<div>
For large drives, RAID 5 is not any longer the best option. Read this article below for more information<br />
<br />
<a href="http://www.hdd-tool.com/raid/raid-5-features-advantage-and-disadvantage.htm" rel="ugc">http://www.hdd-tool.com/raid/raid-5-features-advantage-and-disadvantage.htm</a><br />
<br />
<br />
&nbsp;We now use RAID 10 exclusively on Server 2012 R2 large drive machines.<br />
<br />
<i>VERY expensive </i> &nbsp;&lt;-- Drives are not all that expensive and RAID 10 is a better option.
</div>


<div>
Thanks for the response. I will respec the systems with 10, But again in 10, if you loose two drives the party is over if in the same volume - correct? &nbsp;I would be concerned of performance issues with 6. &nbsp;These are all SAS. &nbsp;Rebuilds never ever took more than a few hours for us with RAID 5.
</div>


<div>
We use top quality drives and have not lost any at any client - drive quality is improving. <br />
<br />
RAID 10 is newer and more robust so that is what we use. RAID 5 works and you can employ that if you so desire. <br />
<br />
Here is an article comparing RAID 5 and RAID 6. <br />
<br />
<a href="http://www.techrepublic.com/blog/the-enterprise-cloud/raid-6-do-you-really-want-it/" rel="ugc">http://www.techrepublic.com/blog/the-enterprise-cloud/raid-6-do-you-really-want-it/</a><br />
<br />
I have scanned through both articles and I am happy with our selection for RAID 10. Our technical consultants recommended RAID 10 as well
</div>


<div>
With the current sizes of hdd raid 5 is no longer a good solution.<br />
Raid 6 is more safe and you loose about 40% of your total hdd space.<br />
<br />
Raid10 is much faster and more reliable but you loose 50% of all available storage space.<br />
<br />
Choices choices....
</div>


<div>
Yes and drives have gotten so inexpensive that the speed (and robustness) of RAID 10 is excellent.<br />
<br />
Even top quality SSD drives are a good investment amortized over 5 years.
</div>


<div>
For systems with 4 drives, you might as well use RAID 10, since that's faster than RAID 6 with the same capacity. &nbsp;Four 4TB disks will get you 8TB of storage, and those are cheap enough. &nbsp; The only reason you have a system with six 2 TB disks to get the 8 TB storage is to improve the speed, but RAID 6 will still slow it down and you'd use more electricity. &nbsp;Not much benefit in that.<br />
<br />
Anything 8 disks or less should probably just be RAID 10. &nbsp;RAID 6 is better for 12 or more disk systems. &nbsp;The reason I say that is that most NAS are 8 disks or less, and run slower in the first place, and the higher end SANS have more disks.<br />
<br />
RAID is for speed, uptime, and/or storage capacity. &nbsp;It is not backup, so not a guaranty against any data loss. &nbsp;You should always have your data backed up separately.
</div>


<div>
we are doing SAS drives which do not come any larger than 900GB in 15k drives. &nbsp;No SATA, these are for their file servers, will only do SAS. &nbsp;I realize it is not a means of backup, just trying to prevent outages. &nbsp;Still thinking RAID 5 is superior to 6 on performance, but appears more people are springing for RAID 10 these days
</div>


<div>
That is because RAID 10 is newer and more robust while RAID 5 is now fairly old technology.
</div>


<div>
Hey John<br />
<br />
It has been around just as long just cost prohibitive. &nbsp;Had one co using it about 12 years ago. &nbsp; I am starting to think I am putting too much time into this because even with RAID 10, If you have two drives to fail - on the same volume, the party is over anyway. So it really is not that much better then unless I am missing something
</div>


<div>
I used RAID 5 a decade ago with Server 2003 and earlier drives. I had one drive fail in a 3 drive set for two clients, replaced the drives and moved on. Drives were not as big and RAID 10 was not out then. <br />
<br />
Yes, RAID 5 and 10 - lose 2 drives and you are done. That is why you need top notch backups which we also use. But I have never lost 2 drives at once. <br />
<br />
You need to use what you wish, but with newer technology, cheaper and bigger drives, I stick with RAID 10.
</div>


