Storage Spaces Performance in Windows Server 2012 Essentials on a HP N40L ProLiant MicroServer

• December 22, 2012

 

clip_image002

After finishing up my test build for Windows Server 2012 Essentials (“S2012E”) in a HP N40L ProLiant MicroServer (“MicroServer”) I wanted to continue some of the testing I had started on Storage Spaces in Installing Windows Server 2012 Essentials on a HP N40L ProLiant MicroServer.

 

Contents:

HP N40L ProLiant MicroServer

Setting up a Virtual Drive in Storage Spaces

Overview of the Drives to be Tested and Test Summaries

· Table 1: Tested S2012E Virtual Drives in MicroServer

Detailed Test Results

Conclusion

Reference

= = = = = =

 

HP N40L ProLiant MicroServer

The MicroServer used in this series of tests is the same MicroServer described in the Installation of Windows Server 2012 Essentials on a HP N40L ProLiant MicroServer.

I am currently operating this MicroServer (with S2012E installed) with three client machines (all VM’s) with about 3TB of data in the ServerFolders.

1. One Windows 8 Pro Machine (VM) is joined to the S2012E domain. The connector is installed and the machine is being successfully backed up regularly. I haven’t implemented Group Policy on it yet.

2. One Windows 8 Pro Machine (VM) was joined then un-joined to the S2012E domain using the process documented by Tinkererguy here, here and explained in the video here. The connector is installed and the machine is being successfully backed up regularly. The Machine is joined to Homegroup.

3. One Windows 8 Pro Machine (VM) was not joined to the S2012E domain by the process documented by Paul Thurrott, TechNET, and here. The connector is installed and the machine is being successfully backed up regularly. The Machine is joined to Homegroup.

clip_image004

Since early November I’ve only had one storage spaces alert and that was the one, noted later in this paper, regarding the diminishing space available in my storage space pool after I had added a fairly significant number of virtual drives to the storage space for the performance testing in this paper.

 

Setting up a Virtual Drive in Storage Spaces

When I first installed S2012E on the MicroServer I had used the setting in Control Panel => System and Security => Storage Spaces => Manage Storage Spaces to build my Storage Pool and Virtual Drive that became my data Drive D: for my server share folders.

clip_image006

When I started working in Storage Spaces again for the series of tests in this paper I found it was much easier (in my opinion) to work in the S2012E Server Manager. Starting at Server Manager => File and Storage Services => Volumes => Storage Pools

clip_image008

To create a new Virtual Disk: at the lower left box listing “VIRTUAL DISKS” (Storage pool on Ranger) just click the “TASK” button at the right top of that box and click the drop down from TASK that says “New Virtual Disk” and follow the wizard:

clip_image010 clip_image012 clip_image014 clip_image016 clip_image018 clip_image020 clip_image022 clip_image024 clip_image026 clip_image028 clip_image030 clip_image032 clip_image034 clip_image036 clip_image038

And that’s it. We’ve just created a new 1TB Simple Storage Space NTFS Drive – Drive E: — that we can also see:

clip_image040

Using the same steps as above I began creating the additional drives in Storage Spaces for testing. With five (or more) physical hard drives making of the Drive Pool the Wizard introduces an additional option with selecting a Mirror Storage Layout:

clip_image042 clip_image044

With five (or more) physical hard drives in the Storage Pool you’ll be given the option of a 2-way or a 3-way mirror.

A view of virtual Disk Drives D: through N: and another view of 13 disks in Ranger in Server Manager just before testing began.

clip_image046 clip_image048

 

 

 

Overview of the Drives to be Tested and Test Summaries

Hard Drives to be Tested in test platform: Windows Server 2012 Essentials (“S2012E”) operating on HP ProLiant N40L MicroServer (“MicroServer”) named RANGER

All physical drives in the MicroServer were attached to SATA II ports.

     Physical Drive(s)

 

Drive C: OS Drive, Crucial M4 128GB SATA III SSD

· ATTO Test Max Read: 281

· ATTO Test Max Write: 187

 

     Virtual Drive(s) in Storage Space are carved out of a Storage Pool made up of five physical ST3000DM001’s attached to SATA II ports in the MicroServer. Storage pool capacity = 13.6TB

Table 1 shows the following Virtual Drives that were tested.

Drive D: is the current working data drive for S2012E on the MicroServer. Drives E: through N: were all created in Storage Space before testing was done on Drives C: through N: — afterwards Drives E: through N: were removed from the Storage Space before the creation and testing of Drives O: and P: because of warnings I received from S2012E that my Storage Space capacity was running low. After testing on Drives O: and P: were completed those drives were also removed from the Storage Space.

