Erichk Knaar
Caldari Noir. Noir. Mercenary Group
|
Posted - 2010.05.08 17:04:00 -
[1]
Originally by: Mme Pinkerton Edited by: Mme Pinkerton on 08/05/2010 16:34:03
Originally by: 45thtiger 0109
What do you do say if you had bad weather in your area where you live and the power went out or you phone lines went down who do you blame then.
Do you blame CCP, your ISP, or your electric company.
Me i would not blame anyone & sometimes nature is unpredictable.
I would of course blame my electric company (the one providing the infrastructure not the one selling the energy, if the two happen to be different) - because they know perfectly well that bad weather can happen and didn't build enough redundancy into their power grid to make it robust against this kind of failure.
Stuff happens but intelligent people take precautions and have failover systems in place in case anything goes wrong.
How do you think people get the 3-5 nine SLAs they keep touting as the gold standard in their rage posts? caertainly not by relying on good luck... servers have 2 power supplies (powered with electricity from two different energy companies), UPS's, 2x the necessary network cards, RAID setups for HDDs, hotplug so you can replace as much as many components on the running system as possible (try swapping a HD or a network card on a running home PC), your database servers are clusters with replication between them so if one goes down you can still get your information from one of the others, you have diesel generators in the backyard, off-site storage for backups and if you are really serious a second datacenter at least a few dozen kilometers away.
And all of this doesn't prevent anything software-side going wrong
"nature is unpredictable" is a heavily over-used excuse - nature is unpredictable in when stuff will happen but it is usually very predictable in what can/could happen.
With software it is much more difficult to get an idea what possibly could go wrong (and thus more difficult to ward against these scenarios) and I think in the light of the uniqueness & complexity of CCP's setup (remember practically all of the software your webhoster runs to support his service has been debugged on hundreds of thousands other machines with similar configurations before) 96% is pretty good.
But if my power would go down just because of bad weather I would get very, very angry at my power company (luckily I cannot remember that happening in the last 10-15 years).
edit: to give this post some kind of direction - I don't emorage when CCP needs a few hours to locate and fix some nasty software bug (because unit tests and fuzz testing can only take you so far); but I get a little angry when CCP tells us several times in a row that their MSSQL failover crashes every time they need it (some months back).
The first scenario has a "**** happens" quality and there's not much you can do (if you run the same software on your production and failsave systems); the second would make me angry because it would/should be preventable (and apparently someone had been planning for that kind of scenario but forgot to properly test the last few meters beforehand).
Pretty much spot on. 5 9s is extremely expensive to achieve. Extremely. I'd much rather have CCP spend the money I give them on more capacity etc. Its a game, not a system carrying launch codes.
|