ZFS includes data integrity verification, protection against data corruption, support for high storage capacities, great performance, replication, snapshots and copy-on-write clones, and self healing which make it a natural choice for data storage.
The most asked question when talking about raids is, "Which raid is the best?" The best depends on what you are trying to achieve and what you are willing to give up. The question we should be asking ourselves is, "How important is our data ?"
Eventually you will lose a drive and depending on the raid solution will decide if data is lost. If the raid loses a single drive and all your files are gone could be a disaster. You may want to look at a safer raid configuration so the raid can lose two or more drives. The problem with higher data safety is you may have to give up speed or capacity or both.
Raids have three main benefits: Performance, Capacity and Integrity. Performance is how fast the raid will read and write data and is measured in megabytes per second as well as in milliseconds of latency. Capacity is how much does the raid hold. Integrity is how many disks can fail before all the data is lost. The problem is you may not be able to take advantage of all three benefits.
The ASCII triangle below shows all three properties. If you put your mouse cursor in the center of the triangle and move to one property, you see the other properties getting farther away. For example, raid0 is both fast and has the highest capacity, but absolutely no data integrity. Raid1 on the other hand has fantastic integrity and fast reads, but slow writes (multiple copies) and has limited capacity.
Raids have three main advantages over using a single disk.
Performance, capacity and data integrity.
capacity
/\
/ \
/ \
/ \
performance /________\ integrity
Be Safe, Not Sorry. When choosing a raid configuration you may look at raidz or RAID5 and see the speed and capacity benefits and decide it is a good choice. From real world experience we highly suggest NOT using raid5. It is simply not safe enough. We recommend using RAID1 mirroring or RAID6 (double parity) or even RAID7 (triple parity). The problem with raid5 is you only have one drive with parity. When one drive dies the raid5 array is degraded and you can not loose another drive before you lose the entire array; i.e. all your data is lost. What happens most of the time is one drive dies, you replace the drive and the array starts resilvering or rebuilding. This is when the other disks are being stressed and during this rebuild is when you have a very good chance (around 8%) that you you will lose another drive. For more information please take some time to read NetApp Weighs In On Disks which was written in 2007, but is still valid today since drive technology has not changed much.
All tests are run on the same day on the same machine. Our goal was to rule out variability in the hardware so that any differences in performance were the result of the ZFS raid configuration itself.
Bonnie++ is a benchmark suite aimed at performing a number of simple tests reporting hard drive and file system performance. The benchmark test uses database type calls to a single file to simulate creation, reading, and deleting of many small files.
The ZFS pools are created with LZ4 compression disabled so the Bonnie++ test data will be written directly to disk. Bonnie++ generates a 16 gigabyte file which was chosen because it could not fit into 16 gig of RAM or the ZFS ARC. Bonnie++ is using four(4) concurrent threads to better simulate real world server loads. A scripted while loop runs bonnie three(3) times and we report the median (middle) performance metric. After each bonnie run the system sleeps for 30 seconds so the load can subside.
Bonnie++ can do asynchronous I/O, which means the local disk cache of the 4k drive is heavily utilized with a flush between ZFS commits once every 30 seconds. Since the disk cache can artificially inflate the results we choose to disable drive caches completely using Bonnie++ in synchronous test mode only. Syncing after each write will result in lower benchmark values, but the numbers will more closely resemble a server which is heavily loaded and using all of its RAM.
bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
The following table shows a summery of all the tests and lists the number of disks, raid types, capacity and performance metrics for easy comparison. The speeds display "w" for write, "rw" for rewrite and "r" for read and throughput is in megabytes per second.
ZFS uses variable-sized blocks of up to 1024 kilobytes. If data compression (LZJB or LZ4) is enabled, variable block sizes are used. If a block can be compressed to fit into a smaller block size, the smaller size is used on the disk to use less storage and improve IO throughput at the cost of increased CPU use for the compression and decompression operations. Take a look at our compression study further down this page.
When building a RAID it is common practice to use the "power of two plus parity" raid infrastructure to maximize parity striping, speed and well as capacity. When using ZFS the standard RAID rules may not apply, especially when LZ4 compression is active. ZFS can vary the size of the stripes on each disk and compression can make those stripes unpredictable. The current rule of thumb when making a ZFS raid is:
The server is setup using an Avago LSI Host Bus Adapter (HBA) and not a raid card. The HBA is connected to the SAS expander using a single multilane cable to control all 24 drives. LZ4 compression is disabled and all writes are synced in real time by the testing suite, Bonnie++. We wanted to test the raid configurations against the drives themselves without LZ4 compression or extra RAM to use for ZFS ARC. You can expect your speeds to easily increase by a factor of two(2) when storing compressible data with LZ4 enabled and when adding RAM at least twice as large as the data sets being read and written.
