Skip to content

Storage concepts for IT Pros-RAID types and IOPS

05/13/2013

In a series of lectures that I taught last year (TechEd 2011, SQLPass # 127, MCT Summit and universities), I discussed the topic on choosing the best storage medium and the RAID types available, with the advantages and disadvantages of each.

It is still a very new for IT Pros on account of not being discussed in college courses, technical training certification only.

In this first article I will address the RAID types and what are IOPS. In a next article comment about architectures and controller types (Fibre Channel, HBA, iSCSI and SMB).

Part 1 – RAID types

RAID is the acronym for Redundant Array of Independent Disk or "redundant Group of independent disks", which indicates the use of a number of disks to create a high availability.

The RAID types can be simplified by the use of 4 main types:

RAID Concepts and Nº Disks Advantages Disadvantages
0 – Stripped without parity

From 2 disks, information is written to both disks of independent form, that is, half of a file on each disk

High-performance
Low cost
Total space available

Without redundancy, any disc that lose the others have no way to recreate the lost data, once the information is in all disks

1 – Mirror

From 2 discs, always in pairs. The data is written to both disks in full

High performance, in reading uses the two discs
High redundancy

Only half of the sum of the disks is available
High cost due to the "lost" space

5 – Stripped with parity

From 3 discs. The information is written similar to RAID 0, but it uses an algorithm that information is written to a disk for reconstruction of any disk with error

Good performance
Good redundancy
Little loses useful space

Always lost the equivalent of a disk
In the event of a loss of more than one disk there is no way to rebuild

6 – Stripped with parity

From 3 discs. The information is written similar to RAID 5, but with 2 parity disks

Good performance
Good redundancy
Minor loss of space that RAID 1

Always lost the equivalent of two disk
In the event of a loss of more than two disk there is no way to rebuild

10 – RAID 0 Mirror

From 4 disks, where each two discs form a RAID 0, the second copy of the first. Is a mix of RAID 0 with RAID 1, but at the level of the whole and not the disc

High-performance
High redundancy

Loss of half of the disks available
High cost due to the loss of useful space

50 and 60 – Parity Mirror

RAID 5 and 6 with the parity disks mirrored

Great redundancy
Good redundancy

Another loss besides the disk that were already parity
Average Performance

The most widely used today is the RAID 5/10, since they have good performance and redundancy, as shown in the chart below:

image

Below a graph of General items and comparison between the types of RAID 5/10/6/50:

image

Note: the above data were collected in the id "Choosing a Member RAID Policy" which is based on Dell Equallogic architecture and it is not necessary to register: http://www.dellstorage.com/WorkArea/DownloadAsset.aspx?id=1066

Part 2-what is IOPS?

Is the number of operations per second that a single disk can get. For example, a 10 k SAS disk can on average 140 IOPS.

This speed is standard in the industry with variations between models, but we can have a basis of what is acceptable and the disc manufacturer can tell you this number.

However, note that the difference is too large, especially taking into account the new SSD disks. For example, the X 25-E Intel (See the pdf with the characteristics in http://download.intel.com/design/flash/nand/extreme/extreme-sata-ssd-datasheet.pdf) reaches the numbers 30 times larger than the SAS and SATA disks.

image

Because the IOPS is so important?

This question is obvious, but the explanation may not be so simple. It turns out that in most cases have the tendency to minimize the issue saying it’s "performance" or "user perception" but actually can impact directly on the running of an application, in some cases up to the end.

For example, an Exchange 2003 environment with 2 thousand mailboxes need 1.5 thousand IOPS and this number is not easy to achieve. SQL Server to a SharePoint database need 5 thousand IOPS for work.

How to calculate the IOPS?

Multiply the total by RAID type and disks will your number. Follow some examples:

image

RAID 1, RAID 10 or RAID 0 will you the largest proportional number of IOPS possible RAID 5, the calculation takes into account the least disk 1 and disk 2 to 50 less RAID parities.

How to get the largest possible IOPS with higher capacity?

We have three ways to do this:

  1. Use high-performance disks, as the 15 k SAS or SSD, but are expensive and only sizes SSD 32/50/64/100 GB
  2. Use the appropriate RAID type for performance and not to the size you want as many today do, which often involves using RAID 10 for total performance rather than RAID 50, we would lose in ability but gain in performance
  3. Buy one that works with virtual LUNs, it allocates the data on the disks as the need for this given and does not need to say the RAID type

Interesting references

How to calculate IOPS for Exchange 2003 http://technet.microsoft.com/en-us/library/bb125019 (EXCHG. 65) .aspx

How to calculate IOPS for Exchange 2010 http://technet.microsoft.com/en-us/library/ee832791.aspx

How to calculate IOPS for the SharePoint 2010 SQL http://technet.microsoft.com/en-us/library/cc298801.aspx

Utility to measure IOPS for SQL Server (SQLIO) http://www.microsoft.com/download/en/details.aspx?displaylang=br&id=20163

Advertisements

From → Hardware

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: