Solved

SAS vs. SATA

Posted on 2010-11-23
17
749 Views
Last Modified: 2012-05-10
I recently built a new server with the following:
Dual Xeon E5620 2.40
24GB of memory
Operating System partition on Raid 1 with 2 450GB SATA Enterprise drives
Storage partition Raid 1+0 four 2TB SATA Enterprise drives
Windows SBS 2008

Running the following
Exchange Server 2007
Accounting Software using Pervasive SQL
3 SQL server applications
General file sharing

The performance is decent by not great. I went with the SATA drives to save some money and to cut down on heat in the server room. Does anyone know how much of a difference 15K SAS drives would make? Am I taking a big performance hit for using RAID 1 for the operating system? It saved me having a huge headache once already, Any suggestions would be appreciated.
0
Comment
Question by:JFasnacht57
  • 3
  • 2
  • 2
  • +7
17 Comments
 
LVL 14

Expert Comment

by:athomsfere
ID: 34198406
You don't say what you are using the SBS server for. Some operations benefit greatly from 15k, other little.
0
 
LVL 47

Accepted Solution

by:
dlethe earned 100 total points
ID: 34198498
Well, best place to see some differences is to go to some place like http://www.seagate.com and look up the specs.

But disk specs are only a small part of the deal.  Interoperability with RAID controllers can add a whole new layer of complexity.

For example, with SAS disks, it is likely that if you have an appropriate controller than each disk can process read requests and there will be load balancing for reads.     The I/O queue depth and I/O queue algorithms are much more intelligent, and it is also possible that I/O queuing is even disabled on your RAID controller.

Real-world, in IOPs, you might see a 4X improvement over what you have now, with greatest improvement being in random.  In throughput, you could easily see 2X improvement.

Data integrity, and relative "snappiness" could also be noticeable, as the error recovery algorithms and timing, and # of ECC bits is probably an order of magnitude better then what you have now ... so fewer recoverable errors and XOR parity remapping.

Bottom line, you'll notice the difference if you have any sort of regular queue depth.
0
 
LVL 6

Assisted Solution

by:djcapone
djcapone earned 100 total points
ID: 34198508
Mail servers tend to see the biggest gains from faster spindle speed hard drives.  This is because much of the I/O for mail servers is small amounts of data being read and written.  Think about how small in filesize the average e-mail message is.  As such, they greatly benefit from reduced seek times that accompany higher spindle speed hdds (15k).
0
 
LVL 1

Assisted Solution

by:nilayaksa
nilayaksa earned 100 total points
ID: 34198716
The price difference will be significant. But in the other hand if you are running that kind of application it would benefit you to use 15K SAS drive and using cached RAID controller. Because SATA is very low on iops compared to SAS + Cached RAID Controller. And 4 disks inRAID 5 can be better if you compared to 4 disk in RAID 10.
0
 
LVL 14

Assisted Solution

by:RickEpnet
RickEpnet earned 100 total points
ID: 34199244
In most SAS installs with mutiplie disk arrays the more drives you put in (the more spindls) the faster your access. So for example everthign esle being equal if you wanted 2TB and you put in 8 300GB drives this woudl be faster that lets say 4 750GB drives.
0
 
LVL 55

Assisted Solution

by:andyalder
andyalder earned 100 total points
ID: 34199516
Depends on the exchange version you're using, 2007 and 2010 require far fewer I/Os than 2000 or 2003 since they save up little I/Os to make one big one. Seeing a lot of big Exchange server configs using 1 and even 2TB disks nowadays.

There again you've got 3 SQL databases and that still needs fast random access although I presume the load is pretty light.

Best thing to do is run perfmon and look at the disk queue length, a length of two I/Os per disk is optimal, so for example with an array of 10 disks you'd like to see a queue of 20 for that array.
0
 

Author Comment

by:JFasnacht57
ID: 34202128
I am running Exchange Server 2007. The server is a application and file server.


 I work in construction management and have a background in IT for small companies. I was using a Dell 2900 server, 32 bit sbs 2003 with 4GB and 15k SCSI drives, which was about 6 years old.  I was expecting a major performance improvement with the new server. It seems to be a little faster, There are only 6 work stations in our network, The load is what I would consider light. Maybe my expecations are a little high, I know that I have better hardware than most local construction companies our size. I plan on putting our old server back online as a second Win 2008 server, The processor is 64 bit and I plan on adding more memory. I am hoping that I lessening the load on the new server will help. I would like to find a way to get the old server to run cooler. I have 8 SCSI drives in it and I think it could hear half of our building. We have a small dedicated server room with AC. The AC couldn't keep up with the old server and the new server runs amazing cool.