ZFS Raid Speed Capacity and Performance Benchmarks
(speeds in megabytes per second)
1x 4TB, single drive, 3.7 TB, w=108MB/s , rw=50MB/s , r=204MB/s
2x 4TB, mirror (raid1), 3.7 TB, w=106MB/s , rw=50MB/s , r=488MB/s
2x 4TB, stripe (raid0), 7.5 TB, w=237MB/s , rw=73MB/s , r=434MB/s
3x 4TB, mirror (raid1), 3.7 TB, w=106MB/s , rw=49MB/s , r=589MB/s
3x 4TB, stripe (raid0), 11.3 TB, w=392MB/s , rw=86MB/s , r=474MB/s
3x 4TB, raidz1 (raid5), 7.5 TB, w=225MB/s , rw=56MB/s , r=619MB/s
4x 4TB, 2 striped mirrors, 7.5 TB, w=226MB/s , rw=53MB/s , r=644MB/s
4x 4TB, raidz2 (raid6), 7.5 TB, w=204MB/s , rw=54MB/s , r=183MB/s
5x 4TB, raidz1 (raid5), 15.0 TB, w=469MB/s , rw=79MB/s , r=598MB/s
5x 4TB, raidz3 (raid7), 7.5 TB, w=116MB/s , rw=45MB/s , r=493MB/s
6x 4TB, 3 striped mirrors, 11.3 TB, w=389MB/s , rw=60MB/s , r=655MB/s
6x 4TB, raidz2 (raid6), 15.0 TB, w=429MB/s , rw=71MB/s , r=488MB/s
10x 4TB, 2 striped 5x raidz, 30.1 TB, w=675MB/s , rw=109MB/s , r=1012MB/s
11x 4TB, raidz3 (raid7), 30.2 TB, w=552MB/s , rw=103MB/s , r=963MB/s
12x 4TB, 6 striped mirrors, 22.6 TB, w=643MB/s , rw=83MB/s , r=962MB/s
12x 4TB, 2 striped 6x raidz2, 30.1 TB, w=638MB/s , rw=105MB/s , r=990MB/s
12x 4TB, raidz (raid5), 41.3 TB, w=689MB/s , rw=118MB/s , r=993MB/s
12x 4TB, raidz2 (raid6), 37.4 TB, w=317MB/s , rw=98MB/s , r=1065MB/s
12x 4TB, raidz3 (raid7), 33.6 TB, w=452MB/s , rw=105MB/s , r=840MB/s
22x 4TB, 2 striped 11x raidz3, 60.4 TB, w=567MB/s , rw=162MB/s , r=1139MB/s
23x 4TB, raidz3 (raid7), 74.9 TB, w=440MB/s , rw=157MB/s , r=1146MB/s
24x 4TB, 12 striped mirrors, 45.2 TB, w=696MB/s , rw=144MB/s , r=898MB/s
24x 4TB, raidz (raid5), 86.4 TB, w=567MB/s , rw=198MB/s , r=1304MB/s
24x 4TB, raidz2 (raid6), 82.0 TB, w=434MB/s , rw=189MB/s , r=1063MB/s
24x 4TB, raidz3 (raid7), 78.1 TB, w=405MB/s , rw=180MB/s , r=1117MB/s
24x 4TB, striped raid0, 90.4 TB, w=692MB/s , rw=260MB/s , r=1377MB/s
######## ###########
######## Below are the zpool commands and RAW Bonnie++ output ###########
######## ###########
###############################################################################
1x 4TB, single drive, 3.7 TB, w=108MB/s , rw=50MB/s , r=204MB/s
root@FreeBSDzfs: zpool create storage da0
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 108834 25 50598 11 204522 9 393.0 4
Latency 1992ms 2372ms 1200ms 289ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
da0 ONLINE 0 0 0
###############################################################################
2x 4TB, mirror (raid1), 3.7 TB, w=106MB/s , rw=50MB/s , r=488MB/s
root@FreeBSDzfs: zpool create storage mirror da0 da1
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 106132 25 50303 10 488762 23 445.7 4
Latency 9435ms 4173ms 220ms 195ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
###############################################################################
2x 4TB, stripe (raid0), 7.5 TB, w=237MB/s , rw=73MB/s , r=434MB/s
root@FreeBSDzfs: zpool create storage da0 da1
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 237933 43 73012 14 434041 20 513.1 5
Latency 505ms 4059ms 212ms 197ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
###############################################################################
3x 4TB, mirror (raid1), 3.7 TB, w=106MB/s , rw=49MB/s , r=589MB/s
root@FreeBSDzfs: zpool create storage mirror da0 da1 da2
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 106749 24 49659 11 589971 27 457.3 5
Latency 12593ms 4069ms 134ms 191ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
###############################################################################
3x 4TB, stripe (raid0), 11.3 TB, w=392MB/s , rw=86MB/s , r=474MB/s
root@FreeBSDzfs: zpool create storage da0 da1 da2
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 392378 59 86799 16 474157 22 829.3 8
Latency 315us 4038ms 129ms 141ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
###############################################################################
3x 4TB, raidz1 (raid5), 7.5 TB, w=225MB/s , rw=56MB/s , r=619MB/s
root@FreeBSDzfs: zpool create storage raidz da0 da1 da2
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 225143 40 56590 11 619315 30 402.9 5
Latency 2204ms 4052ms 896ms 177ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
###############################################################################
4x 4TB, 2 striped mirrors, 7.5 TB, w=226MB/s , rw=53MB/s , r=644MB/s
root@FreeBSDzfs: zpool create storage mirror da0 da1 mirror da2 da3
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 226933 42 53878 9 644043 31 427.8 4
Latency 1356ms 4066ms 1638ms 221ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
mirror-1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
###############################################################################
4x 4TB, raidz2 (raid6), 7.5 TB, w=204MB/s , rw=54MB/s , r=183MB/s
root@FreeBSDzfs: zpool create storage raidz2 da0 da1 da2 da3
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 204686 36 54769 13 183487 9 365.5 6
Latency 5292ms 6356ms 449ms 233ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
###############################################################################
5x 4TB, raidz1 (raid5), 15.0 TB, w=469MB/s , rw=79MB/s , r=598MB/s
root@FreeBSDzfs: zpool create storage raidz1 da0 da1 da2 da3 da4
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 469287 69 79083 15 598770 29 561.2 7
Latency 1202us 4047ms 137ms 157ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
###############################################################################
5x 4TB, raidz3 (raid7), 7.5 TB, w=116MB/s , rw=45MB/s , r=493MB/s
root@FreeBSDzfs: zpool create storage raidz3 da0 da1 da2 da3 da4
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 116395 21 45652 9 493679 25 397.7 5
Latency 27836ms 6418ms 95509us 152ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz3-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
###############################################################################
6x 4TB, 3 striped mirrors, 11.3 TB, w=389MB/s , rw=60MB/s , r=655MB/s
root@FreeBSDzfs: zpool create storage mirror da0 da1 mirror da2 da3 mirror da4 da5
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 389316 62 60493 11 655534 32 804.4 9
