Exploding Heads and Very Good Questions

I shoulda apply the "'What the Hell Are You Thinking' Mind-to-Mind" transfer ring to myself.  Some days I just don't get this "why" I need to push b/c the Meetup I went to today - 2nd lecture blew my brain up into

                 itty

      bitty
                           pieces

Ignoring the fact that I don't know shit about making my own query language structure (so what the hell was I thinking about attending a lecture on one.  I mean, fer chrissakes, I specialize in wearing multi-colored nail polish.

But the talk on google's new product launch, BigTable, very, very interesting.  Especially in light of yesterday's question which asked about that I/O blender effect.  In today's lecture Paul Newson, a Developer Advocate with Google, discussed this in terms of performance impacts because of "bully tenant" / "noisy neighbor" problems.  Anyways, google's rolling out their cloud solution and they've got a developer's kit for you to test performance.  But google's done some algorithmic changes at they've achieved at least three 9s of <well, shit, I forgot exactly what, but I remember thinking, *three 9s, they need five*.>    What the hell.  The free beer was after, so I was only drinking diet coke and it was hot in the room - or I was having hot flashes.  Anyways, they addressed the latency issue to the last 99% of their users... oh yeah... They're up to three 9s on the latency issue when compared to r/w actions of Hbase & Cassandra because of their algorithms and the way they've distributed the loads.  White papers to come at the bottom of this entry.

I was less interested in it from a database problem than I was for the idea of a cloud system which might be able to keep up with the Need For Speed for mobility processing.  But really, what do I know.  Not a fuck of a lot.  That's why I'm out here on Walk About.  But whatever... this was the first time I'd seen the questions of latency addressed.  They also addressed CPU, but as another attendee mentioned, they seemed to have left off addressing questions of R/W actions, bandwidth, etc.  Be that as it may, the distribution, the scaleability of the architecture seems reasonable with the idea of distributed processing, backups, replication, and redundancy.  And now this understanding of the Need for Speed.  Finally.  Something where Cloud Computing looks like it will keep up with mobility demands.  Not that I know anyfink.

So, yes, new metrics need to be added for "tenant" problems - which appears to be the street jargon for the I/O blender effect   However, the fixes appear to be the same budgeting idea that Docker  presented and which Zulily works with, apparently successfully in deploying their iterative application upgrades.

And this link is to a picture worth an extra $500 google credit to add to the $300 when you sign up for their cloud.  And I'd suggest it just so you can get a physical feel.

But there are a lot of smart people out there whose world is different than ours is/was.  I don't know shit, but then I never thought I did.  I just absorbed information.

Original 2006 BigTable WhitePaper
More Big Table Stuff