浙江省高等学校教师教育理论培训

微信搜索“教师资格证岗前培训”小程序

  博客园  :: 首页  :: 新随笔  :: 联系 :: 订阅 订阅  :: 管理

treapdb - A fast and stable key-value store based on Treap data structure - Google Project Hosting

TreapDB is a key-value store based on Treap

News

 

Unit Test

Features

  • can be a embedded library or a standalone RPC-Server
  • can 'talk' memcache and thrift at the same time
  • master-slave replication
  • random "get" speed: 15000 tps(RPC Server); 32000tps(Embedded)
  • bulkPut speed: 17000 tps, bulkGet speed:30000tps

 

 

  • service specification:
  • namespace java fx.sunjoy.server.gen

    struct Pair {
     
    1: string key,
     
    2: binary value,
    }

    service
    TreapDBService{
           
    void put(1:string key, 2:binary value);
            binary
    get(1:string key),
           
    void bulkPut(1:map<string,binary> kvMap);
            list
    <Pair> bulkGet(1:list<string> keyList);
            list
    <Pair> prefix(1:string prefixStr,2:i32 limit,3:string startK,4:bool asc),
            list
    <Pair> bulkPrefix(1:list<string> prefixList,2:i32 limit,3:string startK,4:bool asc),
            list
    <Pair> kmax(1:i32 k),
            list
    <Pair> kmin(1:i32 k),
            list
    <Pair> range(1:string kStart, 2:string kEnd,3:i32 limit),
            list
    <Pair> before(1:string key,2:i32 limit),
            list
    <Pair> after(1:string key,2:i32 limit),
            i32 length
    (),
           
    bool remove(1:string key),
           
    bool removePrefix(1:string key),
           
    void optimize(1:i32 amount)
    }

Operations of TreapDB

 

OperationUsageParameters
putinsert a new key-value pair, or replace an old key-value pairkey and value
getget the value of the indicated keykey
bulkPutinset 2 or more key-value pairs togethermap of key and value
bulkGetget the value of 2 or more indicated keyslist of keys
prefixget the value of the keys with the indicated prefix stringprefix and number of result
bulkPrefixget the value of the keys with the indicated prefix stringslist of prefix, number of result, beginning key, sort the key-value pairs in ascending or descending order
kmaxget the k maximum key-value pairsk
kminget the k minimum key-value pairsk
rangeget the key-value pairs whose key is between the indicated keysbeginning key and ending key
beforeget the key-value pairs whose key is before the indicated key in alphabetical orderkey and number of result
afterget the key-value pairs whose key is after the indicated key in alphabetical orderkey and number of result
lengthget the number of key-value pairsnone
removedelete the indicated key-value pairkey
removePrefixdelete value of the keys with indicated prefix stringprefix
optimizeoptimize the space usage of index filenumber of nodes need to be optimized(1024 is recommended value)

 

 

Sample

http://code.google.com/p/treapdb/wiki/SomeBenchmark

RPC Server

  • memcached protocol compatible server for test usage
  • thrift-based server for production usage

Download & Configure

  • run
    • ./treapdb.sh conf/TreapDBConf.xml(master mode)
    • ./treapdb.sh conf/TreapDBConf_Slave.xml(slave mode)
  • configuration:
TreapDBConf.xml
     
     
<?xml version="1.0" encoding="UTF-8"?>
     
<TreapDB>
       
<Params>
               
<Port>
                       
<Memcache>11811</Memcache>
                         //Listening port of memcache protocol
                       
<Thrift>11812</Thrift>    
                         //Listening port of thrift protocol
               
</Port>
               
<Index>
                       
<FilePath>/var/log/treapdb/master</FilePath>
                          //Index file name
                       
<BlockSize>64</BlockSize>  
                          //size of index-block-item;
                          //default key length is 26(=64-38),38 is the node-overhead,
                          //TreapDB supports max key-length: 127 bytes)

               
</Index>
               
<MMapSize>128</MMapSize> //the more the better, 128MB is enough for 2 Million keys
               
<Replication>
                       
<Role>Master</Role>
               
</Replication>
       
</Params>
       
</TreapDB>

 

TreapDBConf_Slave.xml

 

       <?xml version="1.0" encoding="UTF-8"?>
       
<TreapDB>
         
<Params>
               
<Port>
                       
<Memcache>11911</Memcache>
                       
<Thrift>11912</Thrift>
               
</Port>
               
<Index>
                       
<FilePath>/var/log/treapdb/slave</FilePath>
                       
<BlockSize>64</BlockSize>
               
</Index>
               
<MMapSize>128</MMapSize>
               
<Replication>
                       
<Role>Slave</Role>
                       
<Source>10.61.1.170:11811</Source>
                       //Set master address and port here!
               
</Replication>
         
</Params>
       
</TreapDB>

How to build from source

  1. svn checkout http://treapdb.googlecode.com/svn/trunk/
  2. ant dist
  3. use the jars generated in build/dist

How to write client code

Java:
     String host = "localhost";
     
Integer port = 11812; //thrift port
     
TreapDBClient client = TreapDBClientFactory.getClient(host, port);
     client
.put(...)
     client
.get(...)
     client
.prefix(...)
     
...
Python:
     import pytreap
     client
= pytreap.connect('localhost',11812) //thrift port
     client
.get(...)
     client
.put(...)
     client
.remove(...)

Can TreapDB be used in an embedded way?

Yes,put libtreap-xx.jar in your app's classpath.
    DiskTreap<String, Serializable> treap = new DiskTreap<String,Serializable>
(new File("/usr/local/indexpath"));

    treap
.put(...)
    treap
.get(...)
   
...

   
/* key can be any type which implements Comparable;
    value can be any type which implements Serializable*/

Only Java?

  • TreapDB server is based on thrift, so you can generate your client in

any programming language.

  • For example, generate the python client code.
  • thrift --gen py res/service.txt
  • res/service.txt is a service specification:
Python client: http://code.google.com/p/treapdb/wiki/PythonClientExample

What is treap?

In computer science, the treap and the randomized binary search tree are two closely-related forms of binary search tree data structures that maintain a dynamic set of ordered keys and allow binary searches among the keys. After any sequence of insertions and deletions of keys, the shape of the tree is a random variable with the same probability distribution as a random binary tree; in particular, with high probability its height is proportional to the logarithm of the number of keys, so that each search, insertion, or deletion operation takes logarithmic time to perform.

http://en.wikipedia.org/wiki/Treap

Contact the author

posted on 2012-03-28 20:50  lexus  阅读(269)  评论(0编辑  收藏  举报