Latency 549us 2473ms 1997ms 186ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
mirror-1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
mirror-2 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
###############################################################################
6x 4TB, raidz2 (raid6), 15.0 TB, w=429MB/s , rw=71MB/s , r=488MB/s
root@FreeBSDzfs: zpool create storage raidz2 da0 da1 da2 da3 da4 da5
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 429953 67 71505 14 488952 25 447.7 6
Latency 358us 4057ms 197ms 181ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
###############################################################################
10x 4TB, 2 striped 5x raidz, 30.1 TB, w=675MB/s , rw=109MB/s , r=1012MB/s
root@FreeBSDzfs: zpool create storage raidz da0 da1 da2 da3 da4 raidz da5 da6 da7 da8 da9
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 675195 93 109169 21 1012768 50 817.6 9
Latency 11619us 4471ms 84450us 110ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
raidz1-1 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
###############################################################################
11x 4TB, raidz3 (raid7), 30.2 TB, w=552MB/s , rw=103MB/s , r=963MB/s
root@FreeBSDzfs: zpool create storage raidz3 da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 552264 79 103947 20 963511 48 545.6 8
Latency 7373us 4045ms 84226us 144ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz3-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
###############################################################################
12x 4TB, 6 striped mirrors, 22.6 TB, w=643MB/s , rw=83MB/s , r=962MB/s
root@FreeBSDzfs: zpool create storage mirror da0 da1 mirror da2 da3 mirror da4 da5 mirror da6 da7 mirror da8 da9 mirror da10 da11
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 643886 91 83717 15 962904 47 1257 13
Latency 17335us 4040ms 1884ms 175ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
mirror-1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
mirror-2 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
mirror-3 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
mirror-4 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
mirror-5 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
###############################################################################
12x 4TB, 2 striped 6x raidz2, 30.1 TB, w=638MB/s , rw=105MB/s , r=990MB/s
root@FreeBSDzfs: zpool create storage raidz2 da0 da1 da2 da3 da4 da5 raidz2 da6 da7 da8 da9 da10 da11
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 638918 89 105631 21 990167 49 773.7 10
Latency 15398us 6170ms 104ms 113ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
raidz2-1 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
###############################################################################
12x 4TB, raidz (raid5), 41.3 TB, w=689MB/s , rw=118MB/s , r=993MB/s
root@FreeBSDzfs: zpool create storage raidz da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 689424 96 118573 23 993618 52 647.3 9
Latency 14466us 3700ms 127ms 141ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
###############################################################################
12x 4TB, raidz2 (raid6), 37.4 TB, w=317MB/s , rw=98MB/s , r=1065MB/s
root@FreeBSDzfs: zpool create storage raidz2 da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 317741 45 98974 22 1065349 56 495.5 8
Latency 11742ms 4062ms 90593us 150ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
###############################################################################
12x 4TB, raidz3 (raid7), 33.6 TB, w=452MB/s , rw=105MB/s , r=840MB/s
root@FreeBSDzfs: zpool create storage raidz3 da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 452865 66 105396 24 840136 44 476.2 7
Latency 706us 4069ms 2050ms 165ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz3-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
###############################################################################
22x 4TB, 2 striped 11x raidz3, 60.4 TB, w=567MB/s , rw=162MB/s , r=1139MB/s
root@FreeBSDzfs: zpool create storage raidz3 da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 raidz3 da11 da12 da13 da14 da15 da16 da17 da18 da19 da20 da21
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 567667 83 162462 32 1139088 59 770.3 13
Latency 4581us 2700ms 78597us 116ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz3-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
raidz3-1 ONLINE 0 0 0
da11 ONLINE 0 0 0
da12 ONLINE 0 0 0
da13 ONLINE 0 0 0
da14 ONLINE 0 0 0
da15 ONLINE 0 0 0
da16 ONLINE 0 0 0
da17 ONLINE 0 0 0
da18 ONLINE 0 0 0
da19 ONLINE 0 0 0
da20 ONLINE 0 0 0
da21 ONLINE 0 0 0
###############################################################################
23x 4TB, raidz3 (raid7), 74.9 TB, w=440MB/s , rw=157MB/s , r=1146MB/s
root@FreeBSDzfs: zpool create storage raidz3 da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11 da12 da13 da14 da15 da16 da17 da18 da19 da20 da21 da22
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 440656 64 157066 34 1146275 76 408.8 8
Latency 21417us 2324ms 154ms 195ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz3-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
da12 ONLINE 0 0 0
da13 ONLINE 0 0 0
da14 ONLINE 0 0 0
da15 ONLINE 0 0 0
da16 ONLINE 0 0 0
da17 ONLINE 0 0 0
da18 ONLINE 0 0 0
da19 ONLINE 0 0 0
da20 ONLINE 0 0 0
da21 ONLINE 0 0 0
da22 ONLINE 0 0 0
###############################################################################
24x 4TB, 12 striped mirrors, 45.2 TB, w=696MB/s , rw=144MB/s , r=898MB/s
root@FreeBSDzfs: zpool create storage mirror da0 da1 mirror da2 da3 mirror da4 da5 mirror da6 da7 mirror da8 da9 mirror da10 da11 mirror da12 da13 mirror da14 da15 mirror da16 da17 mirror da18 da19 mirror da20 da21 mirror da22 da23
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 696733 94 144526 27 898137 51 1520 16
Latency 17247us 2850ms 2014ms 79930us
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
mirror-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
mirror-1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
