forked from mirror/ledisdb
benchmark move to wiki
This commit is contained in:
parent
7fed98b957
commit
4daaa9ae37
|
@ -118,7 +118,7 @@ Set slaveof in config or dynamiclly
|
|||
|
||||
## Benchmark
|
||||
|
||||
See benchmark.md for more.
|
||||
See [benchmark](https://github.com/siddontang/ledisdb/wiki/Benchmark) for more.
|
||||
|
||||
## Todo
|
||||
|
||||
|
|
81
benchmark.md
81
benchmark.md
|
@ -1,81 +0,0 @@
|
|||
## Environment
|
||||
|
||||
+ Darwin karamatoMacBook-Air.local 13.1.0 Darwin Kernel Version 13.1.0: Thu Jan 16 19:40:37 PST 2014; root:xnu-2422.90.20~2/RELEASE_X86_64 x86_64
|
||||
+ 2 CPU Intel Core i5 1.7GHz
|
||||
+ 4GB
|
||||
+ SSD 120G
|
||||
|
||||
Using redis-benchmark below:
|
||||
|
||||
redis-benchmark -n 10000 -t set,incr,get,lpush,lpop,lrange,mset -q
|
||||
|
||||
## Config
|
||||
|
||||
+ redis: close save and aof
|
||||
|
||||
+ ssdb: close binlog manually below:
|
||||
|
||||
void BinlogQueue::add_log(char type, char cmd, const leveldb::Slice &key){
|
||||
tran_seq ++;
|
||||
//Binlog log(tran_seq, type, cmd, key);
|
||||
//batch.Put(encode_seq_key(tran_seq), log.repr());
|
||||
}
|
||||
|
||||
+ leveldb:
|
||||
|
||||
compression = false
|
||||
block_size = 32KB
|
||||
write_buffer_size = 64MB
|
||||
cache_size = 500MB
|
||||
|
||||
|
||||
## redis
|
||||
|
||||
SET: 42735.04 requests per second
|
||||
GET: 45871.56 requests per second
|
||||
INCR: 45248.87 requests per second
|
||||
LPUSH: 45045.04 requests per second
|
||||
LPOP: 43103.45 requests per second
|
||||
LPUSH (needed to benchmark LRANGE): 44843.05 requests per second
|
||||
LRANGE_100 (first 100 elements): 14727.54 requests per second
|
||||
LRANGE_300 (first 300 elements): 6915.63 requests per second
|
||||
LRANGE_500 (first 450 elements): 5042.86 requests per second
|
||||
LRANGE_600 (first 600 elements): 3960.40 requests per second
|
||||
MSET (10 keys): 33003.30 requests per second
|
||||
|
||||
## ssdb
|
||||
|
||||
SET: 35971.22 requests per second
|
||||
GET: 47393.37 requests per second
|
||||
INCR: 36630.04 requests per second
|
||||
LPUSH: 37174.72 requests per second
|
||||
LPOP: 38167.94 requests per second
|
||||
LPUSH (needed to benchmark LRANGE): 37593.98 requests per second
|
||||
LRANGE_100 (first 100 elements): 905.55 requests per second
|
||||
LRANGE_300 (first 300 elements): 327.78 requests per second
|
||||
LRANGE_500 (first 450 elements): 222.36 requests per second
|
||||
LRANGE_600 (first 600 elements): 165.30 requests per second
|
||||
MSET (10 keys): 33112.59 requests per second
|
||||
|
||||
## ledisdb
|
||||
|
||||
SET: 38759.69 requests per second
|
||||
GET: 40160.64 requests per second
|
||||
INCR: 36101.08 requests per second
|
||||
LPUSH: 33003.30 requests per second
|
||||
LPOP: 27624.31 requests per second
|
||||
LPUSH (needed to benchmark LRANGE): 32894.74 requests per second
|
||||
LRANGE_100 (first 100 elements): 7352.94 requests per second
|
||||
LRANGE_300 (first 300 elements): 2867.79 requests per second
|
||||
LRANGE_500 (first 450 elements): 1778.41 requests per second
|
||||
LRANGE_600 (first 600 elements): 1590.33 requests per second
|
||||
MSET (10 keys): 21881.84 requests per second
|
||||
|
||||
## Conclusion
|
||||
|
||||
ledisdb is little slower than redis or ssdb, some reasons may cause it:
|
||||
|
||||
+ go is fast, but not faster than c/c++
|
||||
+ ledisdb uses cgo to call leveldb, a big cost.
|
||||
|
||||
However, **ledisdb is still an alternative NoSQL in production for you**.
|
Loading…
Reference in New Issue