Reviewing the Synology DS1817+ with M2D17

Benchmark - Storage Volume with SSD Cache

The statistics of the SSD cache.

We will also be using Fio to benchmark the performance of the DS1817+ under the following scenarios:

  • SSD cache is not enabled;
  • SSD read cache is enabled; and
  • SSD read-write cache is enabled.

The SSD cache works by moving frequently accessed data onto the faster storage (i.e. SSD) so that the system need not wait for the slower storage (i.e regular disks) to respond.

Hence, we will go through a phrase of preconditioning before the actual benchmark test:

  • Preconditioning. 5 sets of writing and reading from pre-prepared files to allow data to be cached;
  • Benchmark. 1 set of writing and reading from the same pre-prepared files.

For each sets, the IO workload will be 70% reads and 30% writes in 8KB blocksize for 60 seconds. The order of reads and writes are randomized.

 

SSD Cache Benchmark Summary

The benchmark results of the DS1817+ under various SSD cache configurations are as follow.

Read (MB/s)Write (MB/s)
Without SSD Cache 4.4 1.88
SSD Read Cache – WD Blue 250GB M.2 8.75 3.79
SSD Read Cache – Samsung 850 EVO 250GB M.2 9.27 3.96
SSD Read-write Cache 16.95 7.24

Avg Read Latency (ms) Max Read Latency (ms) Avg Write Latency (ms) Max Write Latency (ms)
Without SSD Cache 201.51 1,299 189.17 1,295
SSD Read Cache – WD Blue 250GB M.2 34.92 750.63 248.44 1,529
SSD Read Cache – Samsung 850 EVO 250GB M.2 32.37 427.68 239.18 831
SSD Read-write Cache 41.81 2,142.9 74.57 4,806.8

Firstly, don’t be alarmed by the low throughput figures. It appears that fio’s workload and measurement reflected a lower throughput for tests on the storage volume. Our objective is to measure the difference in performance under different SSD cache configurations.

Across the tests, it is clear that the both reading and writing speeds are greatly improved when SSD cache is enabled.

Findings when SSD read cache is enabled

When SSD read cache is enabled, the average latency for read operations drastically dropped. However, the average latency for write operations increased slightly.

We also observed that the performance is better when the Samsung 850 EVO is used as the cache.

Findings when SSD read-write cache is enabled

When SSD read-write cache is enabled, the average latency for writing operation is reduced.

Read 50th Percentile (ms) Read Std Dev (ms) Write 50th Percentile (ms) Write Std Dev (ms)
Without SSD Cache 165.00 109.34 167.00 110.68
SSD Read Cache – WD Blue 250GB M.2156.8046.56221.00146.68
SSD Read Cache – Samsung 850 EVO 250GB M.2 144.00 38.46255.00116.56
SSD Read-write Cache 26.24 43.93 35.58 173.93

We also observed the maximum latency went off the chart during the tests. But when we delved into the statistics, we observed that the maximum readings are just some statistical outliers. The 50th percentile and standard deviation looks pretty ok.

4 thoughts on “Reviewing the Synology DS1817+ with M2D17

  • July 28, 2017 at 11:16 am
    Permalink

    drool ~~~~

  • August 31, 2017 at 4:23 am
    Permalink

    i’ve been looking at upgraded NAS offerings. this seems like a great choice, my workflows are mostly IOPS limited, so this would represent a significant increase in performance and a significant decrease in keyboard smashing stress.

  • January 10, 2018 at 10:50 am
    Permalink

    Good review. Do you know if it possible to use an M.2 SSD installed in the M2D17 card as an extra storage device? Such as using it to run all your app’s etc (docker, virtual machines…) instead of running them off the array. I currently use slot 8 with an SSD installed for this purpose

  • January 10, 2018 at 12:32 pm
    Permalink

    Hi Darren,

    Unfortunately, we can’t use the SSD on the M2D17 as storage volume. It will just appear as SSD cache.

Comments are closed.