mirror-2 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
mirror-3 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
mirror-4 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
mirror-5 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
mirror-6 ONLINE 0 0 0
da12 ONLINE 0 0 0
da13 ONLINE 0 0 0
mirror-7 ONLINE 0 0 0
da14 ONLINE 0 0 0
da15 ONLINE 0 0 0
mirror-8 ONLINE 0 0 0
da16 ONLINE 0 0 0
da17 ONLINE 0 0 0
mirror-9 ONLINE 0 0 0
da18 ONLINE 0 0 0
da19 ONLINE 0 0 0
mirror-10 ONLINE 0 0 0
da20 ONLINE 0 0 0
da21 ONLINE 0 0 0
mirror-11 ONLINE 0 0 0
da22 ONLINE 0 0 0
da23 ONLINE 0 0 0
###############################################################################
24x 4TB, raidz (raid5), 86.4 TB, w=567MB/s , rw=198MB/s , r=1304MB/s
root@FreeBSDzfs: zpool create storage raidz da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11 da12 da13 da14 da15 da16 da17 da18 da19 da20 da21 da22 da23
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 567164 82 198696 42 1304530 81 620.7 11
Latency 36614us 2252ms 70871us 141ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz1-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
da12 ONLINE 0 0 0
da13 ONLINE 0 0 0
da14 ONLINE 0 0 0
da15 ONLINE 0 0 0
da16 ONLINE 0 0 0
da17 ONLINE 0 0 0
da18 ONLINE 0 0 0
da19 ONLINE 0 0 0
da20 ONLINE 0 0 0
da21 ONLINE 0 0 0
da22 ONLINE 0 0 0
da23 ONLINE 0 0 0
###############################################################################
24x 4TB, raidz2 (raid6), 82.0 TB, w=434MB/s , rw=189MB/s , r=1063MB/s
root@FreeBSDzfs: zpool create storage raidz2 da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11 da12 da13 da14 da15 da16 da17 da18 da19 da20 da21 da22 da23
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 434394 63 189521 39 1063626 75 516.9 12
Latency 11369us 2130ms 80053us 153ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
da12 ONLINE 0 0 0
da13 ONLINE 0 0 0
da14 ONLINE 0 0 0
da15 ONLINE 0 0 0
da16 ONLINE 0 0 0
da17 ONLINE 0 0 0
da18 ONLINE 0 0 0
da19 ONLINE 0 0 0
da20 ONLINE 0 0 0
da21 ONLINE 0 0 0
da22 ONLINE 0 0 0
da23 ONLINE 0 0 0
###############################################################################
24x 4TB, raidz3 (raid7), 78.1 TB, w=405MB/s , rw=180MB/s , r=1117MB/s
root@FreeBSDzfs: zpool create storage raidz3 da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11 da12 da13 da14 da15 da16 da17 da18 da19 da20 da21 da22 da23
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 405579 62 180470 39 1117221 70 592.4 13
Latency 622ms 1959ms 94830us 187ms
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
raidz3-0 ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
da12 ONLINE 0 0 0
da13 ONLINE 0 0 0
da14 ONLINE 0 0 0
da15 ONLINE 0 0 0
da16 ONLINE 0 0 0
da17 ONLINE 0 0 0
da18 ONLINE 0 0 0
da19 ONLINE 0 0 0
da20 ONLINE 0 0 0
da21 ONLINE 0 0 0
da22 ONLINE 0 0 0
da23 ONLINE 0 0 0
###############################################################################
24x 4TB, striped raid0, 90.4 TB, w=692MB/s , rw=260MB/s , r=1377MB/s
root@FreeBSDzfs: zpool create storage da0 da1 da2 da3 da4 da5 da6 da7 da8 da9 da10 da11 da12 da13 da14 da15 da16 da17 da18 da19 da20 da21 da22 da23
root@FreeBSDzfs: bonnie++ -u root -r 1024 -s 16384 -d /storage -f -b -n 1 -c 4
Version 1.97 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 4 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
FreeBSDzfs 16G 692351 95 260259 50 1377547 75 1921 19
Latency 12856us 1670ms 49017us 59388us
root@FreeBSDzfs: zpool status
pool: storage
state: ONLINE
scan: none requested
config:
NAME STATE READ WRITE CKSUM
storage ONLINE 0 0 0
da0 ONLINE 0 0 0
da1 ONLINE 0 0 0
da2 ONLINE 0 0 0
da3 ONLINE 0 0 0
da4 ONLINE 0 0 0
da5 ONLINE 0 0 0
da6 ONLINE 0 0 0
da7 ONLINE 0 0 0
da8 ONLINE 0 0 0
da9 ONLINE 0 0 0
da10 ONLINE 0 0 0
da11 ONLINE 0 0 0
da12 ONLINE 0 0 0
da13 ONLINE 0 0 0
da14 ONLINE 0 0 0
da15 ONLINE 0 0 0
da16 ONLINE 0 0 0
da17 ONLINE 0 0 0
da18 ONLINE 0 0 0
da19 ONLINE 0 0 0
da20 ONLINE 0 0 0
da21 ONLINE 0 0 0
da22 ONLINE 0 0 0
da23 ONLINE 0 0 0
###############################################################################
The 24 slot raid chassis is filled with Samsung 840 256GB SSD (MZ-7PD256BW) drives. Drives are connected through an Avago LSI 9207-8i HBA controller installed in a PCIe 16x slot. TRIM is not used and not needed due to ZFS's copy on write drive leveling.
1x 256GB a single drive 232 gigabytes ( w= 441MB/s , rw=224MB/s , r= 506MB/s ) 2x 256GB raid0 striped 464 gigabytes ( w= 933MB/s , rw=457MB/s , r=1020MB/s ) 2x 256GB raid1 mirror 232 gigabytes ( w= 430MB/s , rw=300MB/s , r= 990MB/s ) 3x 256GB raid5, raidz1 466 gigabytes ( w= 751MB/s , rw=485MB/s , r=1427MB/s ) 4x 256GB raid6, raidz2 462 gigabytes ( w= 565MB/s , rw=442MB/s , r=1925MB/s ) 5x 256GB raid5, raidz1 931 gigabytes ( w= 817MB/s , rw=610MB/s , r=1881MB/s ) 5x 256GB raid7, raidz3 464 gigabytes ( w= 424MB/s , rw=316MB/s , r=1209MB/s ) 6x 256GB raid6, raidz2 933 gigabytes ( w= 721MB/s , rw=530MB/s , r=1754MB/s ) 7x 256GB raid7, raidz3 934 gigabytes ( w= 591MB/s , rw=436MB/s , r=1713MB/s ) 9x 256GB raid5, raidz1 1.8 terabytes ( w= 868MB/s , rw=618MB/s , r=1978MB/s ) 10x 256GB raid6, raidz2 1.8 terabytes ( w= 806MB/s , rw=511MB/s , r=1730MB/s ) 11x 256GB raid7, raidz3 1.8 terabytes ( w= 659MB/s , rw=448MB/s , r=1681MB/s ) 17x 256GB raid5, raidz1 3.7 terabytes ( w= 874MB/s , rw=574MB/s , r=1816MB/s ) 18x 256GB raid6, raidz2 3.7 terabytes ( w= 788MB/s , rw=532MB/s , r=1589MB/s ) 19x 256GB raid7, raidz3 3.7 terabytes ( w= 699MB/s , rw=400MB/s , r=1183MB/s ) 24x 256GB raid0 striped 5.5 terabytes ( w=1620MB/s , rw=796MB/s , r=2043MB/s )
Raid 6 or raidz2 is a good mix of all three properties. Good speeds, integrity and capacity. The ability to lose two drives is fine for a home or office raid where you can get to the machine in a short amount of time to put in another disk. For data center use raidz2 (raid6) is good if you have a hot spare in place. Otherwise, raidz3 (raid7) might be better for data center use as it might take an admin a week to get there to replace a disk. Of course, if you use raidz2 (raid6) with a hot spare you might as well use raid7. Raid 7 would offer the same total amount of space and allow the raid to lose ANY three drives.
