Unsolved
This post is more than 5 years old
16 Posts
0
1304
ScaleIO low performance
We using 5 SDS-SDC nodes with 3 disks (480GB SSD).
Configuration:
1 pool with 1 primary domain: 2 datastore - first thin datastore and second thick datastore.
IO priority: rebuild and rebalance - no limit
RAM Read Cache - disabled
Write Handling mode - cached
Use Checksum - disabled
Zero Padding policy - disabled
Performance Profile - high
ScaleIO Benchmark:
| Test | Read (MB/s) | IOPS | Write (MB/s) | IOPS |
| Seq. Q32T1 | 897.1 | 604 | ||
| Random 4KQ32T1 | 132 | 32k | 75.92 | 18k |
| Seq. | 174.7 | 105.9 | ||
| Random 4K | 5.967 | 1456 | 4.010 | 979 |
Single disk Benchmark:
| Test | Read (MB/s) | Write (MB/s) |
| Seq. Q32T1 | 409 | 351 |
| Random 4KQ32T1 | 256 | 178 |
| Seq. | 281 | 348 |
| Random 4K | 23 | 64 |
How to imporve performance for ScaleIO Cluster?
pawelw1
306 Posts
0
July 24th, 2017 05:00
Hi,
Have you configured your system according to the Fine Tuning guide, especially the SSD parts?
Also, when you run an SDS network test, are the results expected for your environment?
Cheers,
Pawel
vearcut
16 Posts
0
July 24th, 2017 22:00
The presentation of ScaleIO says, when using 20x servers with 1 HDD (100 iops and 1TB storage) we get 20 TB raw storage (9.5 TB usable) and 2000 IOPS.
In my case it is - 1 ssd - 50k IOPS and 400 MB/s on read operations. Cluster - 4x nodes with 3x ssd (150k iops per server or 600k IOPS per cluster)
Where is the error in the calculations?