Jun Wu
2015-09-23 00:37:10 UTC
Dear All,
I was doing some test on datastax java driver with Eclipse on Mac Pro. Basically, in
the eclipse, I was using prepared and batch statement to write (id, name), where id
ranges from 10000 to 1000000 and names are student1 - student10000(according to the range of id).
After every time I run the code in eclipse, everything goes well, then I use nodetool cfhistograms and proxyhistograms, and cfstats to check the results. However, it seems to be weird. As I was looking forward to see the read latency and write lantency to increase, as I add the range from 10000, 100000, 1000000, 10000000, the final result turns out to be decreasing on the latency.
For example, for the cfhistograms of 10,000 first table and 100,000 second table:
Percentile SSTables Write Latency
50% 0 14
75% 0 20
95% 0 42
98% 0 60
99% 0 60
Min 0 5
Max 0 179
and 100,000 below
Percentile SSTables Write Latency
50% 0 10
75% 0 12
95% 0 24
98% 0 29
99% 0 42
Min 0 3
Max 0 1109
Also, for the proxyhistogram, 10, 000:
Percentile Read Latency Write Latency
50% 1331 72
75% 1916 103
95% 5722 215
98% 5722 310
99% 5722 372
Min 259 30
Max 5722 11864
and 100,000 below:
Percentile Read Latency Write Latency
50% 1331 42
75% 1916 50
95% 6866 103
98% 6866 149
99% 6866 215
Min 311 18
Max 6866 61214
Anyone has any idea about this?
Thanks in advance!
Steven
I was doing some test on datastax java driver with Eclipse on Mac Pro. Basically, in
the eclipse, I was using prepared and batch statement to write (id, name), where id
ranges from 10000 to 1000000 and names are student1 - student10000(according to the range of id).
After every time I run the code in eclipse, everything goes well, then I use nodetool cfhistograms and proxyhistograms, and cfstats to check the results. However, it seems to be weird. As I was looking forward to see the read latency and write lantency to increase, as I add the range from 10000, 100000, 1000000, 10000000, the final result turns out to be decreasing on the latency.
For example, for the cfhistograms of 10,000 first table and 100,000 second table:
Percentile SSTables Write Latency
50% 0 14
75% 0 20
95% 0 42
98% 0 60
99% 0 60
Min 0 5
Max 0 179
and 100,000 below
Percentile SSTables Write Latency
50% 0 10
75% 0 12
95% 0 24
98% 0 29
99% 0 42
Min 0 3
Max 0 1109
Also, for the proxyhistogram, 10, 000:
Percentile Read Latency Write Latency
50% 1331 72
75% 1916 103
95% 5722 215
98% 5722 310
99% 5722 372
Min 259 30
Max 5722 11864
and 100,000 below:
Percentile Read Latency Write Latency
50% 1331 42
75% 1916 50
95% 6866 103
98% 6866 149
99% 6866 215
Min 311 18
Max 6866 61214
Anyone has any idea about this?
Thanks in advance!
Steven