We found that no matter which spinning hard drive we used, the speeds and latencies were about the same. You do not have the get the most expensive SAS drives like we used. Look for any manufacture which label their drives as RE or Raid Enabled or "For NAS systems." Some Western Digital Red 3TB (WD30EFRX) hard drives performed just as well. 7200 rpm drives are around 10 percent faster then 5400 rpm drives, but the 7200 rpm drives run hotter. We recommend staying away from power saving or ECO mode drives like the Western Digital Green (WD30EZRX) series. These green drives seem to perform poorly in a raid configuration due to their ability to spin down, go into sleep mode and long SATA timings. If you are looking for all out speed, the best performance increase you are going to get is to switch from spinning hard drives to SSDs and/or enable ZFS compression.
Reliability of the pool is directly proportional to the number of disks, quality of the hardware and re-silver time. As the number of disks increase the reliability decreases because there is a higher chance for another disk to fail. Using enterprise SAS drives which have a higher meantime between failure should increase reliability compared to using desktop based personal drives. The longer the raid takes to resilver the higher the chance is another drive will fail during the re-silver process. To deal with the chance of pool failure with more disks we should increase the amount of parity disks. Higher parity count will reduce the MTTDL, Mean Time To Data Loss.
Some Random Observations: Always use LZ4 compression and stick as much RAM as you can afford into the machine. When deciding how many drives to use, drive types do not matter, SAS or SATA are both fine. The more disks in a pool, less space lost to parity equals greater space efficiency. Hard drives die in batches; try to vary the manufacturing date of the drives or buy from different companies. In general, mirrors are better for random IOPs. IOPs do not seem to differ between raidz, raidz2 or raidz3.
Yes, enable LZ4 compression. The command "zfs set compression=lz4 tank" will enable compression on the "tank" pool. The long answer is also yes, and you want to be aware of the positives and negatives to make an informed decision.
Spinning drives in general are really, really slow. The OS wastes a significant amount of time just waiting to get data to and from the hard drives. The main argument against compression on the fly is it uses CPU time when reading or writing to the array and also adds latency. If you have been using computers for a while you might remember "Stacker" from the 1990's. Ideally, we want the system to contact the drives as little as possible because they are so slow.
Compression can reduce the amount of blocks the drive needs to access and the result is faster speeds at the cost of CPU time. What you gain is less blocks read or written to for even moderately compressible files meaning you have effectively increased the areal density of the drive platters. Compression allows the drive heads to move over a smaller area to access the same amount of data.
The developers of ZFS offer lz4 compression which is disabled by default. Take note that ZFS and compression use multiple CPU cores so the more cores and the faster they are the faster compression will be. With LZ4 compression enabled, the system will use around 20% more CPU time to compress and uncompress files compared to handling raw data. On average lz4 will compress data to a maximum of 2x to 3x.
In our tests we show single drive reads go from 150MB/sec to 1,174MB/sec of compressed Bonnie++ test data with lzjb. Even compressed files like bzip2, zip, mp3, avi, mkv and webm will not waste much CPU time as lz4 is a very inexpensive compression method.
In FreeBSD 10 both lzjb and lz4 are available in ZFS. LZJB is fast, but LZ4 is faster with 500MB/sec compression and 1.5GB/sec decompression speeds. This makes LZ4 50% faster on compression and 80% on decompression compared to LZJB. In FreeBSD 9.1 you may need to use "zpool set feature@lz4_compress=enabled poolname" to enable lz4 on your pool as lz4 is not backwards compatible with non-lz4 versions of zfs.
LZ4's performance on incompressible data is equally impressive. LZ4 achieves higher performance by incorporating an "early abort" mechanism which triggers if LZ4 can not reach a minimum compression ratio of 12.5%. LZ4 is perfect for admins who would like to use compression as a default option for all pools. Enabling LZ4 on all volumes means compression will be used where appropriate, while making sure the algorithm will not waste CPU time on incompressible data and reads will not incur a latency penalty due to transparent decompression. Just incredible.
The following table shows a few raid types without compression and then with lzjb and lz4 compression enabled. Bonnie++ tests use database like files which compress fairly well. You will notice as we add more physical disks, compression makes less of an impact on performance. In fact, with lz4 compression we max out read and write speeds with only three disks in raidz (raid5). Our suggestion is to enabled compression on all arrays because the amount of throughput you gain outweighs the CPU usage and you get to compress data which can save space.
