+100.

Problem, though, is the perverse incentive existed in the old pricing
model also -- if one assumes that low infrastructure investment caused
handlers to run slower. I think that was/is the case otherwise there
would be no incentive for GAE to throttle your app when times are
tough.

Of course with the new instance pricing model, the returns on this
"algo-driven" revenue module are potentially spectacular rather than
proportional. If GAE can slow down just enough every half hour or so
such that the scheduler kicks in a new instance for just a few
transactions, then the bean-counters in Finance will be so ecstatic --
I mean really, really ecstatic.

Overall, I try to maintain a sense of humor about how this is rolling
out. Seen a lot of similar situations. Engineers get together to kick-
off what they envision to be a highly disruptive new product. I
imagine Ikai, Guido, et. al. having drinks saying things like, "Yeah,
we can totally slay AWS by pricing according to actual resource usage.
This per hour instance charge is so bogus." They roll with this for a
few years in beta, and everyone using GAE is saying, "Wow, I pay for
what I use. How cool is that. Google rocks."

Then things move toward the product kicking into a P&L group. Larry
says, "Let's have the guys in Finance take a look at this." Previously
if you mentioned GAE to the guys in Fiance, they'd think the SEC had
cooked up something new related to GAAP. The guys in Finance find out
more about GAE, and set up a meeting with Larry, Ikai, Guido, et. al.
Then kick off the PPT on the big screen, and the first slide reads,
"You're leaving a S**T-load of money on the table." They go on to say
things like, "The only guys that can do this are us, MSFT, and Amazon.
Why exactly would you want to make this a price-competition?"

Quite frankly, when the case is made like this, Larry has to side with
Finance. Being nice to small developers is not part of his job
description. Being nice to shareholders is.

So, GAE is likely to cost a bit more than AWS because of the nice
things it does that have been pointed out in other threads. Azure
falls in line, and nobody competes on price, but other things like how
well integrated X is because we have Y. I think Chris Dixon wrote a
blog post related to this -- but that's a very vague recollection.

Course it could all be about RAM as has been said.

Cheers,
Steve

On May 22, 12:38 pm, Anders <[email protected]> wrote:
> Another problem is that with the new GAE price model there is no incentive
> for Google to make the scheduler perform better and better. In fact the
> opposite is the case: the more instances the scheduler creates, the more
> money Google makes. Sounds like a really bad strategy from a customer
> perspective (and from a long-term Google quality perspective).

-- 
You received this message because you are subscribed to the Google Groups 
"Google App Engine" group.
To post to this group, send email to [email protected].
To unsubscribe from this group, send email to 
[email protected].
For more options, visit this group at 
http://groups.google.com/group/google-appengine?hl=en.

Reply via email to