Posted by: danycrit
Posted on: 2007-08-02 05:20:00
Hi All.
I see many posts on the dreamhost status page related to hardware failures. Anyone know what is the medium lifetime of the hardware at dreamhost ?
It would be nice if the notifications states somewhere for how much time the failing component was in service.
Regards.
Posted by: twsepper
Posted on: 2007-08-02 05:59:00
Servers are machines; machines fail. A machine can fail after two months or two years.
Posted by: scjessey
Posted on: 2007-08-02 06:11:00
"I've just picked up a fault in the AE35 unit. It's going to go 100 percent failure within seventy-two hours."
Hardware breaks, and it gets replaced or fixed. DreamHost will replace or repair faulty hardware as soon as a fault appears, usually with no loss of data. Who is going to leave a "failing component" in service?
-- si-blog --
Posted by: Lensman
Posted on: 2007-08-02 07:15:00
Do you think they should do a study and publish a paper like the famous Google paper on disk drive failures?
If you have a suggestion, submit it through the control panel. But what value do you think it would have to us?
Or are you really just interested in what DreamHost's preventitive maintenance and renewal plan is all about?
Posted by: seiler
Posted on: 2007-08-02 07:40:00
By now, I'd guess Dreamhost has around 1,500 servers or so and I believe they add at least 5 each week.
If a unique server were to explode every single day, in order, you'd go over 4 years before the same one failed twice.
Of course, it doesn't work that way, but it should put it into perspective why the status site lists so many failures. Especially when compared to hosts that either don't announce failures (most?) or those that are small and only have a server or two.
Posted by: dinochopins
Posted on: 2007-08-02 09:09:00
In reply to:If a unique server were to explode every single day, in order, you'd go over 4 years before the same one failed twice.
Of course, it doesn't work that way, but it should put it into perspective why the status site lists so many failures. Especially when compared to hosts that either don't announce failures (most?) or those that are small and only have a server or two.
That make sense.
Dino