On 29/04/2021 11:52 pm, Schmid, Michael wrote:
I am new to ceph and at the moment I am doing some performance tests with a 4 node ceph-cluster (pacific, 16.2.1).
Ceph doesn't do well with small numbers, 4 OSD's is really marginal. Your latency isn't crash hot either. What size are you running on the pool? The amount of RAM per node (8GB) would be the bare minimum as well, your ceph setup is really constrained. Do your OSD's have access to the raw device? are they bluestore? Same test on my Cluster * 5 Node, * 20 OSD's (Total) o Mix of SATA and SAS Spinners o WAL/DB on SSD * 64GB RAM Per node * 4 * 1GB Bond rados bench -p ceph 10 write -b 4M -t 16 --no-cleanup hints = 1 Maintaining 16 concurrent writes of 4194304 bytes to objects of size 4194304 for up to 10 seconds or 0 objects Object prefix: benchmark_data_vnh_3642327 sec Cur ops started finished avg MB/s cur MB/s last lat(s) avg lat(s) 0 0 0 0 0 0 - 0 1 16 58 42 167.99 168 0.21848 0.329228 2 16 102 86 171.986 176 0.456715 0.325869 3 16 154 138 183.983 208 0.109888 0.319586 4 16 206 190 189.981 208 0.188891 0.320275 5 16 258 242 193.581 208 0.261014 0.319318 6 16 308 292 194.647 200 0.450672 0.319268 7 16 358 342 195.408 200 0.127415 0.316999 8 16 406 390 194.98 192 0.176382 0.321384 9 16 456 440 195.535 200 0.287347 0.318749 10 16 508 492 196.779 208 0.279796 0.318067 Total time run: 10.2741 Total writes made: 508 Write size: 4194304 Object size: 4194304 Bandwidth (MB/sec): 197.78 Stddev Bandwidth: 14.2111 Max bandwidth (MB/sec): 208 Min bandwidth (MB/sec): 168 Average IOPS: 49 Stddev IOPS: 3.55278 Max IOPS: 52 Min IOPS: 42 Average Latency(s): 0.318968 Stddev Latency(s): 0.137534 Max latency(s): 0.913779 Min latency(s): 0.0933294 -- Lindsay