一、RAC的AWR纵览
二、GCS等待事件
三、Global Cache Load Profile
1.可以大概估算下所用带宽(没有算GCS/GES Message的大小)
(received+send)*db_block_size= 901 * 8k= 7.04MB/s= 56.32Mb/s
注意这仅仅是粗略估算,一般建议private network选用10Gb带宽
2.一条GCS/GES Message大约200 bytes
3.Estd Interconnect traffic(KB)参数计算公式
Estd Interconnect traffic (KB) = (('gc cr blocks received'+ 'gc current blocks received' + 'gc cr blocks served'+ 'gc current blocks served') * Block size) + (('gcs messages sent' + 'ges messages sent' + 'gcs msgs received'+ 'gcs msgs received')*200)/1024/Elapsed Time
4.评估的发送流量(MB) = (((GES_SENT+GCS_SENT)*200) + ((CR_SENT+CURRENT_SENT)* BLOCK_SIZE)) * 1.2 * 8 /1048576
四、Global Cache Efficiency Percentages
1.对RAC而言response time响应时间的要求比单节点更高,所以别指望用烂硬件搭出来的RAC性能比单机好!
2.如果Interconnect的网络延迟 > IO子系统的延迟,那么RAC本身就是性能瓶颈
3.但是IO响应时间对RAC也非常重要,例如上一讲中所述log file sync=> gc buffer busy,所以千万别用garbage storage去搭RAC!!
4.Avg global cache cr block receive time (ms):0.7
Avg global cache current block receive time (ms):1.0
1)关重要的2个指标,结合其他节点的AWR报告一起分析这2个指标,一般要求小于2ms。
2)若在RAC实例之间这2个指标差异很大,一般说明interconnect问题出现于OS buffer层或者网卡上.
5.
Time to process CR block request in the cache = (build time + flush time + send time)
Time to process current block request in the cache= (pin time + flush time + send time)
五、Messaging Statistics
1.Avg message sent queue time 一条信息进入队列到发送它的时间
2.Avg message sent queue time on ksxp 对端收到该信息并返回ACK的时间,这个指标很重要,直接反应了网络延迟,一般小于1ms
3.Send message有2种方式:
kjccsmg() �C send message (FG direct send)
kjccqmg() �C queue message (indirect send by LMS)
4.% of indirect sent messages 间接发送信息一般是排序或大的信息,流控制也可能引起indirect sent message
% of flow controlled messages 流控制最常见的原因是网络状况不佳, % of flow controlled messages应当小于1%
最好是% of direct sent messages所占用比例高。