<div>
Hey John<br />
<br />
Thanks so much for sharing with me. So if you are doing sas 900gb drives, and want 8TB RAW, you would suggest they pay for twice the storage for RAID 10? &nbsp;Or again could go with two RAID 5 volumes with a hotspare split the capacity to 4TB RAW for each RAID 5 vol, and should they lose a drive, could move the VMs over to the other volume until the Rebuild completes. &nbsp;The downside of that is without storage vmotion, will have to bring down the guest for the move. That combined with replicating the VMs to either the second volume or to hostb and the cloud should do it I am thinking.<br />
We can price it both ways. One large RAID 10 or two Raid 5s. &nbsp;Not seeing a lot of added safety with 10 if two drives, (of course on the same &nbsp;volume fails), means the party is over.
</div>


<div>
RAID 10 was used in the 90s, so it's not new. &nbsp;They called it RAID 1+0 back then. &nbsp;RAID 6 is &quot;new&quot; and came on the scene about a decade or so ago. &nbsp;RAID levels 0 - 5 were defined long ago, although I've never really seen anyone use RAID 2 or RAID 4 in production.<br />
<br />
Yes, if you have the money, or don't need that much storage capacity, then you should use RAID 10 over RAID 6 if you have 8 disks or fewer. &nbsp;It's all about risk mitigation and how quickly you can recover your data. &nbsp;I'm not only looking at failure, but also at the data recovery rate. &nbsp;You have to look at the whole picture. &nbsp;RAID 10 should be fast enough to recover from backups in a sufficient time period for most use cases of 8 disks or fewer. &nbsp;You're basically trading the chance of a disk failure in the same RAID 0 with the speed and the speed of data recovery in the event of a failure. &nbsp;Can your system have downtime for that long. &nbsp;If not, then spend more money and get more disks and do RAID 6 or RAID 15 (RAID 1+5, essentially mirrored RAID 5s).<br />
<br />
How much usable capacity do you need? &nbsp;At 8 disks, the 900GB drives are only 7 TB RAW. &nbsp;It's only 3.5 TB usable with RAID 10 and about 5 TB usable with RAID 6. &nbsp;That 3.5 TB is recoverable from backups in about a day or 2 if you have backups on site, or paid for a gigabit connection to your cloud storage. &nbsp;I would still keep backup data on site, even with a cloud storage provider.<br />
<br />
If you want 8 TB of usable space, then you should get more disks and go with RAID 6. &nbsp;For RAID 15, the 8 disks will give you only 2.7 TB of usable storage. &nbsp;If you want more capacity, add a few more disks and you can go with RAID 6. &nbsp;At 12 disks or more you really should start using RAID 6, unless you have the money and are willing to buy even more disks or take a hit on storage space and do RAID 15. &nbsp;Frequently, with that many disks, a 2nd disk will fail within 2 weeks of the 1st disk, sometimes before a rebuild completes. &nbsp;It's all about probability. &nbsp;That's the main reason not to use RAID 5 now.<br />
<br />
4-6 Windows server VMs shouldn't take too much space. &nbsp;You also should create a File based storage with a separate file based backup for your file server and not make a file server a full block based VM. &nbsp;Those become quite slow to recover after a failure and quite slow to recover individual files.<br />
<br />
RAID 5 is probably still fine for the eight 900 GB SAS drives that are 10k or 15k RPM, as long as you have backups and you're ok with them being a day or more older and you're also ok with a few days of downtown to restore from backup. &nbsp;Since they're VM's, you won't be actually filling all your available space or you will have problems with snapshots. &nbsp;The worry with RAID 5 is that the systems won't rebuild in time before the 2nd disk fails. &nbsp;That's more of a problem when you have larger capacities. &nbsp;I wouldn't do RAID 5 for larger storage capacities.
</div>


