9、redis之事务2-Jedis的八种调用方式(事务、管道、分布式)介绍
1、普通同步
1 @Test 2 public void test1Normal() { 3 Jedis jedis = new Jedis("localhost"); 4 long start = System.currentTimeMillis(); 5 for (int i = 0; i < 100000; i++) { 6 String result = jedis.set("n" + i, "n" + i); 7 } 8 long end = System.currentTimeMillis(); 9 System.out.println("Simple SET: " + ((end - start)/1000.0) + " seconds"); 10 jedis.disconnect(); 11 }
例子中的for循环里的每一个set操作都是一个事务。
2、事务方式(Transactions)
redis的事务很简单,他主要目的是保障,一个client发起的事务中的命令可以连续的执行,而中间不会插入其他client的命令。
1 @Test 2 public void test2Trans() { 3 Jedis jedis = new Jedis("localhost"); 4 long start = System.currentTimeMillis(); 5 Transaction tx = jedis.multi(); 6 for (int i = 0; i < 100000; i++) { 7 tx.set("t" + i, "t" + i); 8 } 9 List<Object> results = tx.exec(); 10 long end = System.currentTimeMillis(); 11 System.out.println("Transaction SET: " + ((end - start)/1000.0) + " seconds"); 12 jedis.disconnect(); 13 }
我们调用jedis.watch(keys)
方法来监控key,如果调用后key值发生变化,则整个事务会执行失败。另外,事务中某个操作失败,并不会回滚其他操作。这一点需要注意。还有,我们可以使用discard()
方法来取消事务。
3、管道(Pipelining)
有时,我们需要采用异步方式,一次发送多个指令,不同步等待其返回结果。这样可以取得非常好的执行效率。这就是管道,调用方法如下:
1 @Test 2 public void test3Pipelined() { 3 Jedis jedis = new Jedis("localhost"); 4 Pipeline pipeline = jedis.pipelined(); 5 long start = System.currentTimeMillis(); 6 for (int i = 0; i < 100000; i++) { 7 pipeline.set("p" + i, "p" + i); 8 } 9 List<Object> results = pipeline.syncAndReturnAll(); 10 long end = System.currentTimeMillis(); 11 System.out.println("Pipelined SET: " + ((end - start)/1000.0) + " seconds"); 12 jedis.disconnect(); 13 }
4、管道中调用事务
就Jedis提供的方法而言,是可以做到在管道中使用事务,其代码如下:
1 @Test 2 public void test4combPipelineTrans() { 3 jedis = new Jedis("localhost"); 4 long start = System.currentTimeMillis(); 5 Pipeline pipeline = jedis.pipelined(); 6 pipeline.multi(); 7 for (int i = 0; i < 100000; i++) { 8 pipeline.set("" + i, "" + i); 9 } 10 pipeline.exec(); 11 List<Object> results = pipeline.syncAndReturnAll(); 12 long end = System.currentTimeMillis(); 13 System.out.println("Pipelined transaction: " + ((end - start)/1000.0) + " seconds"); 14 jedis.disconnect(); 15 }
但是经测试(见本文后续部分),发现其效率和单独使用事务差不多,甚至还略微差点。
5、分布式直连同步调用
1 @Test 2 public void test5shardNormal() { 3 List<JedisShardInfo> shards = Arrays.asList( 4 new JedisShardInfo("localhost",6379), 5 new JedisShardInfo("localhost",6380)); 6 7 ShardedJedis sharding = new ShardedJedis(shards); 8 9 long start = System.currentTimeMillis(); 10 for (int i = 0; i < 100000; i++) { 11 String result = sharding.set("sn" + i, "n" + i); 12 } 13 long end = System.currentTimeMillis(); 14 System.out.println("Simple@Sharing SET: " + ((end - start)/1000.0) + " seconds"); 15 16 sharding.disconnect(); 17 }
这个是分布式直接连接,并且是同步调用,每步执行都返回执行结果。类似地,还有异步管道调用
6、分布式直连异步调用
