Software architecture

Let me understand the real difference between "tire" and "layer" terms commonly used in architecture.

tire - referes to physical entity that is PC(s) and Server(s)
e.g - client machine running webbrowser, app serve running the web app, database server hosting the database. 

layer - refers to logical entity that is a component
e.g. userinterface layer, business logic layer, data access layer, database layer etc.

The conventiona asp or servlets, jsp or any web app that returns the html content and on the other side queries the database with SQL queries may not be a pure layered architecture because of the overlap of the layer reponsibilities.

A clean layered architecture could be

client - html, javascript, css client talking with a webservice using HTTP GET or POST, all presentation logic is part of javascript and css

dataaccess - webservice interfaceing with database using stored procedure (not SQL queries) to access data and return the data to the client in JSON or XML formats.

database - servers the webservice for the requested data using the stored procedures

The client machines can be many, the webservice if stateless can be
hosted in many servers for scalability reasons, the database server could be hosted in redudant high end server machines for availability reasons.

There are three logical "layers" and 3-tires but actually involving "n" number of physical machines.

client-HTTP(GET/POST)->webservice-TCP database connection ( to execute Stored Proc) ->DB
client<-http database="database" font="font" recordset-db="recordset-db" webservice="webservice">
 
No HTML is exchanged between client and webservice. 
No SQL command is issued by webservice to DB.

The above is a clean layered and tired architecture if the client, webservice and DB are deployed in different machines.

Reference: http://msdn.microsoft.com/en-us/library/aa905336.aspx

to be continued..

Auto Update

I am trying to understand how this software auto update might be working.
There could be three steps in it
1) Check for the availability of new version
2) Download the latest version
3) Install the downloaded version of software

Next i am going to expand on how each step could be designed or implemented

Check for availability of new version

When the current version of the application is started it could do the following in a background thread 1) Read the current version all the component that could be updated in the software
2) Go the vendor software update web server and check for the availability of newer version
2.1) Only if the exiting software license supports free updates
3) If newers version ar available the goto step2 or do nothing

Download the latest version

Options
1) Update the software user about the availability of newer version by displaying a message box and request the user for a manual update
2) Update the software user about the availability of newer version by dispaying a message box and request the user to intiate the auto download
2.1) Perform the download in the background (HTTP or FTP)

If download is successfully completed then goto to step 3 or display the error message to the user and the user to try again later.

Install the downloaded version of software

Note: Technology should support installation of newer version when current version of the software is running.
Install the downloaded software and request the user to restart the software.

Not sure how much of the above understanding and ideas are correct?

References
http://stackoverflow.com/questions/4499656/how-do-i-architect-automatic-update-of-my-clients