<div>
Serialband<br />
<br />
Excellent response. &nbsp;So tell me, they are shooting for 8-12 TB total space, to allow for future expandability. &nbsp;They will have two hosts, a second will be nothing more than a standby. Will be using Veeam which lets you replicate the VMs as well as file level backup/restores to both hostB and most likely secondly to the cloud.<br />
<br />
They may only be using less than half the space they are requesting initially. &nbsp;Looks like the 900GB 15k SAS drives are the way to go. I have some concerns about performance degradation with RAID 6, they are used to RAID5. &nbsp; RAID 10 to get 8TB would be 18 900s, and would you do a hotspare on RAID 10?<br />
<br />
Budget is always a factor, as you can see even for RAID 6 to get the 8TB raw you are looking at 11 spindles + 1 for a hotspare.<br />
<br />
<br />
So that being said 8-12TB &nbsp;RAW, &nbsp;What is your final answer ;-). &nbsp;This will be to house their VMs.<br />
<br />
Thanks so much!
</div>


<div>
Usually with VMs, you do separate RAID LUNs and split the VMs onto them. &nbsp;It depends on what you have for RAID. &nbsp;You should get a system that can expand.<br />
<br />
With the old RAID SANs, VMWare used to have you split the RAID into multiple RAID 5 groups of 4 or 5 disks, then you split your VMs onto them. &nbsp;The newer systems can aggregate the RAID groups into one larger logical disk and manage them &quot;invisibly&quot; to you. &nbsp;Some newer systems with SSD cache just use larger RAID groups. &nbsp;It all depends on what you currently have.<br />
<br />
Don't buy much more disk than you really need up front. &nbsp;They'll all be in the same purchase year and they'll all have similar failure rates due to higher probabilities that they're in the same batch and more likely to fail at the same time.
</div>


<div>
Should I comment while this 'friendship' is forming? I think I should.<br />
<br />
This question has reached the point where 900gb sas drives are being used, I totally agree with that decision.<br />
I totally disagree for thinking raid five would be acceptable, in the last years hdd manufacturers have been improving there performance but especially their URE. It has been way better then a decade ago but the hdd size has been advanced way harder so the chance off successfully rebuild a raid five array has dropped since.<br />
<br />
Then SAN solutions come to this table we're we recently bought 2 hp San with 14tb storage each.<br />
Hp strongly recommends using raid 10 for all 900gb sas drives and raid 1 for the ssd pool.<br />
Off course we bought all drives in one order to have good and simple administration for warranty and to be 100% sure the storage works perfectly together.<br />
Because of budget we decided to use raid 6 instead simply to use more storage per dollar where raid 5 was denied anonymously by all levels in our it structure.<br />
<br />
That being said, the final decision is for the poster. Hopefully all works well but I am sure you will remember this post once disaster strikes you.<br />
<br />
Just my cup off tea.<br />
<br />
Cheers
</div>


<div>
Hey Patrick<br />
<br />
Thanks much for your input. &nbsp;Yes - ah ha, I think it took a bit to sink in (I was slower than usual today), so the URE has increased along with the drive size introducing more risk in terms of the likelihood of a successful rebuild. I got it. &nbsp;<br />
<br />
Although I can get them to spring for a second server (host), and you may have seen me mention this before, we were going to use Veeam to replicate the VMs to hostb, in the event of a catastrophic failure of hosta. for such a small office with less than 20 people currently - (doing 3d rendering etc), It is difficult to justify a SAN which could add another 26-30k ish for the SAN, fibre channel switches etc. So it looks like we would do local storage. That being said. I would think they are far better off with RAID 6 (but would be concerned about performance), or RAID 10. &nbsp;They are used to running on a server with RAID5 with 15k drives. &nbsp;Do you think they would feel like they would hit a wall with RAID6?<br />
<br />
I am going to price it out as two options Raid 6 and 10. May I ask what HP solution you ended up going with? I was looking at a dell SCV2020<br />
<br />
Thanks again so much for sharing your knowledge with me as well!
</div>


<div>
serialband<br />
<br />
This would be a wee bit different however, because I can't see them splurging on a SAN! &nbsp;I really want to do shared storage for them, but I don't see how I could get the performance needed to run Windows VMs off of a decent quality SAN, with almost local disk performance for less than 25-30k, ish, and it is really overkill for an office of less than 20 people at this time. &nbsp;So it is a difficult sell for the SAN, I am shocked they would spring for a second host as backup. &nbsp;Perhaps I need to push harder ;-)<br />
I have done SANS in larger environments, but not as small as this.
</div>


