URL: https://www.opennet.me/cgi-bin/openforum/vsluhboard.cgi
Форум: vsluhforumID1
Нить номер: 94948
[ Назад ]
Исходное сообщение
"NFS, GlusterFs, Ceph и другие, что выбрать и ...."
Отправлено Alexey , 26-Авг-13 08:44 
Вообщем .. проблемы с nfs меня задрали уже... поэтому решили поднять на тестовом стенде, что либо из распределеной системы хранения данных.

Первое что начал тестировать это был glusterfs

Взлетает на Дебиане все в 2-5 мин =))

перед тем как старотовать volume сделал его сихронизацию... чтобы на обjих peerа была одинаковая инфа
стартанул volume
смонтировал на клиентской машине

делаю чтение запись мелких файлов тут наступает fail

делаю скрипт
#!/bin/bash

for i in {1..1000}; do
#size=$((RANDOM%2+1))
echo "Write"
dd if=/dev/zero of=/storage/test/bigfile${i} count=10 bs=2k
echo "read"
dd if=/storage/test/bigfile${i} of=/dev/zero count=10 bs=2k
done

и вот что получаю
Write
2+0 records in
2+0 records out
4096 bytes (4.1 kB) copied, 0.00204757 s, 2.0 MB/s
read
2+0 records in
2+0 records out
4096 bytes (4.1 kB) copied, 0.00157458 s, 2.6 MB/s
Write
4+0 records in
4+0 records out
8192 bytes (8.2 kB) copied, 0.00197661 s, 4.1 MB/s
read
4+0 records in
4+0 records out
8192 bytes (8.2 kB) copied, 0.00169478 s, 4.8 MB/s

при монтировании
nas.storage:/storage /storage glusterfs defaults,_netdev 0 0

mount /storage

делаем по другому
mount -t nfs -o rw,rsize=8192,wsize=8192,proto=tcp,soft,intr,noatime,noauto,noacl,async,nodirati
me 192.168.15.165:/storage /storage

скорость чтения большая, а вот скорость записи ужас какой-то =(

Write
2+0 records in
2+0 records out
4096 bytes (4.1 kB) copied, 0.136233 s, 30.1 kB/s
read
2+0 records in
2+0 records out
4096 bytes (4.1 kB) copied, 1.3131e-05 s, 312 MB/s
Write
3+0 records in
3+0 records out
6144 bytes (6.1 kB) copied, 0.143521 s, 42.8 kB/s
read
3+0 records in
3+0 records out
6144 bytes (6.1 kB) copied, 1.3899e-05 s, 442 MB/s


на очереди ceph .. сейчас жду когда подключат еще сервак.. и буду тестить ceph

у кого был какой опыт установки и использования таких систем ?

here the result of iozone

Record Size 4 KB
File size set to 4 KB
Command line used: ./iozone -l 2 -u 2 -r 4k -s 4k /storage/
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
Min process = 2
Max process = 2
Throughput test with 2 processes
Each process writes a 4 Kbyte file in 4 Kbyte records

