does anybody here understand what uptime ...
It's gotten to the point where it's bothersome because I'm only capable to achieve 99.73 ( and I run redundant systems ) and my SLA is based at 99.63 ( just need that extra 10 until I can do triple redundancy )
Now I would like to point out that people should look for with these guarantees.
the most important thing to know is the time terms of the guarantee. there are daily, weekly, monthly and yearly. I've only see once a daily SLA term. the most common one is monthly, and the one to avoid is yearly.
now some pointers
1) first there is the network
the odds are that the network will never go down and also if it's a bgp network setup then the odds are even less. One of these great tricks is that a web host will offer 99.999 on the network side, but don't offer anything beond that. SO if the site goes down due to a hack, the guarantee is still valid and there is no recourse to the customer. also you should look to see if they have a minimum ping time statement within that guarantee, most will not. but the best ones do.
2) equipment
equipment fails, is it covered in the SLA? Some firms do cover equiment failures in thier SLA. Look for that ( also see web request below )
3) Web request avalibility
as for some SLA's, the one that you really want is one that had the guarantee based on the avalibility of the site being reachable by end user. This type I have see labled as HTTP SLA or high avalibility SLA.
These superseed the other 2 types of SLA's why? they are most likely redundant systems in an envionment that wants to make sure that your site never comes off line unless you bring it off line.
in true high avalibility systems cost money, Lots of money. currently I had to help a person reach a specific goal and I knew I could not do it on my systems and had to go to another datacenter for help. this was his configs of his request and the cost is at the end
unix platform, zues web serving
ping time for usa lower 48 = 35 ms
london = 52 ms
Melbourne = 98 ms
hong kong = 70 ms
terms for avalibility
120000 page request per hour @ 99.9 % avalibility
or 80000 page request per hour @ 99.95 % avability
his configuration ended up with
11 servers
dual routers
dual load balancers
sweet SAN set up
weirdest bgp set up I've ever seen ( way over my head stuff )
cost per month, on a 1 year contract
99.9 = $ 8,000 plus $1.68 on the B/W per gig
if this guy wanted 99.99 the cost would have been about double, 99.999 and 99.9999 are in the 250K plus range, the only people I read about, that have 99.999 and 99.9999 are people that will loose upwards of $ 5000 per minute of downtime or lack of avalibility. the down time has cascading effects. ( think of an auto plant line, if for some reason the servers slow down and can not respond fast enough to the lines needs, the loss in productivity keeps on escalating, last I read, the average GM USA plant downtime cost about 75K a minute.
