. This happens every day in the evenings when Google updates thousands of their servers with a new index. For several hours when you run a search you play a lottery as it may endup on server with either the old or a new index.
I would never complain, understanding how much efforts it takes for Google to update tens of thousands of servers in just a few hours. However, there is one thing which looks like a serious flaw in the system and dynamic nature of
nigritude ultramarine results, which change daily, only underscores that.
The problem is that when you go from one result page to another you never guaranteed to see continuous results. Rather you see a piece from yesterday SERPs, then a couple of today's SERPs, then yesterday's again. This leads to some sites to be missing from
your session and other sites appearing twice. It looks like Google never anticipated potential problem between nondeterministic ballance loading and nonuniform index across the servers in a datacenter.
Just a year ago when they were doing daily index updates, results would change almost instanteneously for the entite datacenter (Google has at least 12 of them around the world). They would update three datacenters at a time in an half-hour increments. They also would use DNS to redirect all traffic away from updating datacenters about 15 minutes before the update and return it (based on the user locality) 15 minutes after update. This gave a very high chance that any reasonable user session would get consitent SERPs.
My theory is that they used to have enough disk space on the servers to keep both yesterday's and today's indices at the same time and after propagating the indices for a few hours it would take them only one simple command brodcast to switch entire datacenter to a new index. Apparently, now they use asynchronous index update. Possible explanations for that could be:
1. Server disks could no longer hold two indices at the same time.
2. They no longer broadcast index syncroneously (if they ever did).
3. They no longer have a single index (then why same query is dancing?)
4. They still have single master index, they broadcast it, but inependently over multiple network segmentes.
5. They intentionally try to confuse us,
Nigritude Ultramarine SEO folks.
Anyone has any ideas?