Children see throughput for 2 initial writers = 209712.23 KB/sec
Parent sees throughput for 2 initial writers = 98.28 KB/sec
Min throughput per process = 95430.45 KB/sec
Max throughput per process = 114281.78 KB/sec
Avg throughput per process = 104856.11 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 rewriters = 206101.09 KB/sec
Parent sees throughput for 2 rewriters = 109.33 KB/sec
Min throughput per process = 103050.55 KB/sec
Max throughput per process = 103050.55 KB/sec
Avg throughput per process = 103050.55 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 readers = 197831.25 KB/sec
Parent sees throughput for 2 readers = 93.25 KB/sec
Min throughput per process = 197831.25 KB/sec
Max throughput per process = 197831.25 KB/sec
Avg throughput per process = 98915.62 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 re-readers = 366269.22 KB/sec
Parent sees throughput for 2 re-readers = 89.50 KB/sec
Min throughput per process = 366269.22 KB/sec
Max throughput per process = 366269.22 KB/sec
Avg throughput per process = 183134.61 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 reverse readers = 102421.45 KB/sec
Parent sees throughput for 2 reverse readers = 151.60 KB/sec
Min throughput per process = 102421.45 KB/sec
Max throughput per process = 102421.45 KB/sec
Avg throughput per process = 51210.72 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 stride readers = 494861.56 KB/sec
Parent sees throughput for 2 stride readers = 351.51 KB/sec
Min throughput per process = 247430.78 KB/sec
Max throughput per process = 247430.78 KB/sec
Avg throughput per process = 247430.78 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 random readers = 191072.08 KB/sec
Parent sees throughput for 2 random readers = 160.73 KB/sec
Min throughput per process = 191072.08 KB/sec
Max throughput per process = 191072.08 KB/sec
Avg throughput per process = 95536.04 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 mixed workload = 105646.19 KB/sec
Parent sees throughput for 2 mixed workload = 58.51 KB/sec
Min throughput per process = 105646.19 KB/sec
Max throughput per process = 105646.19 KB/sec
Avg throughput per process = 52823.09 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 random writers = 117482.82 KB/sec
Parent sees throughput for 2 random writers = 51.33 KB/sec
Min throughput per process = 117482.82 KB/sec
Max throughput per process = 117482.82 KB/sec
Avg throughput per process = 58741.41 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 pwrite writers = 166431.23 KB/sec
Parent sees throughput for 2 pwrite writers = 51.03 KB/sec
Min throughput per process = 0.00 KB/sec
Max throughput per process = 166431.23 KB/sec
Avg throughput per process = 83215.62 KB/sec
Min xfer = 0.00 KB

Children see throughput for 2 pread readers = 200191.84 KB/sec
Parent sees throughput for 2 pread readers = 75.51 KB/sec
Min throughput per process = 200191.84 KB/sec
Max throughput per process = 200191.84 KB/sec
Avg throughput per process = 100095.92 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 fwriters = 280713.02 KB/sec
Parent sees throughput for 2 fwriters = 150.10 KB/sec
Min throughput per process = 114281.78 KB/sec
Max throughput per process = 166431.23 KB/sec
Avg throughput per process = 140356.51 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 freaders = 351151.31 KB/sec
Parent sees throughput for 2 freaders = 4291.50 KB/sec
Min throughput per process = 160079.23 KB/sec
Max throughput per process = 191072.08 KB/sec
Avg throughput per process = 175575.66 KB/sec
Min xfer = 4.00 KB

iozone test complete.

/storage mounted like
nas.storage:/storage /storage glusterfs defaults,_netdev 0 0


now
mount -t nfs -o rw,rsize=8192,wsize=8192,proto=tcp,soft,intr,noatime,noauto,noacl,async,nodirati
me 192.168.15.165:/storage /storage

nfs server served by glusterd

Record Size 4 KB
File size set to 4 KB
Command line used: ./iozone -l 2 -u 2 -r 4k -s 4k /storage/
Output is in Kbytes/sec
Time Resolution = 0.000001 seconds.
Processor cache size set to 1024 Kbytes.
Processor cache line size set to 32 bytes.
File stride size set to 17 * record size.
Min process = 2
Max process = 2
Throughput test with 2 processes
Each process writes a 4 Kbyte file in 4 Kbyte records

