Portal Home > Knowledgebase > Industry Announcements > Web Hosting Main Forums > Providers and Network Outages and Updates > 4 512GB SSDs in RAID 1 vs RAID 10
4 512GB SSDs in RAID 1 vs RAID 10
| Posted by kuchu, 10-28-2016, 09:06 PM |
|
We have a database server with a lot of user generated content and we are taking extreme measures to make sure that there is minimal chance of losing it. This data is just a few gigs in size and we wanted to go with 4xSSD RAID-1 to have a 4 mirror redundancy but the techs at the data center are telling us that we should go with RAID-10 instead. Trying to understand the pros and cons here.
RAID-1
- More redundancy, makes good use of all 4 disks because we don't need 1 GB of storage space
RAID-10
- More common? Better read and write performance?
|
| Posted by HumaneHostingOwner, 10-28-2016, 09:10 PM |
|
Right, to be honest just do what your needing (Redundancy on "High"). The DC tech probably assumed you're using it just like their 50k+ customers are. With that said it better to make these decisions on your own. Given that it's a unusual requirement.
|
| Posted by Crafted-Servers, 10-28-2016, 09:13 PM |
|
You are correct about better performance for RAID-10, however having 4x disks in RAID-1 is quite wasteful in terms of usage disk. Please also keep in mind RAID does not mean backup, nor does it protect against file system corruptions ect. If you are ok with the waste, I would say using RAID-1 /w a hot spare and single disk as on-server backup (as well as off server) would be your best bet.
.
.
.
|
| Posted by SkylakeDC, 10-28-2016, 09:15 PM |
|
If you have RAID 10 Data Recovery is very fast. If something goes wrong with one of disk you can rebuild your data very fast. Thats the most advantage of Raid 1+0.
|
| Posted by HumaneHostingOwner, 10-28-2016, 09:19 PM |
|
The performance can't be beaten however it does come at a "price".
RAID 1 ANY of the disks can fails and your fine.
RAID 10 if ONE disk fails, your fine no matter what. However any more than that and you're risking your data. This is because even though too often it says "two" can fail... It not really that's simple. That is because if two of the "wrong" ones goes then ALL data is lost.
|
| Posted by BoomHost-Kumar, 10-28-2016, 10:41 PM |
|
Go with RAID10. Setup a proper backup strategy to backup all data to off-server.
|
| Posted by TMS - JoseQ, 10-28-2016, 11:18 PM |
|
I don't believe the controller will let you have a RAID-1 with four drives where three of them are a mirror of the 1st. It is wasteful and unnecessary. At best, you can just have a RAID-1 and have a hot-spare configured so that it jumps in if one fails. That way you can use just three drives and replace any one that fails without downtime.
RAID-10 will also allow you to lose up to two drives (depending on which two) while also giving you twice the performance of RAID-1.
If you want redundancy but speed/size doesn't matter, then use RAID-5 and/or smaller SSDs.
|
| Posted by TMS - JoseQ, 10-28-2016, 11:19 PM |
|
And as mentioned above, ALWAYS keep backups. RAID is a protection strictly against hardware failure. It has no regression capability, it does not protect against user error or security weaknesses. Always, always, always back your data up outside an array. When? Always. But what if- NO! ALWAYS.
|
| Posted by Mike_A, 10-29-2016, 01:24 AM |
|
I'd recommend putting RAID-1 on three drives and mounting the third as a soft backup disk, so you can store a data backup that's separate from RAID.
|
| Posted by kuchu, 10-29-2016, 02:30 AM |
|
Thanks for the replies.
First of all, I reassure everyone that we are definitely planning on doing offline backups and have all the necessary arrangements.
A couple of you mentioned doing RAID 1 with two disks and keeping a third around as a warm backup. Is this something we will have to manage ourselves or is that something you can do as part of a RAID setup?
Again, one of my reservations about doing RAID 10 has been having 1TB of storage, which is a complete waste space for our use case and since even 500GB is way more than we need. Unfortunately, the DC does not offer 256GB SSDs so it is not possible to set up a half TB RAID 10 setup. I'm assuming the people who are still suggesting we do RAID 10 are saying do it for the fault tolerance (being able to lose 2 disks, albeit the right 2 rather than the more cost effective use of disk space.
|
| Posted by Crafted-Servers, 10-29-2016, 02:45 AM |
|
Hot Spares as they are called are setup during the configuration of the array in most cases. They are there, just sitting and waiting. If the controller detects a bad disk and removes it from the array, it would then automatically take the hot spare and 'promote' it to an array member, the array would rebuild and you'd be safe again. You would then remove the bad drive, replace with fresh drive and that would be your new hot spare. For hardware RAID most of this is automatic, for SW RAID in the OS (mdadm for example) there may be some manual intervention here and there.
|
| Posted by HostColor, 10-29-2016, 02:53 AM |
|
As @SkylakeDC and @DH-Owner suggested, the case here is reliability. As you mentioned that "we are taking extreme measures to make sure that there is minimal" the you must gofer RAID 10 setup, no question about that. If you want to save one SSD the follow @Mike_A advise and use RAID1 + hot spare SSD.
|
| Posted by alldedicatedserver, 10-29-2016, 03:55 PM |
|
yes. since you are using 4 disk, so go with raid 5 or raid 10.. if you only have 2 disk, raid 1.
|
| Posted by user54321, 10-29-2016, 05:02 PM |
|
You can also look at filesystems like btrfs or zfs that use checksums to detect data corruption over time. They have also RAID built in and with ECC RAM you are on a good way.
|
| Posted by Srv24x7, 10-30-2016, 01:30 AM |
|
Hi,
Before making any decision, you consider the pros and cons from all the prospectives.
1) RAID-1 or RAID-10 would be great for data redundancy. With this, you will definitely get a good read performance rate, however, you may loose the writing performance a bit, as in order to make data redundant, it will write to all disk in the backend.
2) Cost: If it is small database, using 1 TB would be waste of resource. I am not sure which datacenter you are on, but they should provide you the small disk too. Check if they can provide it on-demand or is it something that you can ship it to them...
3) If data safety is at the highest priority, you can also do a incremental backup at regular period..
|
| Posted by MUmair, 10-31-2016, 01:40 AM |
|
Slightly off-topic,
Would you prefer 4x256GB SSDs in RAID10 or 2x512GB SSD in RAID1 ??
(ie you getting 512GB RAID protected space and 1 drive failure protection in both)
|
| Posted by Coolraul, 10-31-2016, 01:55 AM |
|
Raid-1 with a hot spare, 1 other drive for local backups and another OFFSITE backup.
Writing bad data to 4 disks is as bad as to one only you just wasted money.
|
Add to Favourites 
Print this Article
Also Read
Shared Hosting
Shared Hosting
