问题描述
- leveldb的测试结果怎么理解的问题。
-
最近对leveldb进行了一些测试,因为只是想与leveldb进行对比,不需要对其进行详细的了解,所以想请教测试结果问题。 LevelDB: version 1.15
Date: Tue Nov 17 16:16:36 2015
CPU: 24 * Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz
CPUCache: 15360 KB
Keys: 16 bytes each
Values: 200 bytes each (100 bytes after compression)
Entries: 5000000
RawSize: 1030.0 MB (estimated)
FileSize: 553.1 MB (estimated)WARNING: Snappy compression is not enabled
fillseq : 5.702 micros/op; 36.1 MB/s
fillsync : 964.518 micros/op; 0.2 MB/s (5000 ops)
fillrandom : 18.614 micros/op; 11.1 MB/soverwrite : 23.934 micros/op; 8.6 MB/s
readrandom : 7.279 micros/op; (5000000 of 5000000 found)
readrandom : 5.498 micros/op; (5000000 of 5000000 found)
readseq : 0.346 micros/op; 595.5 MB/sreadreverse : 0.753 micros/op; 273.7 MB/s
compact : 7412687.000 micros/op;
readrandom : 4.926 micros/op; (5000000 of 5000000 found)
readseq : 0.315 micros/op; 654.1 MB/sreadreverse : 0.754 micros/op; 273.2 MB/s
fill100K : 3054.001 micros/op; 31.2 MB/s (5000 ops)
crc32c : 6.280 micros/op; 622.0 MB/s (4K per op)
snappycomp : 6340.000 micros/op; (snappy failure)
snappyuncomp : 6200.000 micros/op; (snappy failure)
acquireload : 0.530 micros/op; (each op is 1000 loads)问题1:这里边如何计算leveldb的读写性能,单位是 op/sec,比如随机读,是不是后边的 MB/s 除以 每个kv item的大小?
问题2:为什么会有两个readseq,这两个有什么区别吗?
问题3:前三个是什么参数 fill...是什么意思?
问题4:怎么计算写性能,是不是使用参数overwrite 的 8.6MB/(16+200)byte