Children see throughput for 2 initial writers = 188920.50 KB/sec
Parent sees throughput for 2 initial writers = 56.48 KB/sec
Min throughput per process = 188920.50 KB/sec
Max throughput per process = 188920.50 KB/sec
Avg throughput per process = 94460.25 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 rewriters = 160079.23 KB/sec
Parent sees throughput for 2 rewriters = 54.93 KB/sec
Min throughput per process = 160079.23 KB/sec
Max throughput per process = 160079.23 KB/sec
Avg throughput per process = 80039.62 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 readers = 210225.80 KB/sec
Parent sees throughput for 2 readers = 72.11 KB/sec
Min throughput per process = 210225.80 KB/sec
Max throughput per process = 210225.80 KB/sec
Avg throughput per process = 105112.90 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 re-readers = 210225.80 KB/sec
Parent sees throughput for 2 re-readers = 281.29 KB/sec
Min throughput per process = 210225.80 KB/sec
Max throughput per process = 210225.80 KB/sec
Avg throughput per process = 105112.90 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 reverse readers = 143633.55 KB/sec
Parent sees throughput for 2 reverse readers = 194.59 KB/sec
Min throughput per process = 143633.55 KB/sec
Max throughput per process = 143633.55 KB/sec
Avg throughput per process = 71816.77 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 stride readers = 267128.91 KB/sec
Parent sees throughput for 2 stride readers = 196.70 KB/sec
Min throughput per process = 267128.91 KB/sec
Max throughput per process = 267128.91 KB/sec
Avg throughput per process = 133564.45 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 random readers = 200191.84 KB/sec
Parent sees throughput for 2 random readers = 159.47 KB/sec
Min throughput per process = 0.00 KB/sec
Max throughput per process = 200191.84 KB/sec
Avg throughput per process = 100095.92 KB/sec
Min xfer = 0.00 KB

Children see throughput for 2 mixed workload = 125384.91 KB/sec
Parent sees throughput for 2 mixed workload = 62.83 KB/sec
Min throughput per process = 125384.91 KB/sec
Max throughput per process = 125384.91 KB/sec
Avg throughput per process = 62692.46 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 random writers = 210225.80 KB/sec
Parent sees throughput for 2 random writers = 51.10 KB/sec
Min throughput per process = 210225.80 KB/sec
Max throughput per process = 210225.80 KB/sec
Avg throughput per process = 105112.90 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 pwrite writers = 166431.23 KB/sec
Parent sees throughput for 2 pwrite writers = 51.74 KB/sec
Min throughput per process = 0.00 KB/sec
Max throughput per process = 166431.23 KB/sec
Avg throughput per process = 83215.62 KB/sec
Min xfer = 0.00 KB

Children see throughput for 2 pread readers = 295679.92 KB/sec
Parent sees throughput for 2 pread readers = 148.73 KB/sec
Min throughput per process = 129248.69 KB/sec
Max throughput per process = 166431.23 KB/sec
Avg throughput per process = 147839.96 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 fwriters = 280151.83 KB/sec
Parent sees throughput for 2 fwriters = 144.09 KB/sec
Min throughput per process = 137737.53 KB/sec
Max throughput per process = 142414.30 KB/sec
Avg throughput per process = 140075.91 KB/sec
Min xfer = 4.00 KB

Children see throughput for 2 freaders = 239228.27 KB/sec
Parent sees throughput for 2 freaders = 6728.64 KB/sec
Min throughput per process = 117482.82 KB/sec
Max throughput per process = 121745.45 KB/sec
Avg throughput per process = 119614.13 KB/sec
Min xfer = 4.00 KB

iozone test complete.

Содержание
Сообщения в этом обсуждении
"NFS, GlusterFs, Ceph и другие, что выбрать и ...."
Отправлено Skif , 01-Сен-13 16:03 
Да никуда не денетесь вы от глюков даже на кластерных фс. Особенно с кучами мелких файлов. Будут задержки и провалы. Если всё так критично - купите нормальную железку.
Чтение ещё фигня. Самый цинус у вас будет при записи. И не на тесте,а в продакшене. Когда оно будет "булькать" данными.
Этим болеют все "коленчатые" решения, что глистер, что дрдб.
"NFS, GlusterFs, Ceph и другие, что выбрать и ...."
Отправлено Alexey , 04-Сен-13 15:53 
Да мы юзаем Дисковую полку с дисками SATA 10k 500Gb
к серверу с nfs подключен через карту Pecl кажется