Geekzone: technology news, blogs, forums
Guest
Welcome Guest.
You haven't logged in yet. If you don't have an account you can register now.
View this topic in a long page with up to 500 replies per page Create new topic
1 | 2 
317 posts

Ultimate Geek
+1 received by user: 75


  Reply # 2077539 22-Aug-2018 15:56
Send private message quote this post

Paul1977:

 

 

 

My only conclusion is that the parity generation on the P420 is VERY FAST, combined with the data being striped across more disks.

 

 

 

 

that would be the FBWC on the controller that would be doing that.




2317 posts

Uber Geek
+1 received by user: 655


  Reply # 2077567 22-Aug-2018 16:40
Send private message quote this post

bagheera:

 

Paul1977:

 

My only conclusion is that the parity generation on the P420 is VERY FAST, combined with the data being striped across more disks.

 

 

that would be the FBWC on the controller that would be doing that.

 

 

I didn't think the 1GB FBWC would make that much of a difference when sequentially reading and writing very large files (files I tested on were approx 20GB). Happy to be wrong about that though.


2437 posts

Uber Geek
+1 received by user: 718

Trusted
Lifetime subscriber

  Reply # 2077568 22-Aug-2018 16:47
Send private message quote this post

Thank you for giving us such great feedback on your testing.

 

Interesting results with the RAID5, given that it has fallen out of favour in recent years.





"4 wheels move the body.  2 wheels move the soul."

“Don't believe anything you read on the net. Except this. Well, including this, I suppose.” Douglas Adams



2317 posts

Uber Geek
+1 received by user: 655


  Reply # 2077579 22-Aug-2018 17:09
Send private message quote this post

Dynamic:

 

Thank you for giving us such great feedback on your testing.

 

Interesting results with the RAID5, given that it has fallen out of favour in recent years.

 

 

No problem.

 

The results would likely be different on faster drives, but on these 7200rpm SATA III drives RAID 5 has come out on top.

 

I didn't bother trying RAID6 as I didn't see how in a 4 disk array it could possibly be any quicker than RAID10 was.

 

I'll say this though, the NZ$120 odd dollars for the HP P420 with 1GB FBWC (Ebay from China) was well worth it.


2012 posts

Uber Geek
+1 received by user: 164

Trusted

  Reply # 2077704 22-Aug-2018 21:27
Send private message quote this post

I thought about this a couple of years ago and went for 4 drives Raid 6.

 

Since then I've added a 5th drive and I really like having the extra redundancy.





Generally known online as OpenMedia, now working for Red Hat New Zealand as a Solution Architect for all things Linux, Virtual and of course Cloud. Still playing with MythTV and digital media on the side.

530 posts

Ultimate Geek
+1 received by user: 125

Subscriber

  Reply # 2077770 23-Aug-2018 08:59
Send private message quote this post

Raid 5/6 sounds good in theory until you work out how long a re-silver would take, risk of additional drive failure etc and effective down time / performance degradation that can exceed that of restoring from backups.



2317 posts

Uber Geek
+1 received by user: 655


  Reply # 2078059 23-Aug-2018 16:15
One person supports this post
Send private message quote this post

solutionz: Raid 5/6 sounds good in theory until you work out how long a re-silver would take, risk of additional drive failure etc and effective down time / performance degradation that can exceed that of restoring from backups.

 

Some fair points.

 

I think I might do a test rebuild on my largest RAID 5 array (not the one I just created). For this particular array I went with RAID5 to maximise usable storage space, but it would be nice to know how long a rebuild takes.

 

And don't worry, I can restore from backup if it all falls over!




2317 posts

Uber Geek
+1 received by user: 655


  Reply # 2078270 24-Aug-2018 09:05
Send private message quote this post

Here's a question for those in the know:

 

When rebuilding a RAID5 array after replacing a disk (specifically on HP controllers with FBWC, if that makes a difference) does the amount of data on the array effect the rebuild time? E.g. if you had an 8TB array, would the rebuild take twice as long if you had 6GB of data as opposed to 3TB?

 

Based on the length of time that the parity initialisation takes when you first create the array (when it has no data), I suspect that the amount of data makes no difference and that the rebuild time is dictated by the size of the disk - but I be interested to know for sure.

 

Can any experts shed some light on this for me?

 

Thanks


317 posts

Ultimate Geek
+1 received by user: 75


  Reply # 2078350 24-Aug-2018 10:43
Send private message quote this post

Paul1977:

 

Here's a question for those in the know:

 

When rebuilding a RAID5 array after replacing a disk (specifically on HP controllers with FBWC, if that makes a difference) does the amount of data on the array effect the rebuild time? E.g. if you had an 8TB array, would the rebuild take twice as long if you had 6GB of data as opposed to 3TB?

