May 30, 201313 yr Anyone know how many drives maximum and space total will be able to be used in the btrfs cache pool?
May 30, 201313 yr Anyone know how many drives maximum and space total will be able to be used in the btrfs cache pool? No, but the new server description indicates it has a 4-drive btrfs cache pool, so it's at least 4
May 30, 201313 yr ... and the status note posted earlier today indicates RC13 is going to be released "later this week" => so within a day or two you should KNOW for sure [Assuming, of course, that RC13 includes the cache pool -- and SURELY it will ... otherwise Tom is planning on releasing v5 Final next week without ever getting feedback on the cache pool !!!]
May 30, 201313 yr Anyone know how many drives maximum and space total will be able to be used in the btrfs cache pool? Can someone explain the practical uses of this cache pool feature?
May 30, 201313 yr Anyone know how many drives maximum and space total will be able to be used in the btrfs cache pool? Can someone explain the practical uses of this cache pool feature? In a word: Fault Tolerance. Right now, when you write to the cache, the data is "at risk" until the Mover moves it to your protected array. With the RAID-1 btrfs cache pool, data in the cache will also be fault tolerant ... i.e. if one of the cache drives fails, you won't lose any data. So once you write data to your UnRAID server, it will be fault-tolerant. This is one reason I do NOT use a cache drive -- I want my data to be protected as soon as it's written to UnRAID. I may add a 2-drive pool after v5 is released now that cached data will no longer be "at risk." In my view, this is a VERY nice improvement.
May 30, 201313 yr I suppose if its a protected area you could also put anything you like on it and have piece of mind that its protected. I may need a bigger case lol, 12 drives max but have 14 SATA Ports to use lol
May 30, 201313 yr it could also be used to increase the speed of the cache. Once faster network cards become common this might be useful.
May 30, 201313 yr it could also be used to increase the speed of the cache. Once faster network cards become common this might be useful. SSD's already easily saturate Gb networks; and 10Gb is a good ways away from being "common" For that matter a good SATA-3 SSD can already do over 4Gb ==> which even if you had a 10Gb network is probably faster than you can "feed" data to UnRAID
May 30, 201313 yr I couldn't warrant putting a couple of SSD's in for cache to many other things to pay for.
May 30, 201313 yr it could also be used to increase the speed of the cache. Once faster network cards become common this might be useful. I might be wrong, but writing to multiple disks will probably not be faster than writing to a single disk, regardless of the file-system involved. I'd not hold my breath waiting for a major performance improvement.
May 30, 201313 yr it could also be used to increase the speed of the cache. Once faster network cards become common this might be useful. I might be wrong, but writing to multiple disks will probably not be faster than writing to a single disk, regardless of the file-system involved. I'd not hold my breath waiting for a major performance improvement. Unless the pool has the option to use RAID-0 you're correct ... and that would, of course eliminate the fault-tolerance, which is BY FAR the #1 advantage of the pool !! Virtually any modern drive can write at very close to Gb Ethernet speeds -- and that's as fast as anyone can send data anyway, so increasing speed isn't an issue.
June 11, 201313 yr Anyone have any updates on this topic? Not really ... what IS clear is that several issues were introduced with RC13 (most seem to be a consequence of the updated Linux kernel), and final will be delayed a bit until those are resolved. The btrfs cache is not going to be introduced until v5 final => and it will be listed as "experimental" when it's released (since it won't have been in any of the release candidates) ... so you'll want to be cautious about what you use it for initially.
Archived
This topic is now archived and is closed to further replies.