# The following Bonnie++ tests show the speeds of volumes with compression off # and then on with compression=lzjb and compression=lz4 of well compressed data. off 1x 2TB a single drive 1.8 terabytes ( w=131MB/s , rw= 66MB/s , r= 150MB/s ) lzjb 1x 2TB a single drive 1.8 terabytes ( w=445MB/s , rw=344MB/s , r=1174MB/s ) lz4 1x 2TB a single drive 1.8 terabytes ( w=471MB/s , rw=351MB/s , r=1542MB/s ) off 1x 256GB a single drive 232 gigabytes ( w=441MB/s , rw=224MB/s , r= 506MB/s ) SSD lzjb 1x 256GB a single drive 232 gigabytes ( w=510MB/s , rw=425MB/s , r=1290MB/s ) SSD off 2x 2TB raid1 mirror 1.8 terabytes ( w=126MB/s , rw= 79MB/s , r= 216MB/s ) lzjb 2x 2TB raid1 mirror 1.8 terabytes ( w=461MB/s , rw=386MB/s , r=1243MB/s ) lz4 2x 2TB raid1 mirror 1.8 terabytes ( w=398MB/s , rw=354MB/s , r=1537MB/s ) off 3x 2TB raid5, raidz1 3.6 terabytes ( w=279MB/s , rw=131MB/s , r= 281MB/s ) lzjb 3x 2TB raid5, raidz1 3.6 terabytes ( w=479MB/s , rw=366MB/s , r=1243MB/s ) lz4 3x 2TB raid5, raidz1 3.6 terabytes ( w=517MB/s , rw=453MB/s , r=1587MB/s ) off 5x 2TB raid5, raidz1 7.1 terabytes ( w=469MB/s , rw=173MB/s , r= 406MB/s ) lzjb 5x 2TB raid5, raidz1 7.1 terabytes ( w=478MB/s , rw=392MB/s , r=1156MB/s ) lz4 5x 2TB raid5, raidz1 7.1 terabytes ( w=516MB/s , rw=437MB/s , r=1560MB/s ) off 5x 256GB raid5, raidz1 931 gigabytes ( w= 817MB/s , rw=610MB/s , r=1881MB/s ) SSD lzjb 5x 256GB raid5, raidz1 931 gigabytes ( w= 515MB/s , rw=415MB/s , r=1223MB/s ) SSD off 7x 2TB raid7, raidz3 7.1 terabytes ( w=393MB/s , rw=169MB/s , r= 423MB/s ) lzjb 7x 2TB raid7, raidz3 7.1 terabytes ( w=469MB/s , rw=378MB/s , r=1127MB/s ) lz4 7x 2TB raid7, raidz3 7.1 terabytes ( w=507MB/s , rw=436MB/s , r=1532MB/s ) off 12x 2TB raid5, raidz1 19 terabytes ( w=521MB/s , rw=272MB/s , r= 738MB/s ) lzjb 12x 2TB raid5, raidz1 19 terabytes ( w=487MB/s , rw=391MB/s , r=1105MB/s ) lz4 12x 2TB raid5, raidz1 19 terabytes ( w=517MB/s , rw=441MB/s , r=1557MB/s ) off 17x 2TB raid5, raidz1 28 terabytes ( w=468MB/s , rw=267MB/s , r= 874MB/s ) lzjb 17x 2TB raid5, raidz1 28 terabytes ( w=478MB/s , rw=380MB/s , r=1096MB/s ) lz4 17x 2TB raid5, raidz1 28 terabytes ( w=502MB/s , rw=430MB/s , r=1473MB/s ) off 24x 2TB raid5, raidz1 40 terabytes ( w=528MB/s , rw=291MB/s , r= 929MB/s ) lzjb 24x 2TB raid5, raidz1 40 terabytes ( w=478MB/s , rw=382MB/s , r=1081MB/s ) lz4 24x 2TB raid5, raidz1 40 terabytes ( w=504MB/s , rw=431MB/s , r=1507MB/s ) off 24x 256GB raid0 stripe 5.5 terabytes ( w=1340MB/s , rw=796MB/s , r=2037MB/s ) SSD lzjb 24x 256GB raid0 stripe 5.5 terabytes ( w=1032MB/s , rw=844MB/s , r=2597MB/s ) SSD # The raw bonnie output 1x 2TB a single drive 1.8 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 445544 78 344412 67 1174294 80 285.9 3 Latency 6118ms 5142ms 273ms 502ms 1x 2TB a single drive 1.8 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 471248 76 351900 61 1542189 88 14053 178 Latency 8638ms 9357ms 720ms 44143us 1x 256GB a single drive 232 terabytes SSD # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsSSD 80G 510613 88 425359 81 1290425 86 9716 140 Latency 509ms 553ms 277ms 13602us 2x 2TB raid1 mirror 1.8 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 461530 83 386759 78 1243555 85 397.2 15 Latency 5031ms 4847ms 276ms 304ms 2x 2TB raid1 mirror 1.8 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 398572 67 354963 64 1537069 90 1157 14 Latency 10477ms 7119ms 795ms 76540us 3x 2TB raid5, raidz1 3.6 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 479200 85 366110 75 1243482 87 260.8 3 Latency 3580ms 3717ms 500ms 342ms 3x 2TB raid5, raidz1 3.6 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 517140 87 453732 82 1587063 95 899.5 10 Latency 471ms 627ms 325ms 70999us 5x 2TB raid5, raidz1 7.1 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 478755 87 392926 81 1156163 83 256.9 4 Latency 807ms 1269ms 338ms 417ms 5x 2TB raid5, raidz1 7.1 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 516294 88 437536 80 1560222 93 1190 14 Latency 522ms 2311ms 301ms 61362us 5x 2TB raid5, raidz1 931 gigabytes SSD # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsSSD 80G 515770 88 415428 79 1223482 85 9288 147 Latency 495ms 1192ms 87411us 12334us 7x 2TB raid7, raidz3 7.1 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 469870 86 378324 78 1127670 81 260.9 5 Latency 1862ms 1055ms 210ms 371ms 7x 2TB raid7, raidz3 7.1 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 507960 88 436097 82 1532414 92 614.7 10 Latency 509ms 1576ms 187ms 61843us 12x 2TB raid5, raidz1 19 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 487671 87 391920 81 1105113 83 248.9 4 Latency 503ms 1128ms 409ms 323ms 12x 2TB raid5, raidz1 19 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 517168 89 441478 81 1557655 93 1335 22 Latency 462ms 847ms 267ms 61475us 17x 2TB raid5, raidz1 28 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 478964 87 380927 80 1096636 82 280.7 5 Latency 2051ms 802ms 179ms 363ms 17x 2TB raid5, raidz1 28 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 502536 88 430811 81 1473594 89 1217 14 Latency 489ms 726ms 261ms 58670us 24x 2TB raid5, raidz1 40 terabytes # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 478082 87 382614 79 1081919 80 278.9 6 Latency 990ms 1013ms 185ms 492ms 24x 2TB raid5, raidz1 40 terabytes # zfs set compression=lz4 tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsFBSD10 80G 504239 87 431002 81 1507097 90 1645 22 Latency 514ms 844ms 58128us 43527us 24x 2TB raid0 stripe 5.5 terabytes SSD # zfs set compression=off tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 587993 97 359645 65 1040082 61 7410 108 Latency 109ms 827ms 577ms 13935us 24x 2TB raid0 stripe 5.5 terabytes SSD # zfs set compression=lzjb tank Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 1032588 88 844972 82 2597192 85 10210 142 Latency 538ms 571ms 235ms 14151us
The performance of your RAID is highly dependent on your hardware and OS drivers. If you are using the motherboard SATA connectors, they are not going to perform as well as a SATA expander or a dedicated raid card. Just because the SATA port says SATA 6 and comes with fancy cables does not mean the port can move data quickly. The onboard chipsets are normally the cheapest silicon the manufacturer can get away with. This sentiment is true for most of the substandard on motherboard network ports too.
