MongoDb和YCSB: workload A:为什么更新延迟这么小？

Question

我在通过10G以太网连接的两个节点上运行MongoDb和YCSB，工作负载A(读取:更新= 50:50)。

ycsb输出结果看起来不错，但只有一件事:为什么更新延迟这么小？更新延迟是否仅包括MongoDb本地操作延迟？

例如，在下面的运行中，更新延迟仅为4微秒。但我的网络延迟是100+微秒。所以在我看来这很奇怪。

谢谢!

YCSB Client 0.1
Command line: -db com.yahoo.ycsb.db.AsyncMongoDbClient -s -P LisWorkloadA -p mongodb.url=mongodb://192.168.4.168:27017/ycsb?w=0 -threads 1 -t
mongo connection created with mongodb://192.168.4.168:27017/ycsb?w=0
18:09:22.758 [Thread-1] DEBUG c.a.m.c.c.b.BootstrapConnectionFactory - Simple MongoDB bootstrap to 192.168.4.168/192.168.4.168:27017.
18:24:23.109 [Thread-1] DEBUG c.a.mongodb.client.ClientImpl - MongoDB Connection closed: MongoDB(56194-->192.168.4.168/192.168.4.168:27017)
[OVERALL], RunTime(ms), 900433.0
[OVERALL], Throughput(ops/sec), 2878.1730567404793
[READ], Operations, 1296287.0

***[READ], AverageLatency(us), 687.2264621954861***

[READ], MinLatency(us), 105.0
[READ], MaxLatency(us), 5455871.0
[READ], 95thPercentileLatency(us), 1220.0
[READ], 99thPercentileLatency(us), 1749.0
[READ], Return=OK, 1296287
[CLEANUP], Operations, 1.0
[CLEANUP], AverageLatency(us), 1672.0
[CLEANUP], MinLatency(us), 1672.0
[CLEANUP], MaxLatency(us), 1672.0
[CLEANUP], 95thPercentileLatency(us), 1672.0
[CLEANUP], 99thPercentileLatency(us), 1672.0
[UPDATE], Operations, 1295315.0

***[UPDATE], AverageLatency(us), 4.323250329070535***

[UPDATE], MinLatency(us), 1.0
[UPDATE], MaxLatency(us), 42015.0
[UPDATE], 95thPercentileLatency(us), 6.0
[UPDATE], 99thPercentileLatency(us), 10.0
[UPDATE], Return=NOT_FOUND, 1295315

Answer 1

10G以太网的延迟可以低至2微秒，因此MongoDB报告的数字是可能的。您确定您的100+微秒测量正确吗？它可能会包含一些开销。