8Disk RAID10 or (2) 4Disk RAID10

[rockettpc]

Hey guys, I am looking to config a RAID for an SQL DB here at work. Currently we have the DB installed all on the C:/

Yes I know thats not where it goes. HAHA i didnt put it there, the DEV did. Blows my mind. (this is a production server, the DB should be on one set of drives, and the LOG should be on another)

I was wondering, should I RAID10 (8) WD1002FAEX drives and partition it as two logicaly drives and place the DB on D:/ and the LOG on E:/

or

should I create (2) seprate RAID10's with (4) WD1002FAEX drives each and place the DB on on array and the LOG on the other array?

 

[choucove]

Either way should work fine for what you are needing to do. However, the key difference here is that a larger set of RAID 10 drives is going to give you greater overall throughput, even if you have that array split into multiple partitions. For example, a set of four of those WD Black 1 TB drives in RAID 10 should give you 250 MBps and possibly upwards of 280 MBps. However, a set of eight would give you nearly double that throughput. I would be cautious about using that many WD Black drives in such a large RAID volume. That number of drives in a production environment really should be utilizing at least the WD RE4 drives.

 

[rockettpc]

Really? I just thought that the RE4's were more for NAS and NAS is no good for SQL DB's. The RE4's claim to be designed for NAS with NASware on them. I was trying to avoid them for that reason. The RE4's are about the same price as the blacks and I just thought the blacks had better performance for storage environments.

Either way should work fine for what you are needing to do. However, the key difference here is that a larger set of RAID 10 drives is going to give you greater overall throughput, even if you have that array split into multiple partitions. For example, a set of four of those WD Black 1 TB drives in RAID 10 should give you 250 MBps and possibly upwards of 280 MBps. However, a set of eight would give you nearly double that throughput. I would be cautious about using that many WD Black drives in such a large RAID volume. That number of drives in a production environment really should be utilizing at least the WD RE4 drives.

 

[FireWire2]

There are some factors that you need to consider beside the rule of thumb more spindles = more speed

1_ Controller bandwidth: 4x lanes 8x lanes of PCIe Gen1 or PCIe Gen2
2_ Hardware RAID, Hardware Assist, or software RAID
3_ Does your controller support NCQ?

If
1 = 8x lanes Gen1, or 4x Lane Gen2,
2 = Hardware RAID
3 = Yes

Then you can do what ever you want you still very fast
BTW I use 1000's of desktop HDD for RAID5/6/10 there is no issue, just have extra invnetory, because over 2TB the reliability goes down.

But if your DB is less than 512GB, I would go for Mirror 2x 512GB SSD...
There is nothing in 3.5 can compare to SSD in random read/write.. which is exactly SQL DB is all about.

 

[rockettpc]

So here are my Server's Mobo Specs, we plan on using the Adaptec RAID 6805. Our DB today is 4GB its about 4 months old. Our biggest problem right now is printing on the client side. Each Client machine has a Intel Pentium G630 2.70GHz with 4GB of DDR3. When we print reports it takes forever! Again currently all DB Files and LOGS and TEMPDB are on a single HDD on our servers C:/

I am hoping to faster query times using a RAID10, wether it be with SSD's or HDD's. I ran SP_Blitz on my Master DB. There are some problems with the DB that we are working on. Yet i think one of the big problems is I/O.

There are some factors that you need to consider beside the rule of thumb more spindles = more speed

1_ Controller bandwidth: 4x lanes 8x lanes of PCIe Gen1 or PCIe Gen2
2_ Hardware RAID, Hardware Assist, or software RAID
3_ Does your controller support NCQ?

If
1 = 8x lanes Gen1, or 4x Lane Gen2,
2 = Hardware RAID
3 = Yes

Then you can do what ever you want you still very fast
BTW I use 1000's of desktop HDD for RAID5/6/10 there is no issue, just have extra invnetory, because over 2TB the reliability goes down.

But if your DB is less than 512GB, I would go for Mirror 2x 512GB SSD...
There is nothing in 3.5 can compare to SSD in random read/write.. which is exactly SQL DB is all about.

 

[FireWire2]

I think your issue is not in DB but rather the network protocols, i can be wrong 🙂

Perform this test, it will tell you ASAP.

Get a SSD, move DB to SSD , access from there... Just make sure the SSD run as AHCI - get a additional card if need to.

see whether the issue till there or not, you can always re-use the SSD for OS other system

 

[rockettpc]

So I placed an order for some parts. We have decided to upgrade our current Server.

New Parts:
(8) Western Digital WD RE4 1TB 7200RPM 64MB Cache SATA 3.0GB/s
(1) Corsair HX Series 850W ATX PSU
(1) Supermicro SNK
(2) Samsung 16GB 240-Pin DDR3 SDRAM 1333 ECC
(2) Supermicro 1.6ft SATA / SAS Cable
(1) Intel Xeon E5-2640 Sandy Bridge-EP 2.5GHz (3GHz Turbo Boost)
(1) Supermicro MBD-X9DR3-F-O
(1) Rosewell Blackhawk-Ultra

Got everthing for $3200.00 off Newegg.com

Going to install Windows Small Business Server 2008 R2, and MS SQL Server 2012 Enterprise Edition. I have 2 WD500AAKX-001CA0 that i will use in RAID1 for the OS and the other 8 will be RAID10 for the DB files etc.

 

[popatim]

Two 4 drive arrays is the way to go. otherwise your singel array will do some DB work, stop & move the heads, write a log, move the heads, DB, move the heads, Log... over and over.

With two arrays, data on one, Log on the other, both can be accessed simultaneously.

 

[rockettpc]

I was under the impression that an 8 disk RAID10 partitioned would yield better results. Each disk has 2 platters and 4 heads. So combined there would be 32 heads and 16 platters working simultaneously. If the RAID is partitioned then the LOG and DB should still be able to write simultaneously... Am I wrong?

Two 4 drive arrays is the way to go. otherwise your singel array will do some DB work, stop & move the heads, write a log, move the heads, DB, move the heads, Log... over and over.

With two arrays, data on one, Log on the other, both can be accessed simultaneously.

 

[popatim]

you are forgetting that all disks in the raid act as one so if it has to write to another part, all the drives in the array have to reposition their heads. Raids increase thoroughput but increase access time also which is why it typically slows down things that access lots of small files, like windows startup.