RAID 5 - 4 drives vs 5 drives.

Posted on 2010-08-17
Last Modified: 2016-12-08
I am currently rebuilding a RAID 5 array.

I would like to use one of the available drives as a hot spare, but not if there will be a performance impact.

Is there a performance difference in a RAID 5 array operating with 4 drives and a hot spare vs 5 drives without a hot spare?
Question by:exp_exch1
LVL 10

Accepted Solution

jramsier earned 100 total points
ID: 33454881
It is said that you will get around ~8% better performance with 4 7200RPM drives compared to 3.  This is basically because you are adding another hard drive to the mix and if 2 go out at the same time your screwed.
LVL 47

Assisted Solution

dlethe earned 100 total points
ID: 33455038
Above is not necessarily true.  Performance could very well decrease.   Performance is measured in I/Os per second as well as throughput.  IOPs and throughput are mutually exclusive. Furthermore, whatever block size the RAID is configured at  has a sweet spot, and performance characteristics change depending on reads or writes.   The native block size of the applications and file system are a huge factor.

So the rule of thumb is that there is NOT a real-world rule of thumb to assess overall performance impact of adding a drive.  Performance will be different. That is all you can get based on info you supplied.

It is NOT a good idea to use a hot spare, because the point of it is in a disaster recovery scenario.  The chances of getting a degraded array increase as you add members to the array, and the number of hours you are exposed (due to the rebuild time) also increases.

So no matter what, you put your data at a higher probability of being at risk (due to drive failure), and you increase the exposure as well.   If you value your data then don't do it.


Expert Comment

ID: 33455213
Yes, there will be a performance impact.  How much impact depends on several variables.  The only way to know for sure is to benchmark each configuration.  Even then, raw perf stats may not matter.  It also would be usefull to gather application performance statistics.

I would guess that the difference will not be too great.  Maybe a cold spare on the shelf is good enough if you want to use all drives?
PRTG Network Monitor: Intuitive Network Monitoring

Network Monitoring is essential to ensure that computer systems and network devices are running. Use PRTG to monitor LANs, servers, websites, applications and devices, bandwidth, virtual environments, remote systems, IoT, and many more. PRTG is easy to set up & use.


Expert Comment

ID: 33455417

This is my point of view only, but it may help you.
I never used a spare drive because for a server, a drive being quite expensive, I don't see why I shouldn't use it right away.
Using RAID5 gives me enough security as long as only 1 drive fails.
Anyway you spare drive could have only cover 1 failing drive.

For RAID5, the more drive you add, the more secure you are because data is plitted on more drives.
But then again, the more you split, comes a moment where performance declines.
It also depends on your RAID hardware, and the cache size.

Using 4+1 offers a different security in case 1 of the 4 fails... which can happen tomorrow or in 3 years depending on the age and usage of you disks.
But, will that spare drive still be operational after sitting there in the dust of the array for a long time ? And wil it keep up or also fail after a few days ?

If you have at least 128MB of cache on the array, I would advise you to use you 5 disks array for security and performance instead a 4+1.
The performance impact between the 2 configs will be seemlessly minimal.

For more than 5 disks, I use RAID1+0 (RAID10).

Keep it simple, you have 5 disks, use them all ... that's what I would do ;-)

LVL 47

Expert Comment

ID: 33455587
Actually, since you have a hot spare, then if you want best possible data integrity & reliability, then just incorporate the hot spare and resilver the RAID5 + hot spare into a RAID6 with no hot spare.  Then the added parity information is available 24x7.  In event of a drive failure, it is essentially already rebuilt for you .. instantly.

RAID6 does have slight performance disadvantage of RAID5 on writes, but this varies greatly depending on make/model.

However, if this was my data, then I would not hesitate to migrate from a RAID5 + spare to a RAID6.

Assisted Solution

drcheap earned 100 total points
ID: 33457015
It really depends on your ultimate goals and why you are using RAID.  You then need to weigh the advantages/disadvantages with respect to those goals.

4 disk RAID5 + 1 hot spare  vs. a 5 disk RAID5

Performance - You would in theory get slightly better performance by having all 5 disks active due to the fact that the stripes span more spindles.