<div>
<i>May I ask what HP solution you ended up going with? I was looking at a dell SCV2020</i><br />
<br />
We are using Lenovo Server 2012 R2 at all clients
</div>


<div>
HP wants you to spend more money ;)<br />
<br />
I agree that RAID 5 is not a good choice for those larger storage arrays, but if you separate them into the 4 disk groups where it's only 2.7 TB of usable storage, it's still somewhat passable, just not current best practice. &nbsp;You can recover that amount of data in under a day. &nbsp;For use with VMs, you'd never really fill that space, if you're it doing properly, so that you have room for snapshots. &nbsp;They used to do 4 disk RAID 5 groups with with 400 GB and 600 GB SAS drives not that long ago as best practice. &nbsp;It's not really that far from the 900 GB SAS. &nbsp;As long as you're satisfied with the recovery time when it does fail, it's not that great of a problem. &nbsp;The question here is whether you want the speed or uptime. &nbsp;Which is more important to the client? &nbsp;Performance or availability?<br />
<br />
Your client just needs to know what to expect and understand the amount of downtime that would happen. &nbsp;How much money, do they want to put into it and how much downtime can they live with when it happens. &nbsp;RAID 6 can still fail when 3 or more disks fail. &nbsp;While the probability is less likely, if it ever does fail, the downtime for recovery will be even greater when the array holds even more data in a single array. &nbsp;It's all about what you can expect and how much down time you can live with.<br />
<br />
If you have the money, you'd duplicate your main RAID, as HP suggested, with one as a live backup that can be swapped in when the other system fails, and you'd have another slower backup on an onsite NAS as well as an offsite copy of that. &nbsp;I wouldn't live mirror the 2 systems as that would cause the same amount of wear on those disks, but if you need the uptime, then maybe you do.<br />
<br />
My clients also get multiple backups. &nbsp;That's more important than the RAID levels, although RAID levels are important. &nbsp;The layman has an incomplete concept of what RAID is for. &nbsp;It's not backup, but they treat is as such. &nbsp;It can, and will fail at some point. &nbsp;RAID was never originally meant to safeguard data, and it still really isn't meant for that. &nbsp;They was originally used to increase storage space. &nbsp;They added parity to reduce the increase rate of possible array failure because of the introduction of additional drives. &nbsp;That's why there's still necessary backups, multiple backups if you can convince the client.<br />
<br />
**<br />
Just to repeat: &nbsp;You should <b>not</b> make a RAID 5 of larger disk groups. &nbsp;I never said to do that with larger groups. &nbsp;Four disks of 900GB is still passable in RAID 5 depending on actual usage of capacity and recovery time. &nbsp;It's not that much more time to restore 2.7 TB versus 1.8 TB from a decently fast local backup. &nbsp;With snapshots and VM practice, you won't actually be recovering quite that much data. &nbsp;It's also stupid to do RAID 6 with just 4 disks. &nbsp;It will be slow. &nbsp;You should use RAID 10 instead.
</div>


<div>
Hey there<br />
<br />
Got it! &nbsp;Wasn't thinking of URE increasing with larger disks, so glad that was pointed out. &nbsp;However then what do you think if they wanted about 8TB RAW - I would split int to RAID volumes RAID 6 or potentially one RAID 10? &nbsp;Do you ever do a hotspare with RAID10? Thinking..... I think I had it that way on an IBM SAN Years ago... &nbsp;Based on all I have learned here it sounds as if really RAID 10 is the best in terms of reliability and recoverability. &nbsp; &nbsp;So I would do one large RAID 10 with 10-12 disks x 2. &nbsp; If they wont spring for 10, what would you go with then? &nbsp;<br />
Remember they are targeting 8TB Raw for long term expandibility, EX - they want to add AutoDesk Vault for 3d files etc. so would spin up a new VM.
</div>


<div>
This is evolving as a good threat, love to see it.<br />
The hp solution we ended up with was a so called poor mans San. The hp msa240 with 10gb copper nics.<br />
<br />
Even it is raid 6 backed by raid one ssd's it is doing with great ease 10000 iops a second.<br />
<br />
Cheers
</div>


<div>
You should definitely do RAID 10 with the larger set if you have the money. &nbsp;RAID 6 is the compromise when you want the storage space without spending quite as much money.
</div>


