Examining the Dell PERC H310 Controller
-
@scottalanmiller said:
Defeats the point of SSDs quite a bit, though, and increases wear and tear on them dramatically.
Why is that?
-
@BRRABill said:
@scottalanmiller said:
Defeats the point of SSDs quite a bit, though, and increases wear and tear on them dramatically.
Why is that?
Because the RAID cache is a major component of speed by moving things into memory. And the wear and tear is because with SSDs you set the cache to be primarily for writes and many of the writes, especially when you have RAID 5 which suffers from 400% write expansion, are absorbed by the RAID controller. If a single block is changed 20 times, the controller might absorb all of those writes and keep them from going to the disks at all. And it can queue things for efficient writing. Very important with SSDs and parity arrays.
-
Wouldn't it have the same issue with "spinning rust" as you guys call it?
-
@BRRABill said:
Wouldn't it have the same issue with "spinning rust" as you guys call it?
Except there is no appreciable wear and tear from writes with spinning rust.
-
@scottalanmiller said:
Except there is no appreciable wear and tear from writes with spinning rust.
Is it proven (questioning the theory, not you) that is really a concern with SSDs? Especially server grade SSDs?
-
@BRRABill said:
Is it proven (questioning the theory, not you) that is really a concern with SSDs? Especially server grade SSDs?
That writes wear them out? Yes, it is very well established that writes are the only significant reliability concern to SSDs. Shock, temperature, operating duration, read frequency all have effectively zero effect on them. Writes alone cause them measurable wear.
-
The risk is far lower than people like to make it out to be and enterprise drives are much better than non-enterprise drives, but normally drives do not take direct writes in any serious server situation. Having enterprise drives without a cache in front of them is an odd pairing and not something that we would ever expect to see in an enterprise scenario. RAID array cache is one of the most significant features looked for in servers. 1GB of cache is normally a minimum today.
Add to that parity write expansion and you might have a lot more writes than is normally expected.
-
So it sounds like my options are...
(I only have the H310 in hand)
- keep the H310, and get 10K SAS DELL drives
- buy a H710, throw the H310 away (or maybe eBay), and go with the EDGE SSDs
WWSD?
-
To say that those are the only options is a bit extreme. But it is true that spending money on fast drives with an H310 controller doesn't make sense. H310 is cheap for people cutting all possible corners, even 10K SAS drives does not match. With an H310 we would expect NL-SAS (aka 7200 RPM SAS) at maximum and SATA drives more likely.
-
I contacted xByte to see if they would have any interest in swapping out my H310 for a H710.
If not, it's a sub $200 loss.
-
Mint is up on the DELL T320 with PERC H310.
Let me know what you'd like to see.
-
Let's start with ...
sudo fdisk -l -
Disk /dev/sda: 499.6 GB, 499558383616 bytes
255 heads, 63 sectors/track, 60734 cylinders, total 975699968 sectors
Units = sectors of 1 * 512 = 512 bytes
Sector size (logical/physical): 512 bytes / 512 bytes
I/O size (minimum/optimal): 512 bytes / 512 bytes
Disk identifier: 0x23492907Device Boot Start End Blocks Id System
/dev/sda1 * 2048 718847 358400 7 HPFS/NTFS/exFAT
/dev/sda2 718848 245759999 122520576 7 HPFS/NTFS/exFAT -
There are 2 500GB SATA drives hooked into this H310 in a RAID1 array.
Server 2012 is currently installed on them.
But I can blow it away if needed.
-
@BRRABill said:
There are 2 500GB SATA drives hooked into this H310 in a RAID1 array.
Server 2012 is currently installed on them.
But I can blow it away if needed.
They are in RAID 1?
-
-
That the config screen during POST.
-
Cool, thanks.
-
Let me know what you want next, Guv'na!
-
What? No more testing? Are we that MEH about this controller?
