leveldb初探
近来各种nosql像雨后春笋一样，最近听说HyperDex比较火爆，为避免落伍，今天闲来无事，玩了一把leveldb，选择leveldb把玩的原因三点：
0.Google出品；
1.作者Jeff Dean；
2.玩不起HBase；
于是leveldb成了一个很好的材料；
下面进入正题，或者FAQ，很多人都在讲，抄来用用：
Q：什么是leveldb？
A：LevelDB is a fast key-value storage library written at Google that provides an ordered mapping from string keys to string values.
也可以说Leveldb是一个Google实现的高效kv数据库，重点是开源。据说v1.2目前能够支持十亿级别数据量。
Q：leveldb都有哪些特点？
A：从官方摘抄如下的Features：
0.Keys and values are arbitrary byte arrays.
1.Data is stored sorted by key.
2.Callers can provide a custom comparison function to override the sort order.
3.The basic operations are Put(key,value), Get(key), Delete(key).
4.Multiple changes can be made in one atomic batch.
5.Users can create a transient snapshot to get a consistent view of data.
6.Forward and backward iteration is supported over the data.
7.Data is automatically compressed using the Snappy compression library.
8.External activity (file system operations etc.) is relayed through a virtual interface so 9.users can customize the operating system interactions.
10.Detailed documentation about how to use the library is included with the source code.
Q：小白如何checkout&编译code？
A：
checkout:
git clone https://code.google.com/p/leveldb/
svn co http://leveldb.googlecode.com/svn/trunk/ leveldb
编译方式很多，这里列出我的环境：
Linux **** 2.6.32-28-generic #55-Ubuntu SMP Mon Jan 10 21:21:01 UTC 2011 i686 GNU/Linux MemTotal : 1026484 kB model name : Intel(R) Core(TM)2 Duo CPU P9400 @ 2.40GHz stepping : 10 cpu MHz : 2298.463 cache size : 6144 KB gcc version 4.4.3 (Ubuntu 4.4.3-4ubuntu5)
剩下的都懂得开始make^-^
Q：Benchmark数据？
A：条件所限，在虚拟机上进行了测试，系统信息如上，为控制文件数量，特别将sst文件大小设为32M，默认2M：static const int kTargetFileSize = 32 * 1048576; via version_set.cc：23,这里列出两组测试数据：
$./db_bench --num=100000 --write_buffer_size=$((256*1024*1024)) LevelDB: version 1.2 Date: Sun Feb 26 16:53:11 2012 CPU: 1 * Intel(R) Core(TM)2 Duo CPU P9400 @ 2.40GHz CPUCache: 6144 KB Keys: 16 bytes each Values: 100 bytes each (50 bytes after compression) Entries: 100000 RawSize: 11.1 MB (estimated) FileSize: 6.3 MB (estimated) WARNING: Snappy compression is not enabled ------------------------------------------------ fillseq : 14.581 micros/op; 7.6 MB/s fillsync : 91140.200 micros/op; 0.0 MB/s (100 ops) fillrandom : 24.181 micros/op; 4.6 MB/s overwrite : 32.435 micros/op; 3.4 MB/s readrandom : 25.788 micros/op; readrandom : 19.767 micros/op; readseq : 6.858 micros/op; 16.1 MB/s readreverse : 54.625 micros/op; 2.0 MB/s compact : 2306046.000 micros/op; readrandom : 54.663 micros/op; readseq : 2.142 micros/op; 51.6 MB/s readreverse : 8.731 micros/op; 12.7 MB/s fill100K : 5607.600 micros/op; 17.0 MB/s (100 ops) crc32c : 21.006 micros/op; 186.0 MB/s (4K per op) snappycomp : 204255.000 micros/op; (snappy failure) snappyuncomp : 175050.000 micros/op; (snappy failure) acquireload : 4.271 micros/op; (each op is 1000 loads) $./db_bench --num=1000000 --write_buffer_size=$((256*1024*1024)) LevelDB: version 1.2 Date: Sat Feb 25 17:26:31 2012 CPU: 1 * Intel(R) Core(TM)2 Duo CPU P9400 @ 2.40GHz CPUCache: 6144 KB Keys: 16 bytes each Values: 100 bytes each (50 bytes after compression) Entries: 1000000 RawSize: 110.6 MB (estimated) FileSize: 62.9 MB (estimated) WARNING: Snappy compression is not enabled ------------------------------------------------ fillseq : 10.199 micros/op; 10.8 MB/s fillsync : 87355.685 micros/op; 0.0 MB/s (1000 ops) fillrandom : 16.159 micros/op; 6.8 MB/s overwrite : 23.154 micros/op; 4.8 MB/s readrandom : 32.558 micros/op; readrandom : 35.783 micros/op; readseq : 4.380 micros/op; 25.3 MB/s readreverse : 6.911 micros/op; 16.0 MB/s compact : 7270752.000 micros/op; readrandom : 28.186 micros/op; readseq : 0.837 micros/op; 132.2 MB/s readreverse : 2.349 micros/op; 47.1 MB/s fill100K : 2924.748 micros/op; 32.6 MB/s (1000 ops) crc32c : 10.733 micros/op; 363.9 MB/s (4K per op) snappycomp : 34784.000 micros/op; (snappy failure) snappyuncomp : 46721.000 micros/op; (snappy failure) acquireload : 2.106 micros/op; (each op is 1000 loads)
总结：Leveldb轻量级kv本地存储库，读性能一般、随机写不错。接下来消化实现原理。
更多测试结果访问http://leveldb.googlecode.com/svn/trunk/doc/benchmark.html