<div>
Hey Andy/Pat/Serial<br />
<br />
Thanks for the clarification on RAID 10 disk failures. &nbsp;I don't see using VMware VSan because you need a third host and it is not supported in a VM, so don't want to go there. However you brought up something I was unaware of and was wondering if that is where you would go with this. &nbsp;SAS host connect? &nbsp;So would you consider that that in prod? &nbsp;I have to look at costs. &nbsp;Do you do two controllers per server to connect to what I would assume are two SAS controllers with dual ports on each controller on the SAN so you have fully redundant paths?<br />
Maybe consider 1.8TB 10k SAS drives in a RAID 10 config. &nbsp;<br />
Sounds interesting actually. It could give them an option for shared storage, yes they would pay more for VSphere, but then can then use vmotion etc. &nbsp;We were going to go with a Dell solution, I have to look again, didn't see any such SAS controllers. &nbsp;<br />
We would have two options then to put on the table for them.<br />
I. RAID 10 10 1.8TB 10k drives local replicating via veeam to hostB.<br />
<br />
2. Or possible a SAS SAN with RAID 10 again with 10 1.8TB 10k drives. &nbsp;<br />
<br />
You guys have been a tremendous help on this. Tremendous!
</div>


<div>
You mentioned Dell SCv2000, you can have iSCSI, FC or SAS host interface with that. You can use two SAS Host Bus Adapters in each server for redundancy but one dual-port one is just about as good. You don't get card redundancy with just one dual-port one but you haven't got motherboard redundancy anyway. We call them HBAs rather than controllers because they're dumb interface cards, they don't do RAID as that is done inside the SCv2000 SAN. <br />
<br />
I'd prefer MD3400 to SCv2000 but that's mainly because I know LSI Engenio (now NetApp E-series) and don't know Compellant. Dynamic Disk Pooling on the MD3400 may be of interest, the main advantage is failed disk rebuilds, in traditional RAID you rebuild onto the failed disk so if it is 900GB then 900GB will have to be written to it before you recover parity, with DDP it'll rebuild a bit onto each drive in a pool of 10 so only 90GB to be written to each one before redundancy is restored. DDP is limited to RAID6 type redundancy. Of course any storage that uses distributed spares has similar fast rebuild times.
</div>


<div>
Wasn't sure if they were still referred to as HBAs in SAS. &nbsp;Would you stay purely HP then on the server side as well?<br />
Is the config command line or gui based, I would have to get up to speed on either platform. Was working a bit with EMC over the years, but been a couple of years now. &nbsp;<br />
FYI - RAID 10 appears to be more cost effective and better performance even with the 10k drives! So that looks like the way to go, regardless of SAN or local storage.<br />
Thanks Andy!<br />
Going back a few years, I was told to avoid HP SANS because they were very problematic. &nbsp;I don't have experience with them either way, so dont know first hand. &nbsp;<br />
<br />
I now have to confirm if the &nbsp;512e or 4Kn drives are supported with vSphere 6.<br />
Def want to sell the client at this point on RAID 10 though!<br />
Then will just give them both options two hosts, local storage replicating to hostb, or<br />
Vsphere Essentials Plus, with Vmotion, and SAN I am going to look at the MD3400 a little later today!<br />
<br />
Thanks again!
</div>


<div>
I didn't mention HP, SCv2000 and MD3 are both Dell. I'd stick to the same make of server as storage, makes support easier. If it was HP I'd go for the MSA2040 (Badged from Dot Hill), the original firmware on the early MSA2000s was atrocious, power them off and on and they didn't wait for the disks to spin up but that's been fixed for ages. Possibly why you were told to avoid HP. Pretty unlikely it was an EVA, they're solid as a rock but out of your customer's price range.<br />
<br />
Regarding 512e or 4Kn VMware doesn't know anything about the physical disks, all it sees are logical disks whether in a SAN or on a local RAID controller. You can have a RAID controller connected to 4Kn disks and still see a 512 byte sectored logical disk.
</div>