1 @Test 2 public void test6shardpipelined() { 3 List<JedisShardInfo> shards = Arrays.asList( 4 new JedisShardInfo("localhost",6379), 5 new JedisShardInfo("localhost",6380)); 6 7 ShardedJedis sharding = new ShardedJedis(shards); 8 9 ShardedJedisPipeline pipeline = sharding.pipelined(); 10 long start = System.currentTimeMillis(); 11 for (int i = 0; i < 100000; i++) { 12 pipeline.set("sp" + i, "p" + i); 13 } 14 List<Object> results = pipeline.syncAndReturnAll(); 15 long end = System.currentTimeMillis(); 16 System.out.println("Pipelined@Sharing SET: " + ((end - start)/1000.0) + " seconds"); 17 18 sharding.disconnect(); 19 }
7、分布式连接池同步调用
如果,你的分布式调用代码是运行在线程中,那么上面两个直连调用方式就不合适了,因为直连方式是非线程安全的,这个时候,你就必须选择连接池调用
1 @Test 2 public void test7shardSimplePool() { 3 List<JedisShardInfo> shards = Arrays.asList( 4 new JedisShardInfo("localhost",6379), 5 new JedisShardInfo("localhost",6380)); 6 7 ShardedJedisPool pool = new ShardedJedisPool(new JedisPoolConfig(), shards); 8 9 ShardedJedis one = pool.getResource(); 10 11 long start = System.currentTimeMillis(); 12 for (int i = 0; i < 100000; i++) { 13 String result = one.set("spn" + i, "n" + i); 14 } 15 long end = System.currentTimeMillis(); 16 pool.returnResource(one); 17 System.out.println("Simple@Pool SET: " + ((end - start)/1000.0) + " seconds"); 18 19 pool.destroy(); 20 }
上面是同步方式,当然还有异步方式
8、分布式连接池异步调用
1 @Test 2 public void test8shardPipelinedPool() { 3 List<JedisShardInfo> shards = Arrays.asList( 4 new JedisShardInfo("localhost",6379), 5 new JedisShardInfo("localhost",6380)); 6 7 ShardedJedisPool pool = new ShardedJedisPool(new JedisPoolConfig(), shards); 8 9 ShardedJedis one = pool.getResource(); 10 11 ShardedJedisPipeline pipeline = one.pipelined(); 12 13 long start = System.currentTimeMillis(); 14 for (int i = 0; i < 100000; i++) { 15 pipeline.set("sppn" + i, "n" + i); 16 } 17 List<Object> results = pipeline.syncAndReturnAll(); 18 long end = System.currentTimeMillis(); 19 pool.returnResource(one); 20 System.out.println("Pipelined@Pool SET: " + ((end - start)/1000.0) + " seconds"); 21 pool.destroy(); 22 }
需要注意的地方:
1)事务和管道都是异步模式。在事务和管道中不能同步查询结果。比如下面两个调用,都是不允许的,如下:
1 Transaction tx = jedis.multi(); 2 for (int i = 0; i < 100000; i++) { 3 tx.set("t" + i, "t" + i); 4 } 5 System.out.println(tx.get("t1000").get()); //不允许 6 List<Object> results = tx.exec(); 7 … 8 … 9 Pipeline pipeline = jedis.pipelined(); 10 long start = System.currentTimeMillis(); 11 for (int i = 0; i < 100000; i++) { 12 pipeline.set("p" + i, "p" + i); 13 } 14 System.out.println(pipeline.get("p1000").get()); //不允许 15 List<Object> results = pipeline.syncAndReturnAll();
2)事务和管道都是异步的,个人感觉,在管道中再进行事务调用,没有必要,不如直接进行事务模式。
3)分布式中,连接池的性能比直连的性能略好(见后续测试部分)。
4)分布式调用中不支持事务,因为事务是在服务器端实现,而在分布式中,每批次的调用对象都可能访问不同的机器,所以,没法进行事务。
转载自:http://www.open-open.com/lib/view/open1410485827242.html