Are there any other suggestions on how to increase performance.
Any other possible bottlenecks I should be looking at?
Would our server come to a screeching halt if I had 50 work stations with a heavy load?
Would newer SCSI drives help the old server run cooler?
0
Windows Server 2016: All you need to know

Learn about Hyper-V features that increase functionality and usability of Microsoft Windows Server 2016. Also, throughout this eBook, you’ll find some basic PowerShell examples that will help you leverage the scripts in your environments!

 
LVL 57

Expert Comment

by:giltjr
ID: 34202167
You need to first define your current "performance?"

What are you doing?  

What was the response time with the old server and what is the response time with the new server?

Define "heavy load"?

Is the majority of work CPU heavy, disk I/O heavy, or  network data transfer heavy?  If network data transfers heavy, did you upgrade the network?
0
 
LVL 55

Expert Comment

by:andyalder
ID: 34204934
You had 8 * 10K or 15K disks and you've replaced them with 4 * 7.2K disks which are about 1/3 of the speed. That's probably made the disk subsystem the bottleneck, run perfmon and look at disk queue length, with 4 disks an average queue of 8 would be a good figure.
0
 

Author Comment

by:JFasnacht57
ID: 34205458
Drive C: Raid 1 two 450 GB drives
Disk Queue average .6 max 1.3

Drive D: Raid 1+0 four 2TB drives
Disk Queue average .2 max 1.7

0
 
LVL 4

Expert Comment

by:fr0nk
ID: 34219959
Your disk queue is fine.

Please consider the following:

You can easily identify the bottleneck, if there's any.
Monitor this 3 values from the "outside" (means: in the performance chart of vmware server).

    * CPU Ready
    * CPU Running
    * CPU Waiting

They will tell you the following:
The usual lifespan of a process inside your Windows guest OS is:
1. a new process is created
2. The process asks the scheduler of your guest OS (Win) to get CPU cycles. This changes the process to the "Ready" state.
3. The scheduler of your guest OS will decide that the process now can have CPU cycles. This changes the process to the "Running" state.
4. In almost every case the process decides for itself that no CPU cycles are required anymore and gives control back to the scheduler. If not, the scheduler will take it away eventually.
4.1 The reasony why the process decides that it doesn't need the CPU anymore are: it is waiting for something: user input, data from disk, data from the network, etc.

Now you can interpret this values:
High Ready = CPU contention
High Waiting = I/O contention of any kind.

Please be aware of the fact, that idle time is also counted as waiting time, so evaluate the (disk, net, etc.) I/O queues inside the guest OS if you're getting high waiting times.

Hope this helps.
0
 
LVL 4

Expert Comment

by:fr0nk
ID: 34219961
If you don't have a VMware Server or ESX or something like that, monitor this values from within perfmon.
0
 
LVL 14

Expert Comment

by:RickEpnet
ID: 34220078
I think you maybe looking in the wrong place. Your server may not be the issue. What do you have for networking. Switches, Network Cards in the desktops etc. Are the desktops old? Your server could be lighting fast but if everything else is slow it will not matter.
0
 

Author Comment

by:JFasnacht57
ID: 34302862
You say to monitor the following:
 
    * CPU Ready
    * CPU Running
    * CPU Waiting

Could you be more specific. Which counters should I add in Perfmon?
I do not have a VMware Server or ESX

Thanks...
0
 
LVL 33

Expert Comment

by:digitap
ID: 34505902
This question has been classified as abandoned and is being closed as part of the Cleanup Program.  See my comment at the end of the question for more details.
0

Featured Post

Top 6 Sources for Identifying Threat Actor TTPs

Understanding your enemy is essential. These six sources will help you identify the most popular threat actor tactics, techniques, and procedures (TTPs).

Join & Write a Comment

More or less everybody in the IT market understands the basics of Networking, however when we start talking about Storage Networks, things get a bit dizzier, and this is where I would like to help.
#Citrix #Citrix Netscaler #HTTP Compression #Load Balance
Sending a Secure fax is easy with eFax Corporate (http://www.enterprise.efax.com). First, Just open a new email message.  In the To field, type your recipient's fax number @efaxsend.com. You can even send a secure international fax — just include t…
Get a first impression of how PRTG looks and learn how it works.   This video is a short introduction to PRTG, as an initial overview or as a quick start for new PRTG users.

747 members asked questions and received personalized solutions in the past 7 days.

Join the community of 500,000 technology professionals and ask your questions.

Join & Ask a Question

Need Help in Real-Time?

Connect with top rated Experts

11 Experts available now in Live!

Get 1:1 Help Now