<div>
Hey Andy<br />
<br />
Thanks again for the wealth of info. &nbsp;So the rebuilds on RAID 10 are not from parity they just just rebuild from the mirror so more likely to rebuild rather than having to read every single block in a RAID 5 array from every single disk &nbsp;- correct? &nbsp; <br />
Def pushing for RAID 10. &nbsp;It really does appear performance with the 10ks in RAID 10 is better than 15k in RAID 6. and you get a better bang for your buck.<br />
Thanks
</div>


<div>
Guys<br />
<br />
Thank you so much for clearing this up for me, you guys were a huge help.<br />
Be well
</div>


<div>
<div class="content wysiwyg-content">
Small environment esx host (a second host for replicating to), They want close to 8 TB of storage, figure 4-6 Windows server VMs.<br />
<br />
For about 20 years have used RAID 5 with a hotspare, with one incident of data loss.<br />
These will be SAS drives hanging off of a Perc 730, can split it to two volumes but they want to have about 8TB of usable for future growth<br />
From a cost perspective it appears that SSDs are still more expensive, and figure they will have the server in prod for 7 years.<br />
I am not sure yet about MTBF for SSDs long term so you might subject yourself to multiple failures over 10 years, introducing more risks on rebuilds.<br />
<br />
Option 1 for VMs RAID 5 WITH HOT SPARE<br />
Option 2 for VMs RAID 6 with hot spare (concerned about performance loss on writes and excessive rebuild times.<br />
Option 3 for VMs RAID 10 - VERY expensive and if I am correct if you lose two drives in a volume, the party is over any way.<br />
<br />
Cost is a factor, no SATA only SAS across the boards/hardware RAID<br />
They will most likely go with essentials so no storage vmotion because of cost.<br />
If I split into two volumes and lose a drive, can bring down guest and move to other volume while volume with failed drive rebuilds reducing the liklihood of losing the RAID 5 volumes if the rebuild fails.<br />
<br />
They will also have a second standby host, where the vms will be replicated to in the event of catastrophic failure.<br />
What do you guys think in this case regarding the required capacity, concerns about performance and limited budet?<br />
<br />
Thanks guys!!!
</div>
</div>


Small environment esx host (a second host for replicating to), They want close to 8 TB of storage, figure 4-6 Windows server VMs.

For about 20 years have used RAID 5 with a hotspare, with one incident of data loss.
These will be SAS drives hanging off of a Perc 730, can split it to two volumes but they want to have about 8TB of usable for future growth
From a cost perspective it appears that SSDs are still more expensive, and figure they will have the server in prod for 7 years.
I am not sure yet about MTBF for SSDs long term so you might subject yourself to multiple failures over 10 years, introducing more risks on rebuilds.

Option 1 for VMs RAID 5 WITH HOT SPARE
Option 2 for VMs RAID 6 with hot spare (concerned about performance loss on writes and excessive rebuild times.
Option 3 for VMs RAID 10 - VERY expensive and if I am correct if you lose two drives in a volume, the party is over any way.

Cost is a factor, no SATA only SAS across the boards/hardware RAID
They will most likely go with essentials so no storage vmotion because of cost.
If I split into two volumes and lose a drive, can bring down guest and move to other volume while volume with failed drive rebuilds reducing the liklihood of losing the RAID 5 volumes if the rebuild fails.

They will also have a second standby host, where the vms will be replicated to in the event of catastrophic failure.
What do you guys think in this case regarding the required capacity, concerns about performance and limited budet?

Thanks guys!!!

Raid - 5 6 or 10 undecided... Thoughts?

RAID stands for Redundant Array of Independent Disks.

RAID is data storage technology that allows multiple drives to be used together as a single virtual drive for reasons such as fault tolerance, reliability and performance. 

There are several different levels of RAID that determine how the data is stored and the level of redundancy achieved across the drives.

RAID

This topic area includes legacy versions of Windows prior to Windows 2000: Windows 3/3.1, Windows 95 and Windows 98, plus any other Windows-related versions including Windows Mobile.

Windows OS

Computer data storage, often called storage or memory, is a technology consisting of computer components and recording media used to retain digital data. In addition to local storage devices like CD and DVD readers, hard drives and flash drives, solid state drives can hold enormous amounts of data in a very small device. Cloud services and other new forms of remote storage also add to the capacity of devices and their ability to access more data without building additional data storage into a device.