Reply
Topic Options
- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page
Big Spindle Disk Array
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-09
01:52 PM
2013-04-09
01:52 PM
Big Spindle Disk Array
Busy with configuring and testing a 3200 with 12x 2TB SATA disks. The unit will be used for offloading some minor backup files (VEEAM), storing of replicas (VEEAM) and running passive node of a Exchange 2007 CCR cluster and one domain controller. Nothing to hectic unless all the replicas are started and the passive Exchange cluster node becomes active.
Nonetheless - I was wondering that nowhere in the ReadyNAS documentation there is no mentioning on how performance scaling is effected by the amount of hard disk. Under normal circumstances on heavy I/O systems I would not configure 12x 2TB SATA disks into one array (except maybe NetAPP units with their proven property raiding method) with traditional raiding. I am hoping that X-RAID2 (NFS) is going to deliver but also cannot help but feel I might be disappointed - especially when a) workload starts to increase (linear by adding more virtual machine loads and not necessarily workload increase on one virtual machine) b) during post disk failure rebuilding those huge 2TB slow hard disks. I made the maths of converting to tradditional raid but anything except RAID6 and RAID5 is too much of a storage loss.
Without going into the whole discussion of why I cannot have performance and storage efficiency both - I am wondering if X-RAID2 (in my case dual redundancy) is actually suited for big spindle disk array like with my setup? Currently cannot really see what the true performance is as my X-RAID2 array is currently still in the process to sync. Current disk latency is a little bit concerning the least with values up to 200ms+ (very spikey).
Nonetheless - I was wondering that nowhere in the ReadyNAS documentation there is no mentioning on how performance scaling is effected by the amount of hard disk. Under normal circumstances on heavy I/O systems I would not configure 12x 2TB SATA disks into one array (except maybe NetAPP units with their proven property raiding method) with traditional raiding. I am hoping that X-RAID2 (NFS) is going to deliver but also cannot help but feel I might be disappointed - especially when a) workload starts to increase (linear by adding more virtual machine loads and not necessarily workload increase on one virtual machine) b) during post disk failure rebuilding those huge 2TB slow hard disks. I made the maths of converting to tradditional raid but anything except RAID6 and RAID5 is too much of a storage loss.
Without going into the whole discussion of why I cannot have performance and storage efficiency both - I am wondering if X-RAID2 (in my case dual redundancy) is actually suited for big spindle disk array like with my setup? Currently cannot really see what the true performance is as my X-RAID2 array is currently still in the process to sync. Current disk latency is a little bit concerning the least with values up to 200ms+ (very spikey).
Message 1 of 8
Labels:
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-09
01:58 PM
2013-04-09
01:58 PM
Re: Big Spindle Disk Array
XRAID2 is either RAID-5 (single redundancy) or RAID-6 (dual redundancy). It's a set of wrappers that manage expansion, etc. for standard RAID formats.
Message 2 of 8
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-09
02:04 PM
2013-04-09
02:04 PM
Re: Big Spindle Disk Array
So would this mean general best practice on amount of disks in an array applies to X-RAID2 as well (e.g. RAID5 recommended 5 disk with performance hit or no gain observed at 10+ disks)?
Message 3 of 8
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-09
02:11 PM
2013-04-09
02:11 PM
Re: Big Spindle Disk Array
I'm just a home user, so I haven't tried XRAID-2 with more than 6 disks on a ReadyNAS.
But I believe general best practice is correct. Maybe Chirpa or someone from Netgear can comment?
But I believe general best practice is correct. Maybe Chirpa or someone from Netgear can comment?
Message 4 of 8
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-10
01:21 PM
2013-04-10
01:21 PM
Re: Big Spindle Disk Array
Well something does not seem right - copying data (500GB+) from a ReadyNAS Pro to the 3200 from the same storage network gives me a transfer of about 1MB/s. Both are running X-RAID2, teamed network cards, 1000mb/s, connected on same Cisco 3700 series switch. Looks like might have to convert the X-RAID2 to a RAID10 and loose a lot of storage in the process (about half). I have not enabled jumbo frames on the ReadyNAS and VMware side of things. Would this help a lot or worth it trying out or should I not waste time and just convert to RAID10?
Message 5 of 8
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-11
03:04 PM
2013-04-11
03:04 PM
Re: Big Spindle Disk Array
No technical insight in this? Went through the documentation again but cannot find anything relevant.
Message 6 of 8
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-11
03:36 PM
2013-04-11
03:36 PM
Re: Big Spindle Disk Array
1mbs is definitely not normal and I would hope is not related to the array. There should be zero performance between xraid2 and manual raid5/6 modes. Raid 10 of course has different performance/storage/redundancy characteristics that are obviously different than the other modes.
You could try the standard network trouble shooting, removing 1 complexity at a time;
- make sure flow control is enable on switch/lag
- test/replace each cable
- test 1 device to some other device (ie test 3200 to a windows server, and/or the pro to a windows server)
- unteam 1 or both, use a single nic in standard config
- test a different switch
if the speed problem is vmware specific, ie drag/drop in windows to/from nas is fine, but vm perf is slow, I don't have anything to suggest other to suggest it does not appear to be nas or network problem at that point.
You could try the standard network trouble shooting, removing 1 complexity at a time;
- make sure flow control is enable on switch/lag
- test/replace each cable
- test 1 device to some other device (ie test 3200 to a windows server, and/or the pro to a windows server)
- unteam 1 or both, use a single nic in standard config
- test a different switch
if the speed problem is vmware specific, ie drag/drop in windows to/from nas is fine, but vm perf is slow, I don't have anything to suggest other to suggest it does not appear to be nas or network problem at that point.
Message 7 of 8
- Mark as New
- Bookmark
- Subscribe
- Subscribe to RSS Feed
- Permalink
- Report Inappropriate Content
2013-04-11
03:52 PM
2013-04-11
03:52 PM
Re: Big Spindle Disk Array
Maybe try the copy again with teaming off?
Message 8 of 8