AW wrote: > "B. M." wrote: > > optimal partitioning scheme which should last for the > > next 10 years :-) > > I've found that using lvm is a great idea. Resizing volumes is incredibly > easy. You can even easily resize a volume to occupy a portion of a new HDD. > So, my recommendation for new installs is always to use lvm. I almost always set up LVM on system these days. > > Is it still a good idea to put /var on an HDD, not a SSD? I am going to voice an opinion that is counter to what I have read through this thread. If it is a good quality SSD then I bet that SSD outlives the hard disk even if you have it on the busiest file system. With a mixed SSD and spinning hard disk I would probably put /var on the spinning hard disk. Mostly because I wouldn't care about the performance there and so the much slower hard disk would be fine. Leave the fast SSD for something you care about. > My understanding of SSD operation is that they are excellent for > write once, read often scenarios. Data that are going to be changed > often may not be good candidates for an SSD. Because there is so much caching involved the raw flash nand write performance is hidden. Instead because of all of the caching the SSD just seems like a very fast disk regardless of read versus write. I suppose one of us should run some file system performance tests for some real data. If only there were more time in the day to do all of these things. > There's also a tendency for SSD operation to slow if the drive is > re-written often. My information may be a bit outdated as SSD > technology has improved. If this is a problem it would be a vendor specific problem dependent upon the vendor and the firmware. This is definitely not a problem with the Intel SSDs I have been using. If your SSD slows down as it is used then mark that vendor as substandard and buy a better device from a different vendor. Early SSDs vendors that came up from the camera side of things had notable bad performance. Cameras have a completely different data flow from a computer system. But I think most vendors must have solved those problems by now and performance numbers will have converged significantly between vendors. There was a discussion of SSD performance just last month in this thread. https://lists.debian.org/debian-user/2014/07/msg00257.html > > /video HDD, btrfs, 560 GB > > for video editing or series, no backup, not encrypted > > /data HDD, btrfs, encrypted, keyfile, RAID1 (2 x 700 GB). > > With subvolumes for digikam archive, movie archive and music > > I don't know much about software RAID. I usually run a hardware RAID. > However, I'm not so sure it's a good idea to run a software RAID1 on a drive > that will also be used for video editing. It seems to me that the RAID1 > operation would drive down the performance of the HDD enough to significantly > affect video editing. I am surprised by your statement. Can you point to some benchmarks? Mirroring does not improve the write performance but neither does it significantly decrease it. Write performance is pretty much flat between a system with no mirroring, just one disk, and a system with two disks in a software RAID1 mirror. The data must be written to both disks. Read performance is increased with mirroring because the reads can use alternate spindles in parallel. Bob
Attachment:
signature.asc
Description: Digital signature