Cassandra节点中的高负载平均值

Question

我们正在观察4节点Cassandra集群的高CPU平均负载。 RF主要是2，我们使用LOCAL_ONE来实现读写一致性。

sjk ttop输出主要显示了sharedpool-worker线程。

    2016-06-24T11:59:08.894+0000 Process summary
      process cpu=704.99%
      application cpu=712.74% (user=678.15% sys=34.60%)
      other: cpu=-7.75%
      heap allocation rate 174mb/s
    [387229] user=32.01% sys= 1.49% alloc=   58mb/s - CompactionExecutor:12419
    [000314] user=17.52% sys=-0.17% alloc= 3136kb/s - SharedPool-Worker-3
    [000317] user=17.42% sys=-0.19% alloc= 3897kb/s - SharedPool-Worker-5
    [000315] user=15.90% sys= 0.52% alloc= 1546kb/s - SharedPool-Worker-4
    [000313] user=15.80% sys= 0.13% alloc= 2519kb/s - SharedPool-Worker-2
    [000312] user=15.70% sys= 0.09% alloc= 1826kb/s - SharedPool-Worker-1
    [000316] user=13.57% sys= 0.74% alloc= 2568kb/s - SharedPool-Worker-6
    [000326] user=13.57% sys= 0.08% alloc=  832kb/s - SharedPool-Worker-13
    [000320] user=13.57% sys= 0.03% alloc= 3024kb/s - SharedPool-Worker-7
    [000325] user=13.07% sys= 0.23% alloc= 1781kb/s - SharedPool-Worker-12
    [000318] user=12.46% sys= 0.74% alloc= 2174kb/s - SharedPool-Worker-8
    [000327] user=11.65% sys= 0.46% alloc= 2371kb/s - SharedPool-Worker-11
    [000328] user=11.45% sys= 0.21% alloc= 1233kb/s - SharedPool-Worker-24
    [000319] user=11.65% sys= 0.00% alloc= 1861kb/s - SharedPool-Worker-10
    [000321] user=11.34% sys= 0.14% alloc= 1037kb/s - SharedPool-Worker-14
    [000322] user=11.45% sys=-0.05% alloc= 2120kb/s - SharedPool-Worker-9
    [000324] user=10.84% sys= 0.12% alloc= 1860kb/s - SharedPool-Worker-15
    [000332] user=10.74% sys= 0.09% alloc= 1219kb/s - SharedPool-Worker-22
    [000335] user=10.43% sys= 0.19% alloc=  956kb/s - SharedPool-Worker-19
    [000323] user=10.23% sys= 0.19% alloc=  688kb/s - SharedPool-Worker-16
    [000333] user=10.13% sys= 0.03% alloc= 1450kb/s - SharedPool-Worker-21
    [000330] user= 9.42% sys= 0.25% alloc= 1507kb/s - SharedPool-Worker-26
    [000337] user= 9.32% sys= 0.29% alloc=  780kb/s - SharedPool-Worker-17
    [000334] user= 9.32% sys= 0.06% alloc= 1698kb/s - SharedPool-Worker-20
    [000339] user= 9.42% sys=-0.06% alloc= 1514kb/s - SharedPool-Worker-27
    [000345] user= 8.61% sys= 0.15% alloc=  510kb/s - SharedPool-Worker-34
    [000329] user= 8.41% sys= 0.34% alloc=  760kb/s - SharedPool-Worker-23
    [000598] user= 8.31% sys= 0.06% alloc= 1124kb/s - SharedPool-Worker-39
    [000336] user= 8.20% sys= 0.12% alloc= 1895kb/s - SharedPool-Worker-18
    [000590] user= 8.00% sys= 0.22% alloc=  953kb/s - SharedPool-Worker-35

群集中的读写请求是：

我正在寻找如何进一步调试高负载的建议。我想在决定扩展集群之前消除任何调整可能性。