استوریج 3PAR HP
With our previous installation project (part 1 & part 2) we installed the complete HP blade server infrastructure with Generation 9 blade servers and Virtual Connect FlexFabric modules up&running in the rack. They have been on display in the PROOFMARK portal ever since and quite a few of you have already seen them in action. Now it is time to add some serious storage to the picture!
First, I would like to have a moment and give a short introduction speech about storage array concept and the vast advantages of having a centralized storage solution instead of so called direct-attached storage or DAS, “normal” hard drives that are installed directly into the server.
Lets imagine a situation with 10 servers, each of which need 1TB of disk capacity. Using the most traditional approach, we would install individual hard disks directly into the servers. For simplicity, let’s use one 1TB hard disk in each of the ten servers. Now, the first and immediate problem is if one of those hard drives breaks down. In the worst case we will lose all the data in that hard drive if we didn’t happen to have a backup copy of the data and even if we did, there will be a loooong service break before we get the new hard drive, operating system, applications, services etc installed.
Then there’s the capacity utilization problem. At the time of acquisition, it is very hard to predict how much disk capacity will be eventually needed. That’s why people tend to buy a bit more disk capacity up-front to make sure they don’t run out of capacity while only few servers will ever use all that capacity. If, for example, one server is using only 500GB of the available 1TB, you’ll pay for the other 500GB for nothing. On the other hand, if that 1TB is not enough for another server, there’s no way to borrow that leftover 500GB from the first server and you’l have no choice but to buy more or bigger hard drives, shut down the server, install the hard drive and boot the server up again.
Performance-wise 10 disks must be better than just 1, right? Correct. Kinda. Only not when they are installed in the different servers. If there’s a very disk intensive task executed on one of the servers, once again that one server can not borrow disk performance from other servers’ disks.
Now, instead of 10 servers imagine the same situation with hundreds or thousands of servers. Would you volunteer administering that? I wouldn’t.
There’s a lot of built-in redundancy in storage arrays so losing one or two disks usually doesn’t matter at all. All you need to do is replace the broken disk at any convenient point while the servers will never skip a beat.
Of the total 5TB we have in the storage array we can initially allocate, say, 200GB for each server. Servers running out of capacity is not a problem either since it is very easy to allocate more capacity with great granularity from the storage array. Allocating less capacity is as easy if needed. There’s also a ton of other features that optimise the capacity utilization like thin provisioning, compression and de-duplication but I’ll leave those for the “advanced course”.
Sophisticated storage arrays can also leverage all disks for all read and write operations by any server so performance is also greatly boosted.
How about administering hundreds or thousands of servers using storage arrays. No problem, show me to my station!
And THAT, ladies and gentlemen, was a short introduction.
Now, let’s get down to business! What we have today is two brand new HP 3PAR StoreServ 7200c storage arrays.
