git.net

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Urgent help needed : Cassandra benchmarking with Cassandra Stress tool


Hi all

 

Hope you all doing good

I am working on benchmarking Cassandra using Cassandra stress tool and results are not very impressive. I really some help to tune Cassandra to get great results. Below is my environment and test results and attached is Cassandra.yaml I am using for the test.

 

I am running single node Cassandra virtual machine and 1 SSD disk in VMWare ESXi 6.7 eventually I want to scale the database to 4 fours once I get the expected results in the single node. Attached is the Cassandra.yaml file is used for configuring database.

 

The result I am getting is not very impressive operations/sec is very low and the latency is very high for the reads here is the results it got from the test.

Load command: Cassandra-stress write n=4000000000 -rate threads=96 -node $node1 -log file=load_400M.log

 

Results:

Op rate                   :    2,347 op/s  [READ: 2,347 op/s]

Partition rate            :    2,347 pk/s  [READ: 2,347 pk/s]

Row rate                  :    2,347 row/s [READ: 2,347 row/s]

Latency mean              :  381.1 ms [READ: 381.1 ms]

Latency median            :  327.4 ms [READ: 327.4 ms]

Latency 95th percentile   :  836.2 ms [READ: 836.2 ms]

Latency 99th percentile   : 1138.8 ms [READ: 1,138.8 ms]

Latency 99.9th percentile : 1695.5 ms [READ: 1,695.5 ms]

Latency max               : 6253.7 ms [READ: 6,253.7 ms]

Total partitions          :  1,000,000 [READ: 1,000,000]

Total errors              :          0 [READ: 0]

Total GC count            : 0

Total GC memory           : 0.000 KiB

Total GC time             :    0.0 seconds

Avg GC time               :    NaN ms

StdDev GC time            :    0.0 ms

Total operation time      : 00:07:06

 

The write performance is really impressive with less latency and good operation/sec and when we compare it with reads I feel something is wrong.

Read command: Cassandra-stress read n=100000000  no-warmup cl=one -mode native cql3 -schema keyspace=”keyspace1” -rate threads=96 -node $node1 -log file=read_10M_test.log

 

Results:

Op rate                   :   35,979 op/s  [WRITE: 35,979 op/s]

Partition rate            :   35,979 pk/s  [WRITE: 35,979 pk/s]

Row rate                  :   35,979 row/s [WRITE: 35,979 row/s]

Latency mean              :    3.3 ms [WRITE: 3.3 ms]

Latency median            :    2.2 ms [WRITE: 2.2 ms]

Latency 95th percentile   :    8.2 ms [WRITE: 8.2 ms]

Latency 99th percentile   :   17.6 ms [WRITE: 17.6 ms]

Latency 99.9th percentile :   75.6 ms [WRITE: 75.6 ms]

Latency max               : 1840.3 ms [WRITE: 1,840.3 ms]

Total partitions          : 1,600,000,000 [WRITE: 1,600,000,000]

Total errors              :          0 [WRITE: 0]

Total GC count            : 0

Total GC memory           : 0.000 KiB

Total GC time             :    0.0 seconds

Avg GC time               :    NaN ms

StdDev GC time            :    0.0 ms

Total operation time      : 12:21:10

 

 

When I compare the both read and write performance they are not at all consistent and I feel there is some kind of bottleneck with the reads during test.

 

Could you please look into these details and let me know where I am missing I have been struggling hard to get good results out of Cassandra database but I couldn’t figure out where I am missing.

 

Please let me know if you need additional information about the Cassandra setup.

 

 

Thanks and Regards

Vinay Kumar Vongour

 

 

 

Thanks and Regards

Vinay Kumar Vongour

 

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@xxxxxxxxxxxxxxxxxxxx
For additional commands, e-mail: dev-help@xxxxxxxxxxxxxxxxxxxx