 

Based on the length of time that the parity initialisation takes when you first create the array (when it has no data), I suspect that the amount of data makes no difference and that the rebuild time is dictated by the size of the disk - but I be interested to know for sure.

 

Can any experts shed some light on this for me?

 

Thanks

 

 

 

 

never look into this before but my guess longer as it has to rebuild from the parity bits but it may need to do it for the blank space too, if so then you would be right, normal plugin, wait 5 min for it to get a good estimate on time and that how long it will take - so depend alot on how much the drive is being used at the time too. Also from past experiences when you get 1 failed disk, the rest are in dodge state too, and the rebuild overhead could cause another disk to fail (hence why raid 5 is falling out of favour )


467 posts

Ultimate Geek
+1 received by user: 141


  Reply # 2078435 24-Aug-2018 12:06
Send private message quote this post

Paul1977:

 

Here's a question for those in the know:

 

When rebuilding a RAID5 array after replacing a disk (specifically on HP controllers with FBWC, if that makes a difference) does the amount of data on the array effect the rebuild time? E.g. if you had an 8TB array, would the rebuild take twice as long if you had 6GB of data as opposed to 3TB?

 

 

I am not an expert, but the time taken is going to be implementation dependent. Some RAID systems will rebuild everything while others may flag blocks as unused and skip those during a rebuild. How many blocks end up being used would depend on how data is mapped within the RAID, and how the file system on top of it all behaves. At the more advanced end of the spectrum, ZFS can rebuild only the data due to the integration of RAID with the file system.

 

Assuming that every block must be touched during a rebuild, the time taken will be proportional to the size of the disks rather than the amount of data stored on them. The time taken to replace a drive in an eight disk RAID won't be much different from a four disk RAID with the same drive type, irrespective of how much data is on the disk.


78 posts

Master Geek
+1 received by user: 4


  Reply # 2078455 24-Aug-2018 12:21
Send private message quote this post

SirHumphreyAppleby:

 

Assuming that every block must be touched during a rebuild, the time taken will be proportional to the size of the disks rather than the amount of data stored on them. The time taken to replace a drive in an eight disk RAID won't be much different from a four disk RAID with the same drive type, irrespective of how much data is on the disk.

 

 

My experience bears this out. It's the size of the disk being rebuilt rather than the data in them.

 

 




2317 posts

Uber Geek
+1 received by user: 655


  Reply # 2078488 24-Aug-2018 13:06
Send private message quote this post

nitro:

 

SirHumphreyAppleby:

 

Assuming that every block must be touched during a rebuild, the time taken will be proportional to the size of the disks rather than the amount of data stored on them. The time taken to replace a drive in an eight disk RAID won't be much different from a four disk RAID with the same drive type, irrespective of how much data is on the disk.

 

 

My experience bears this out. It's the size of the disk being rebuilt rather than the data in them.

 

 

Thanks guys, that's what suspected.

 

I'm doing a test rebuild right now.

 

I have a RAID5 array made up of 3x 12TB disks giving me 24TB (well, actually 21.86TB) of usable space. Yes, these disks are way bigger than what a lot people say you should put in a RAID5 due to the long rebuild time.

 

I started the rebuild at 10:15pm last night, and as of right now the rebuild is at 83%. Based on that I estimate a total rebuild time of about 18 hours. If my estimate turns out to be accurate, I think that's actually pretty good. Apart from the rebuild, the array has been mostly idle; so that has probably helped with the rebuild time.

 

The array only has 6.59TB of data on it, but based on the discussion above it sounds like the rebuild time would be the same even if it had 20TB of data.

 

Write performance doesn't seem adversely affected and is still well over 300MB/s (although I only tested a couple of large sequential writes, about 15-20GB each).

 

Read performance takes a big hit and is about 50MB/s which is slow but (for me) perfectly usable (again, I only tested a couple of large sequential reads).




2317 posts

Uber Geek
+1 received by user: 655


  Reply # 2078649 24-Aug-2018 17:07
Send private message quote this post

Rebuild complete with a time of a bit under 17 hours. Considering the size of the disks in the array (12TB) I think that is pretty respectable!


1 | 2 
View this topic in a long page with up to 500 replies per page Create new topic

Twitter »

Follow us to receive Twitter updates when new discussions are posted in our forums:



Follow us to receive Twitter updates when news items and blogs are posted in our frontpage:



Follow us to receive Twitter updates when tech item prices are listed in our price comparison site:



Geekzone Live »

Try automatic live updates from Geekzone directly in your browser, without refreshing the page, with Geekzone Live now.



Are you subscribed to our RSS feed? You can download the latest headlines and summaries from our stories directly to your computer or smartphone by using a feed reader.

Alternatively, you can receive a daily email with Geekzone updates.