I hope this is not a repeated post. I would like to know the exact idea/technology behind developing a push server.
The exact query i have is technology wise whats the exact implementation difference between push server and polling client ?
Push server basically does not require much at all download bandwith, when polling requires bandwith for each update request. But push server requires open connection per active client, which might be in the millions at the same time. Pull server does not need active connections to be open all the time. Only when a request happens. For example web servers use pull method to service more clients with less cpu and memory usage because the connections are actually only open when http request/reply happens. Games and chats use push method to move information to the server and to push information given by another client to other clients. But for the server to push data to clients the server at least needs to have the ip address of the client with udp or open socket with tcp.
http://en.wikipedia.org/wiki/Push_technology
Basically clients subscribe to channels they want to be kept up to date with and they get notified of changes.
They're being "pushed" information.
Behind the myriad of complex implementations, it's very simple.
Have a service allowing subscriptions and removals, a channel system (handling multiple lists of clients) - although not necessarily - and the server a method to push the info out to the clients.
From there it's all about getting smarter about data... Throttling data, chunking client lists, periodically testing if clients are still alive, etc. But overall, simple.