You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@dubbo.apache.org by "wangruyu1 (GitHub)" <gi...@apache.org> on 2019/01/04 09:24:46 UTC
[GitHub] [incubator-dubbo] wangruyu1 opened issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
虽然dubbo不建议大数据,但是我是测试的localhost.
用10个并发,每个并发循环访问20次（其实相当于200个并发）,服务器直接返回3M字符串，不做任何处理,处理时间都需要220ms左右。
相关配置:
producer:
dubbo.service.connections = 100
dubbo.service.protocol=dubbo
consumer:
dubbo.reference.connections = 100
我的疑问:既然我走的localhost,带宽不用考虑吧,长连接设置的也是100个,每次请求使用一个长连接这处理应该会超级快啊？但是最终结果耗时还是在数据传输,200多ms.
这个问题哪位大神能解答下，很急（大数据传输无法避免，上线项目需要用到）。谢谢
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] wangruyu1 commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "wangruyu1 (GitHub)" <gi...@apache.org>.
> Dubbo protocol use "single long tcp connection", so it is not fit to transfer file or data of large size.
我也看了文档，建议短连接。
但是我现在有个疑问。假如说我对consumer1设置共享的长连接是100，用100个并发请求consumer1的service1，这样100个并发应该是会复用长连接，所以100个并发实际使用到的长连接并没有100个（这个是长连接复用规则导致的吧？源码我不清楚。。。）。所以我想问能不能配置100个并发在请求时循环使用新的长连接？~~~~(>_<)~~~~，还能只能修改源码呢？？？谢谢
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] wangruyu1 commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "wangruyu1 (GitHub)" <gi...@apache.org>.
> > > @wangruyu1 `<dubbo:reference connections="100">`在远程引用的时候就会创建完成，然后请求会轮询这100个连接，代码在DubboInvoker.doInvoke
> > > > 所以我想问能不能配置100个并发在请求时循环使用新的长连接
> > >
> > >
> > > 不是很懂你这句话，你的意思是继续创建新连接吗
> >
> >
> > 我的意思是100个并发同时第一次请求时,是不是都是用的这100个长连接（因为出现长连接复用会小于100个）？当101次请求来时才会出现长连接复用。因为我从现在测试判断出100个并发并没有都被用到了，应该是出现了长连接复用。
> > 我现在想实现每个请求都是优先使用100个长连接中没被用到，如果100个都被用到了，那就循环的复用。相当于,for(i=0;i<100;i++) con=cons[i %100]这样获取长连接。
> > 这能不能配置还是只能改源码？
>
> 目前消费端源码就是这样实现的；
> DubboInvoker
>
> ```java
> @Override
> protected Result doInvoke(final Invocation invocation) throws Throwable {
> RpcInvocation inv = (RpcInvocation) invocation;
> final String methodName = RpcUtils.getMethodName(invocation);
> inv.setAttachment(Constants.PATH_KEY, getUrl().getPath());
> inv.setAttachment(Constants.VERSION_KEY, version);
>
> ExchangeClient currentClient;
> if (clients.length == 1) {
> currentClient = clients[0];
> } else {
> currentClient = clients[index.getAndIncrement() % clients.length];
> }
> ...
> ```
>
> `currentClient = clients[index.getAndIncrement() % clients.length];`已经实现了你所说的复用。
恩，谢谢，估计是我配置有问题。
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] gudegg commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "gudegg (GitHub)" <gi...@apache.org>.
@wangruyu1 `<dubbo:reference connections="100">`在远程引用的时候就会创建完成，然后请求会轮询这100个连接，代码在DubboInvoker.doInvoke
> 所以我想问能不能配置100个并发在请求时循环使用新的长连接
不是很懂你这句话，你的意思是继续创建新连接吗
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] tswstarplanet commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "tswstarplanet (GitHub)" <gi...@apache.org>.
Dubbo protocol use "single long tcp connection", so it is not fit to transfer file or data of large size.
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] tswstarplanet commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "tswstarplanet (GitHub)" <gi...@apache.org>.
Dubbo suggests that transfering data of large size should use protocols that do not contains dubbo.
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] tswstarplanet commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "tswstarplanet (GitHub)" <gi...@apache.org>.
Dubbo suggests that you can transfer file or data of large size with protocol like hessian or http. You can try it.
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] wangruyu1 commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "wangruyu1 (GitHub)" <gi...@apache.org>.
谢谢大神们帮我解释下。。。
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] gudegg commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "gudegg (GitHub)" <gi...@apache.org>.
> > @wangruyu1 `<dubbo:reference connections="100">`在远程引用的时候就会创建完成，然后请求会轮询这100个连接，代码在DubboInvoker.doInvoke
> > > 所以我想问能不能配置100个并发在请求时循环使用新的长连接
> >
> >
> > 不是很懂你这句话，你的意思是继续创建新连接吗
>
> 我的意思是100个并发同时第一次请求时,是不是都是用的这100个长连接（因为出现长连接复用会小于100个）？当101次请求来时才会出现长连接复用。因为我从现在测试判断出100个并发并没有都被用到了，应该是出现了长连接复用。
> 我现在想实现每个请求都是优先使用100个长连接中没被用到，如果100个都被用到了，那就循环的复用。相当于,for(i=0;i<100;i++) con=cons[i %100]这样获取长连接。
> 这能不能配置还是只能改源码？
目前消费端源码就是这样实现的；
DubboInvoker
```java
@Override
protected Result doInvoke(final Invocation invocation) throws Throwable {
RpcInvocation inv = (RpcInvocation) invocation;
final String methodName = RpcUtils.getMethodName(invocation);
inv.setAttachment(Constants.PATH_KEY, getUrl().getPath());
inv.setAttachment(Constants.VERSION_KEY, version);
ExchangeClient currentClient;
if (clients.length == 1) {
currentClient = clients[0];
} else {
currentClient = clients[index.getAndIncrement() % clients.length];
}
...
```
`currentClient = clients[index.getAndIncrement() % clients.length];`已经实现了你所说的复用。
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] wangruyu1 commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "wangruyu1 (GitHub)" <gi...@apache.org>.
> @wangruyu1 `<dubbo:reference connections="100">`在远程引用的时候就会创建完成，然后请求会轮询这100个连接，代码在DubboInvoker.doInvoke
>
> > 所以我想问能不能配置100个并发在请求时循环使用新的长连接
>
> 不是很懂你这句话，你的意思是继续创建新连接吗
我的意思是100个并发同时第一次请求时,是不是都是用的这100个长连接（因为出现长连接复用会小于100个）？当101次请求来时才会出现长连接复用。因为我从现在测试判断出100个并发并没有都被用到了，应该是出现了长连接复用。
我现在想实现每个请求都是优先使用100个长连接中没被用到，如果100个都被用到了，那就循环的复用。相当于,for(i=0;i<100;i++) con=cons[i %100]这样获取长连接。
这能不能配置还是只能改源码？
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org
[GitHub] [incubator-dubbo] Kiddinglife commented on issue #3145: dubbo本地传输3M大小的数据为什么这么慢？
Posted by "Kiddinglife (GitHub)" <gi...@apache.org>.
Hi @wangchangbing @gudegg,
Did you guys find out the reason why it is slow to transfer 3m data in local network with long connection?
220ms latency really surprised me...
regards.
[ Full content available at: https://github.com/apache/incubator-dubbo/issues/3145 ]
This message was relayed via gitbox.apache.org for notifications@dubbo.apache.org