Ask HN: Technology behind Push Notification

Push server requires constant connection with the client and the server pushes new information to clients without requiring the client to first ask for updates. Polling requires the client to ask for updates, but does not require constant connection. The connection with polling server is required only while the poll is happening.

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.

The best is to have a read of Wikipedia's article.

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.

"push" requires that an open connection be maintained, while "poll" means you connect multiple times.

Thanks a lot everyone.