Table 1: Tested S2012E Virtual Drives in MicroServer

Virtual Drive No. HDD’s in Storage Space Provisioning Capacity Resiliency Type File System ATTOTest

Max

Read

ATTOTest

Max

Write

D 5 Thin 6.8TB 2-Way Mirror NTFS 381 208
E: 5 Thin 1TB Simple NTFS 700 647
F: 5 Thin 1TB Simple ReFS 645 606
G: 5 Thin 1TB 2-Way Mirror NTFS 456 308
H: 5 Thin 1TB 2-Way Mirror ReFS 1,504 50
I: 5 Thin 1TB 3-Way Mirror NTFS 450 173
J: 5 Thin 1TB 3-Way Mirror ReFS 1,589 50
K: 5 Thin 1TB Parity NTFS 463 37
L: 5 Thin 1TB Parity ReFS 467 46
M: 5 Fixed 1TB 2-Way Mirror NTFS 444 306
N: 5 Fixed 1TB 2-Way Mirror ReFS 1,570 50
O: 5 Fixed 1TB Parity NTFS 591 43
P: 5 Fixed 1TB Parity ReFS 479 45

Note: In the November 9, 2012, test on Drive D: published the ATTO Test Max Read was 342 and the ATTO Test Max Write was 191.

Detailed Test Results

Drives C: through P: tested in ATTO Disk Benchmark:

clip_image049 clip_image050 clip_image051 clip_image052 clip_image053 clip_image054 clip_image055 clip_image056 clip_image057 clip_image058 clip_image059 clip_image060 clip_image061 clip_image062

Conclusion

While I was expecting poorer write speeds with Parity drives (based on past testing) I was surprised by the relative poor write speeds for 2-way and 3-way Mirror drives with ReFS. I simply don’t understand how ReFS’s performance is so heavily impacted.

For good overall performance with some resiliency the 2-Way & 3-Way Mirrors with NTFS would seem to be good choices to consider and offered protection against one or more failures of the physical drives making up the underlying drive pool.

A Storage Spaces Virtual Drive with 2-Way Mirror resiliency, NTFS and Thin provisioning is what I plan to continue using for the foreseeable future in S2012E.

The Simple Resiliency drives in either NTFS or ReFS offered very good performance and may be useful in gaming as well as other temporary applications that need high disk IO but they offer no resiliency if any of the physical drives making up the underlying drive pool should fail.

At TechNet’s Wiki on Storage Spaces the best uses of simple, mirror, and parity spaces are listed as:

= = =

Storage Spaces offers simple spaces, mirror spaces, and parity spaces, each with unique characteristics.

· Simple spaces are designed for increased performance without increased resiliency. They are best suited for temporary data, such as video rendering files, image editor scratch files, and intermediary compiler object files.

· Mirror spaces are designed for increased performance and increased resiliency. Two-way mirror spaces can tolerate one disk failure and three-way mirror spaces can tolerate two disk failures. They are well suited to storing a broad range of data, from a general-purpose file share to a VHD library. When a mirror space is formatted with the Resilient File System (ReFS), Windows offers automatic data integrity maintenance. This is a layer of resiliency is above and beyond the resiliency achieved from maintaining multiple data copies to tolerate drive failure.

· Parity spaces are designed for capacity efficiency and increased resiliency. Parity spaces are best suited for archival data and streaming media, such as music and videos.

= = =

Interesting, drives D: and G:’s performance (see Table 1) are different by such a wide margin even though they are the same except for the capacity (6.8TB vs. 1TB) and the amount of free space (~56% vs. ~99%) on each drive. This may (speculation on my part) mean that as the virtual drives become bigger and/or are more filled their performance may be depressed. If true, this wouldn’t bode well for the long term Write performance for any Parity drives or any Mirrored ReFS drives. More testing is necessary in this area. As the Note to Table 1 shows, D:’s ATTO performance improved slightly since the test early in November.

As can been seen below drive D: has a substantial number of Share that in the aggregate contain about 3TB of data as compared to virtually no data on drive G:

clip_image064

Reference

Installing Windows Server 2012 Essentials on a HP N40L ProLiant MicroServer

More Interesting MicroServer Links

HP MicroServer

Storage Spaces Overview

Storage Spaces Frequently Asked Questions (FAQ)

Windows Server 2012 Essentials

Storage: Windows Server 2012

Storage Spaces explained: a great feature, when it works

Check out other Storage Spaces Stories

Share

Tags: , , , , , , , , , , , , , , , , , ,

Category: Hard Drives, Home Server Storage, HP Microserver, Review, SSD, Windows Home Server

Comments (5)

Trackback URL | Comments RSS Feed

  1. Hardened says:

    Nice results. It would be interesting to compare to old school SW raid solution if it is still available for W2012e through drive management snap-in.

  2. Scott Krueger says:

    Joe, I pretty much built a system just like described. I flashed the BIOS via Diehard and installed windows 8 on an SSD drive on port 5, and have a 3.0 TB drive in port 4 in the icy dock. Both show up in the BIOS. I changed the SouthBridge settings as per described and for some reason the 4 3.0 TB Seagate drives in the internal drive bays, don't show up in the BIOS or under manage, drives in Windows. Any thoughts on what I might be missing? Thanks!

  3. Martin says:

    Testing Storage Spaces myself, the write speed seemed to change a lot over time (I was using Refs, on a 2 way mirror). I copied about 200GB of test data from another NTFS drive. For 10 minutes or so it copied at a steady 9-10MB/sec – a bit slow I thought. I came back half an hour later and it was running at a steady 25Mb/sec.

×

Shop and help out Home Server Show. Drag this box to your BookMarks Bar. Amazon