Will 32GB Ram be enough for my planned upgrade?

[Cheatdeath]

My current build (specs below) has ten 3TB drives in a RaidZ2 and has worked perfectly. However I am getting close to 80% capacity and have purchased eight 10TB drives to handle my future needs. I am worried about how my system will fair with a 8 drive RaidZ2 with 72TB's. I am not able to upgrade this current setup beyond 32GB of ram. I am on a 10Gb network so performance is a worry but not hugely critical. I have never used ssd cache dives etc, would those help me? Would it be advised to even try this much storage with 32GB of ram? Primary purpose of this system is to provide trusted long term storage to my data. Secondary purpose is editing photos and video off the system because of the massive performance gain over my other servers file system. I am hoping 32GB will work or with some ssd cache drives will help, it would make this system last me several more years before a major upgrade is needed.

I am not an IT professional and have self taught myself everything I now about this just a FYI. If I could get some solid insight or advice that would be much appreciated.

Thank you


MB: Supermicro X10SLL-F
CPU: Intel Xeon E3-1231V3 Haswell 3.4GHz
RAM: 32GB Crucial DDR3 ECC 4x8GB
HBA: LSI 9211-8i flashed with IT Firmware

 

[Chris Moore]

I have never used ssd cache dives etc, would those help me?

Not likely, but it depends on how you use the system.

Would it be advised to even try this much storage with 32GB of ram?

Again, it depends more on how you use the system. If you are just using it as a file server, that should be fine.

Secondary purpose is editing photos and video off the system because of the massive performance gain over my other servers file system.

The new drives should be faster than the old ones, so you might actually see some improvement in file transfer speed. If speed is a goal, you may need to revise the build in other ways, but if it is already doing the thing you want it to do, replacing the existing drives with larger ones should not cause a negative impact on that performance. If you find that you want the system to be faster than it is, having more RAM would be the starting point for change, but that would require a system board swap. Depending on the case you have the system mounted in, that might not be a big undertaking. FreeNAS will not be troubled by it.

 

[Cheatdeath]

Not likely, but it depends on how you use the system.

Again, it depends more on how you use the system. If you are just using it as a file server, that should be fine.

The new drives should be faster than the old ones, so you might actually see some improvement in file transfer speed. If speed is a goal, you may need to revise the build in other ways, but if it is already doing the thing you want it to do, replacing the existing drives with larger ones should not cause a negative impact on that performance. If you find that you want the system to be faster than it is, having more RAM would be the starting point for change, but that would require a system board swap. Depending on the case you have the system mounted in, that might not be a big undertaking. FreeNAS will not be troubled by it.

Thank you for your reply. My current drives are 7200RPM with 64MB of cache, the new drives are 5400RPM with 256MB of cache. I am getting 600-700MBps (30-50GB video files) sustained file read transfer speeds on the 10Gb network with the current system and it is overkill tbh. I have it for the same reason someone owns a super car for day to day driving "because I can". The drives should be here during the next week and will go ahead and try it out.

 

[Chris Moore]

Thank you for your reply. My current drives are 7200RPM with 64MB of cache, the new drives are 5400RPM with 256MB of cache. I am getting 600-700MBps (30-50GB video files) sustained file transfer speeds on the 10Gb network with the current system and it is overkill tbh. I have it for the same reason someone owns a super car for day to day driving "because I can". The drives should be here during the next week and will go ahead and try it out.

Your pool layout will have some effect on this. Is this pool of drives broken into multiple vdevs? Also, the transfer rate of the individual drive plays into the equation. I have some older drives that do well to maintain a transfer rate of 120MB/s where the newer 10TB drives are easily sustaining a transfer rate of 230MB/s, and the rotational speed of the platter is not as significant as it might sound because the areal density of the drive is so great.

 

[Cheatdeath]

Your pool layout will have some effect on this. Is this pool of drives broken into multiple vdevs? Also, the transfer rate of the individual drive plays into the equation. I have some older drives that do well to maintain a transfer rate of 120MB/s where the newer 10TB drives are easily sustaining a transfer rate of 230MB/s, and the rotational speed of the platter is not as significant as it might sound because the areal density of the drive is so great.

My ten 3TB drive pool is one vdev and the new pool will be eight 10TB drives in one vdev. The 3TB drives aro going into another system. These 3TB drives are able to provide 190MB/s reads and 180MB/s writes from a single drive in another system and the new 10TB should be slightly faster. My write speeds to the current pool after the drive cache fills drops to about 200MB/s would a ssd cache (ZIL?) in freenas help with writes at all for my setup? As I said above my read speeds from this system seems solid at 600-700MB/s to a nvme drive in my workstation. These speeds are based on larger video file transfers.

 

[Chris Moore]

would a ssd cache (ZIL?) in freenas help with writes at all for my setup?

No. ZIL (ZFS Intent Log) when you put it in a separate device, is called SLOG (for Separate LOG) but this is a LOG, not an actual cache. The thing that would potentially make writes appear faster is if you had enough RAM for the write to be cached in RAM while it is spooled to the disks at the speed of the disks. It would only make it appear faster interactively, but like the super car, it isn't something you need.

my read speeds from this system seems solid at 600-700MB/s to a nvme drive in my workstation

Large file, sequential read, is the fastest kind of file activity.
For both read and write, you might see as much as double the performance if you went to two vdevs of six drives each instead of one vdev because IOPS are connected to vdev count. Each vdev you add to the pool multiplies the effective speed by the number of vdevs. To give you an example, I manage a couple of servers at work that both have 60 drives. One server has those drives broken into 10 vdevs of 6 drives each and the other server has the drives broken into four vdevs of 15 drives each. The server with 10 vdevs can complete a scrub in 18 hours where the server with 4 vdevs takes 97 hours to complete a scrub and the server that takes more time actually has less data in the pool. The faster performance is largely down to the smaller vdevs and higher vdev count.