IPDaemon Component

Properties   Methods   Events   Configuration Settings   Errors  

The IPDaemon Component is a generic TCP server component based on an asynchronous, event-driven architecture. It is designed to balance the load between connections for a fast, powerful server.

Syntax

nsoftware.IPWorks.Ipdaemon

Remarks

IPDaemon is the server complement of IPPort (which is used to create client applications). They share a common design philosophy and interface. We expect you will find IPDaemon as easy to use as IPPort.

By default, each instance of IPDaemon can handle up to 1,000 simultaneous incoming connections (this number may be increased up to 100,000 or decreased to a lower value by using the MaxConnections configuration setting).

The connections are identified by a ConnectionId, an id generated by the component to identify each connection. This id is unique to each connection. IPDaemon's events also have ConnectionId as a parameter to identify the connection they relate to.

Our main goal in designing IPDaemon was to make it easy to use without sacrificing performance. The component has a minimum of properties, and six events: ConnectionRequest, Connected, DataIn, Disconnected, ReadyToSend, and Error.

IPDaemon can start to listen on a port by setting the Listening property to True. When a remote host asks for a connection, the ConnectionRequest event is fired. At that point, the connection can either be accepted or rejected. If the connection is accepted, a ConnectionId is assigned, and communication can start. From this point on, the operation is very similar to IPPort. Data is sent by assigning the data string to the DataToSend property. The address and port of the incoming connection can be found by querying the RemoteHost and RemotePort properties.

Threading in Server Applications

In .NET, all socket-based components perform all socket IO through asynchronous methods. Each call consumes a thread from the system's thread pool, and thus the number of concurrent calls is limited to the number of worker threads returned by System.Threading.ThreadPool.GetMaxThreads. When using several components that can each create a socket connection or a single component that can create multiple socket connections, the application may enter a state where all of the pool threads are in use for receiving. In a scenario such as this, the remote hosts may be waiting for data from the application before sending, but the application cannot send that data because all the threads are tied up waiting for inbound data. This is particularly important for server applications that use a daemon component.

For example, say a server application 'S' has a max thread pool size of three, and clients 'A' and 'B' connect to it and each client uploads a large file. After all uploads are complete, 'S' enqueues an asynchronous read request in the thread pool for each client. Since there is no work being done, the thread pool immediately consumes two threads which wait for any further data from 'A' and 'B'. While 'S' is still processing the first two files it has received, client 'C' connects and uploads a relatively small file. Similarly, once that upload has completed 'S' enqueues a third read request, and the thread pool immediately consumes the third and final thread to wait for further data from 'C'. Once 'S' completes processing any of the files it will enqueue a send request. However, since all three threads are in use and waiting for data from the clients, the send operation cannot be completed. The entire system becomes deadlocked because each of the clients is waiting for data from the server, which cannot send that data since all of its pool threads are in turn waiting for data from the clients.

To avoid a deadlock situation such as this, it is necessary to ensure that the maximum number of socket connections an application can make is always at least one less than the number of available worker threads. One method to achieve this is to call System.Threading.ThreadPool.SetMaxThreads during the application's startup to set the maximum number of threads to a value high enough to support the application for the duration of its execution. Another option is to trap an event such as ConnectionRequest and Disconnected from each component instance with logic to manage the max thread pool size according to the application's current needs.