Hey,
I wonder if people can update me/us on how the
integration with the
Amsterdam cluster is going. Are there any lessons? Are we still
committed to our multiple-data-center strategy? Is there anything new
which we have learned?
Sure, that light speed sucks, if it concerns information, and that having
content as close to people, increases user satisfaction a lot. Having
squid caches in distributed locations (so at least majority of
requests do not cross oceans) is a really win-win situation.
We did enjoy having good hardware with good management possibilities,
so... we do not need man on site for most of operations and we do not
need various power switches (APCs) or serial consoles.
We still didn't solve problems of distributed apaches or distributed
databases, but that requires quite careful planning and lots of
determination and motivation, as it would complicate our bits a lot.
I am working with Yahoo to finalize the details of the
South Korean
facility... what have we learned which might help with this?
Well, first of all, Asia needs this and it will definately improve
the service. Though, as we have issues with speed of light,
we're not sure yet how to deal with it in most effective way.
What might we ask of a large potential donor next?
We are not small anymore. We might be wanting to use
industry-standard technogy for our database failover
(eh, some mid-range storage;-). Maybe. I'm just raving ;-)
As well, we might be wanting to have high-availability network
with redundant switches for our main servers (did I say we
need new Cisco 3560? :)
It will be much more interesting to analyze our future needs
after 1.5 deployment and tuning. But we might be wanting
to go high-availability soon. And sure, increase our capacity.
We still have our media storage orphaned and unaddressed,
but hopefully Mark will be trying to put all his summer efforts
there.
All this higher availability stuff is not only to please our users
(though, lately they have been spoiled ;-), but for less stress
and more concentration in development.
And sure, we will need more application and db servers in
Florida, some day. As it is summer season and usage is slightly
decreased, our next order might be in... September?
Maybe we'll have enough of planning time in Frankfurt,
and I hope as much as possible of development and tech
people to be there.
Anyway, we've been doing not that bad lately, so it might be
a bit more difficult to do a fundraising.. :)
Domas