On mailing lists and forums there are posts which state ZFS is slow and unresponsive. We have shown in the previous section you can get incredible speeds out of the file system if you understand the limitations of your hardware and how to properly setup your raid. We suspect that many of the objectors of ZFS have setup their ZFS system using slow or otherwise substandard I/O subsystems.
Here we test the SATA throughput of the same physical Western Digital Black 2TB SATA6 (WD2002FAEX) spinning hard drive and Samsung 840 PRO 256GB SSD on three(3) different interfaces using FreeBSD. We look at the SATA 6 Gbit/s port of a common gaming motherboard by Asus, a Supermicro server motherboard and an LSI MegaRAID raid card. All machines have at least a six(6) core CPU and 16 gig of ram. According to Western Digital the Black series 2TB (WD2002FAEX) should be able to theoretically sustain 150MB/sec sequential read and writes. The Samsung 840 Pro is rated at 540MB/sec reads and 520MB/sec writes. Notice the throughput between motherboard based SATA ports (sata 3 versus sata 6) and a dedicated card.
1x 2TB a single drive - 1.8 terabytes - Western Digital Black 2TB (WD2002FAEX) Asus Sabertooth 990FX sata6 onboard ( w= 39MB/s , rw= 25MB/s , r= 91MB/s ) SuperMicro X9SRE sata3 onboard ( w= 31MB/s , rw= 22MB/s , r= 89MB/s ) LSI MegaRAID 9265-8i sata6 "JBOD" ( w=130MB/s , rw= 66MB/s , r=150MB/s ) 1x 256GB a single drive - 232 gigabytes - Samsung 840 PRO 256GB (MZ-7PD256BW) Asus Sabertooth 990FX sata6 onboard ( w=242MB/s , rw=158MB/s , r=533MB/s ) LSI MegaRAID 9265-8i sata6 "JBOD" ( w=438MB/s , rw=233MB/s , r=514MB/s ) # The raw bonnie output on FreeBSD 9.1 1x 2TB a single drive 1.8 terabytes (Asus Sabertooth 990FX on board SATA6) Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsASUS 80G 39139 11 25902 8 91764 8 122.1 2 Latency 4015ms 3250ms 2263ms 443ms 1x 2TB a single drive 1.8 terabytes (Supermicro X9SRE on board SATA6) Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP zfsSuper 80G 31342 6 22420 4 89085 4 180.6 1 Latency 4709ms 4699ms 1263ms 568ms 1x 2TB a single drive 1.8 terabytes (LSI MegaRAID 9265-8i JBOD SATA6) Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 130325 23 66041 11 150329 8 144.8 6 Latency 12251ms 5102ms 1381ms 611ms 1x 256GB a single drive - 232 gigabytes (Asus Sabertooth 990FX on board SATA6) Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 224826 42 158018 30 533094 38 3788 49 Latency 3736ms 1422ms 652ms 26953us 1x 256GB a single drive - 232 gigabytes (LSI MegaRAID 9265-8i JBOD SATA6) Version 1.96 ------Sequential Output------ --Sequential Input- --Random- Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks-- Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP moneyslow.com 80G 438424 70 233976 41 514946 29 3807 57 Latency 70402us 1714ms 770ms 28608us
By default, older hard drives have a sector size of 512 bytes. Sectors sizes changed when drives became large enough that the overhead of keeping track of sectors consumed a high percentage of storage space. Many modern drives are now specified as advanced format drives which means they have 4096 byte sector sizes (4KiB). Modern 4K drives includes all SSDs and most 2TB and above magnetic drives. You can check on the manufacture website for your drives "sector size" or "sector alignment".
ZFS will query the underlying device to understand how large the sectors are and use this information to determine the size of its dynamic width stripes. A fine idea unless the hardware lies. Today, drive hardware lies more often than not. Drives claim to have a logical sector size of 512 bytes (ashift=9 or 2^9=512) while the physical sectors are 4Kib (ashift=12 or 2^12=4096). Since the drive lied, ZFS will incorrectly make stripes aligned to 512 bytes. This means stripes will almost always be non-aligned forcing the underlying device to use internal processing time which will slightly degrade write performance. We say slightly because in practice a misaligned drive only writes less then 10% slower then a properly aligned drive. Still, every byte per second counts right?
So, we are forced to manually tell ZFS exactly how we want the sectors aligned since we can not trust the hardware to tell us the truth. The procedure is not hard, but it is a little abstract. Lets setup a single drive in the ZFS pool called "tank." Using gnop we can create .nop devices to force ZFS to align physical sectors to 4K no matter what the hard drives tell us.