Probability of failure - As the number of disks in an array increases, the chance of data loss due to multiple drive failures is increased. This is because there is a reduced ratio of "losable" drives (the number of drives which may fail before data is lost) to total drives.

Failure recovery time - The one thing the hot spare gives you is quick recovery.  A corrolary to Murphy's law states that if there is a worst possible time for something to go wrong, that is when it will go wrong.  As soon as your drive fails at 3am the night after you leave for your week long vacation, the array is rebuilt onto hot spare.  In the 5 disk version, it may be a week or two before you are able to plug in s fresh drive.  During that week or two, your performance suffers and your protection against failure is lost.

If your goal is performance, go with the 5 disk.  If your primary objective is availability, go with the 4+1 (or RAID6 if the controller even supports it).  And if you are somewhere in the middle and need performance, and can stretch the availability thing a bit (i.e. longer time to recover being acceptable), then the 5 disk may still be a viable option.
LVL 55

Expert Comment

ID: 34199567
@dlethe > IOPs and throughput are mutually exclusive

You keep making that statement and as far as I'm concerned it's total tosh. If I've got twice as many disks in one array than I have in another then assuming everything else to be equal I'd expect more IOPS and greater throughput from the array that has more disks in it.

Can you please either stop saying it or post something to justify it bearing in mind that the question is about the number of disks and not about the size of individual I/Os.
LVL 47

Expert Comment

ID: 34199791
You do not magically increase IOPs and do not increase throughput for all (controller stripe)  sizes and types of I/O for all host-generated I/O requests by adding disks.  Some I/O operations will result in less throughput and/or less total throughput.

In a perfect world, where there is no saturation, and all I/Os use different parity drives and those I/Os are balanced equally so queue depth is equal among all disks, and the physical I/O size is is exactly the same as the stripe size of the RAID set in the controller, and so on... then there will be such balance.

Real world, it doesn't work that way.  
LVL 55

Assisted Solution

andyalder earned 100 total points
ID: 34200764
In most cases you do increase both throughput and IOPS when you add more disks though don't you?

Simple yes/no will suffice, not a load of waffle. That's already excluded as everything else is assumed equal and the disk sub-system is assumed to be the bottleneck.
LVL 17

Assisted Solution

by:Gerald Connolly
Gerald Connolly earned 100 total points
ID: 34205135
I agree with Andy, in the real world, if you add another spindle to the stripe you increase the number of IOPS that the stripe can handle. Increasing the number of IOPS will increase the throughput for real world jobs.

A 4D+P RAID-5 RAIDset should perform better than a 3D+P RAID-5 plus a spare.

You still have the same number of spindles as before (5) so you shouldnt have any issues with the more drives the bigger the risk of failure.

And of course a 3D+2P RAID-6 would be better for availability than any of the other configurations but may give you slower write performance depending on the controller
LVL 88

Expert Comment

ID: 34924268
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.

Featured Post

Best Practices: Disaster Recovery Testing

Besides backup, any IT division should have a disaster recovery plan. You will find a few tips below relating to the development of such a plan and to what issues one should pay special attention in the course of backup planning.

Question has a verified solution.

If you are experiencing a similar issue, please ask a related question

Suggested Solutions

Title # Comments Views Activity
In Citrix PVS, where do new reads come from? 3 65
Calibrating x & y screen  pixel coordinates 1 42
Storage Options for VHD File in Disk2VHD 2 30
Best backup solution 8 42
Great sound, comfort and fit, excellent build quality, versatility, compatibility. These are just some of the many reasons for choosing a headset from Sennheiser.
In this article we will learn how to backup a VMware farm using Nakivo Backup & Replication. In this tutorial we will install the software on a Windows 2012 R2 Server.
This video teaches viewers how to encrypt an external drive that requires a password to read and edit the drive. All tasks are done in Disk Utility. Plug in the external drive you wish to encrypt: Make sure all previous data on the drive has been …
This Micro Tutorial will teach you how to reformat your flash drive. Sometimes your flash drive may have issues carrying files so this will completely restore it to manufacturing settings. Make sure to backup all files before reformatting. This w…

820 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