HoW are data centers upgraded?

Toggle sidebar Toggle sidebar
A

Atreyo Bhattacharjee

Commendable
Feb 7, 2017
207
0
1,690
How do data centers upgrade their hardware?
Lets say there is a huge technological improvement allowing for much more data to be stored in a hard drive, how would a big company like Youtube or Steam upgrade their multi petabyte data centers? Would their servers be down for extended periods of time?
 
Solution
T
You'd be absolutely amazed how much old hardware is still in use. One of the reasons they have an Enterprise classification for hardware. Anyway, they typically add capacity as they go, rather than replace. There isn't usually downtime even due to drive failures, because of redundancy. Also, they tend to use hot-swappable drive enclosures so if a drive fails they simply pull it out and slap a new one in. Here's a couple pictures to help explain things:

https://www.backblaze.com/blog/wp-content/uploads/2014/08/blog-270TB-storage-pod.jpg

https://www.backblaze.com/blog/wp-content/uploads/2014/08/blog-petabyte-comparison.jpg
Sort by date Sort by votes
greens

greens

Splendid
Jan 27, 2012
3,773
7
26,965
That is too small of a scope.

Think of it more like a trucking company. When a new great awesome truck comes out, do you sell your 25 old trucks to buy 2 new expensive trucks? no of course not. You keep your 25 old trucks and when the time comes for you to need more trucks - you buy the best trucks you can get.

Same with datacenters. You add a center or a cluster and you use the best technology you can.
Only when the old stuff can't keep up, doesn't meet spec, or breaks down - does it ever go offline.

They go offline slower than they are added so you end up always having enough room - or enough trucks to get the jobs done, while slowly improving the "fleet" over time.



 
Upvote 1 Downvote
T

TMTOWTSAC

Distinguished
Jun 27, 2015
1,462
7
20,215
You'd be absolutely amazed how much old hardware is still in use. One of the reasons they have an Enterprise classification for hardware. Anyway, they typically add capacity as they go, rather than replace. There isn't usually downtime even due to drive failures, because of redundancy. Also, they tend to use hot-swappable drive enclosures so if a drive fails they simply pull it out and slap a new one in. Here's a couple pictures to help explain things:

https://www.backblaze.com/blog/wp-content/uploads/2014/08/blog-270TB-storage-pod.jpg

https://www.backblaze.com/blog/wp-content/uploads/2014/08/blog-petabyte-comparison.jpg
 
Upvote 0 Downvote
Solution
jsmithepa

jsmithepa

Polypheme
Jun 14, 2014
13,096
263
60,840

Simple. Because any company of any size have mirror sites, mirror data centers, so they schedule this when traffic is light and they down the site that needs maintenance and things keep running on the remaining mirror sites. When the upgrade sites comes back on line it does through a sync process, but all these stuff is more of less transparent to the customer.
 
Upvote 0 Downvote
You must log in or register to reply here.

TRENDING THREADS

Latest posts

Moderators online

Share this page

COMPANY
RESOURCES
FOLLOW
Top Bottom