# Setting up a single drive ZFS pool called "tank" which is 4K aligned
# If you already have an unaligned pool you need to
# destroy it. Backup your data if need to.
zpool destroy tank
# A 4K gnop device needs to be created which will be chained to the ZFS pool
# This will create the "tank" pool with *.nop devices forcing ZFS to have 4K
# physical sectors.
gnop create -S 4096 /dev/mfid0
zpool create tank /dev/mfid0.nop
# Next, the pool needs to be exported and the temporary gnop
# devices removed
zpool export tank
gnop destroy /dev/mfid0.nop
# Now, the aligned pool can be imported from the "raw" devices
zpool import tank
# All done. Query ZFS to make sure the ashift equals 12 which means this pool
# is 4K aligned. If you see "ashift: 9" then you are 512b aligned.
zdb -C tank | grep ashift
ashift: 12
So, what performance difference will I see for a properly aligned 4K sector sized ZFS pool? Truthfully, we did not see much difference with one or two spinning disks. Normally less then a 10% speed variation. With alignment you may see a slight increase in hard drive speeds and your setup will be more efficient because the hard drive itself will not have to do any sector splitting or read-modify-write operations.
The performance difference seems to also depend on the type of drives you use. Green or ECO drives perform poorly even with 4K alignment compared to RE (Raid Enabled) drives. To make things more complicated some drives are simply slower then others even from the same manufacturer or family of drives. The only way to make sure you are getting the most out of your equipment is to format the drive as 512b and 4K and do the bonnie++ test on each. If you have many drives, test each one individually and put the "fast" ones in one raid and the "slower" ones in another. Mixing fast and slow drives will make the entire setup slow. It is a pain, we know, but the individual test method works reliably.
For a a new boot drive or raid we still suggest aligning the sectors. If you have a unaligned drive and want to wipe the data and re-align, keep these numbers in mind, test your own equipment and make an informed decision.
Western Digital Black 2TB (WD2001FYYG)
512b 1x 2TB a single drive 1.8 terabytes ( w=130MB/s , rw= 65MB/s , r=146MB/s )
4K 1x 2TB a single drive 1.8 terabytes ( w=131MB/s , rw= 66MB/s , r=150MB/s )
Samsung 840 PRO 256GB (MZ-7PD256BW)
512b 1x 256GB a single drive 232 gigabytes ( w=446MB/s , rw=232MB/s , r=520MB/s )
4K 1x 256GB a single drive 232 gigabytes ( w=438MB/s , rw=233MB/s , r=514MB/s )
##### 512 byte UN-aligned sectors Western Digital Black 2TB
#
# zdb -C tank | grep ashift
ashift: 9
Version 1.96 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
moneyslow.com 80G 130463 23 65792 11 146863 8 120.6 1
Latency 11645ms 4845ms 2505ms 353ms
##### 4K aligned sectors Western Digital Black 2TB
#
# zdb -C tank | grep ashift
ashift: 12
Version 1.96 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
moneyslow.com 80G 131335 23 66141 11 150777 9 148.2 5
Latency 13651ms 5009ms 1189ms 625ms
##### 512 byte UN-aligned sectors Samsung 840 PRO 256GB
#
# zdb -C tank | grep ashift
ashift: 9
Version 1.96 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
moneyslow.com 80G 446299 71 232491 40 520952 30 1977 68
Latency 96173us 1963ms 104ms 25309us
##### 4K aligned sectors Samsung 840 PRO 256GB
#
# zdb -C tank | grep ashift
ashift: 12
Version 1.96 ------Sequential Output------ --Sequential Input- --Random-
Concurrency 1 -Per Chr- --Block-- -Rewrite- -Per Chr- --Block-- --Seeks--
Machine Size K/sec %CP K/sec %CP K/sec %CP K/sec %CP K/sec %CP /sec %CP
moneyslow.com 80G 438424 70 233976 41 514946 29 3807 57
Latency 70402us 1714ms 770ms 28608us
Do you have a ZFS Health Script we can use ?
Take a look at our ZFS Health Check and Status script. The script checks disk and volume errors as well as the health of the pool and even when the last zpool scrub was done.
Why is ZFS import not using GPT labels ?
When importing a volume you must point ZFS to the directory with the GPT labels using the "-d" argument. For example our GPT labels start with data1-XX and under Ubuntu the device labels are in /dev/disk/by-partlabel/ .
# import pool with GPT labels on Ubuntu for example.
root@zfsRAID:~# zpool import -d /dev/disk/by-partlabel/ data1
root@zfsRAID:~# zpool status
pool: data1
state: ONLINE
scan: scrub repaired 0 in 0h0m with 0 errors on Mon Apr 8 14:36:17 2033
config:
NAME STATE READ WRITE CKSUM
data1 ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
data1-00 ONLINE 0 0 0
data1-01 ONLINE 0 0 0
data1-02 ONLINE 0 0 0
data1-03 ONLINE 0 0 0
data1-04 ONLINE 0 0 0
data1-05 ONLINE 0 0 0
data1-06 ONLINE 0 0 0
data1-07 ONLINE 0 0 0
data1-08 ONLINE 0 0 0
data1-09 ONLINE 0 0 0
data1-10 ONLINE 0 0 0
data1-11 ONLINE 0 0 0
# If you do a straight zpool import ZFS will mount the drives with the
# raw device names which we do not want.
root@zfsRAID:~# zpool import data1
root@zfsRAID:~# zpool status
pool: data1
state: ONLINE
scan: resilvered 0 in 0h0m with 0 errors on Mon Apr 8 14:33:58 2033
config:
NAME STATE READ WRITE CKSUM
data1 ONLINE 0 0 0
raidz2-0 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe327b8d44f-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe427c0c540-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe427c9410f-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe527d1dc47-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5c8456d9cb386-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe627e2e502-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe627eb60b9-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe727fa267a-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe82802c19f-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abe92811a6f3-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abea28205d1e-part1 ONLINE 0 0 0
scsi-3600605b00512e8c018f5abeb282e7776-part1 ONLINE 0 0 0