Memory usage, scalability and session

Hi, I read this blog

http://blog.philipphauer.de/evaluating-vaadin-strengths-weaknesses/

about Vaadin and the following discussion and a little bit curious about this topic.

According to Vaddin doc

"User Session

A user session begins when a user first makes a request to a Vaadin servlet or portlet by opening the URL for a particularUI. All server requests belonging to a particular UI class are processed by the VaadinServlet or VaadinPortlet class. When a new client connects, it creates a new user session, represented by an instance of VaadinSession. Sessions are tracked using cookies stored in the browser."

Q1. So dose that mean even without login/count system, every client computer will have his proper “session” and I , as developper, can get the session value separately?

Q2. If my understanding about Q1 is right, is there a difference if I build up login/count system? and whether or not this will take more server memory.

Q3. I’m building a web app aiming to serve over 1 million user (not shop on line) every day. The most content will only be text and a little picture, and according to Vaadin analyses


https://vaadin.com/wiki?p_p_id=36&p_r_p_185834411_title=Scalable+web+applications&p_r_p_185834411_nodeName=vaadin.com+wiki

What size of server memory would achieve the job? or is it even possible?like 100 servers * 200 GB memory?

Q1: Yes, when a user’s browser opens a URL of a Vaadin app, a new session is created.

Q2: It depends on authentication architechture: if I remember correctly for example Spring Security creates first one session and then another is created for Vaadin after succesfull login. But, even in that case the memory overhead of first session should be minimal.

Q3: It depends… So you would have 1 million visits everyday. But how many of those are concurrently using your web application at peak hour? If for example there is 50 000 concurrent sessions and your session size is 150KB, that would be about 7.5GB of memory…

thanks a lot. I know better the calculation about the memory size now, it is the concurrency at peak hour that matters.