1. TCP/IP heartbeat packet. xe2 datasnap itself is built in. You can set several attributes. This has been tested. You do not need to write data by yourself as you do in Delphi 2010 datasnap.Code.
2. the built-in thread pool of xe2 datasnap is the indy10 thread pool used. Indy10 is a communication control that blocks and multithreading. In actual tests, datasnap is a thread corresponding to a client connection,
The thread pool is returned until the client connects to the disconnected thread. Even if a client does nothing after connection, it will occupy a thread object on the server. How many connected customers are required for the server?
Thread. If the number of customers is large, this is very inefficient. There are many threads enabled, and the CPU time is spent on context switching of the thread. Therefore, when there are a large number of concurrent connections, the port is preferred.
3. After the datasnap thread pool is full, new customers cannot connect to the server. At this time, the customer's mouse remains in the waiting status and no prompt is displayed. If a customer closes the connection at this time,
Then the customer in the waiting state will automatically connect to the server. Therefore, the maximum number of concurrent connections is set after the maxthreads attribute of the thread pool is set.
4. in actual tests, about four hundred clients are opened. Each client sends (query, add, modify, and submit) requests to the server at random every second. After a long period of operation, the application server is still robust
The SQL Server 2000 database server does not support running. An error occurs when the queue waiting for processing is too long. SQL Server 2000 automatically suspends the service and must be started manually.
This shows that a large number of concurrent (query, add, modify, and submit) operation data events will cause the database to become a machine. For enterprise applications, data must be processed. For enterprise application architecture
Not only the application server in the middle layer, but also the capacity of the database server.