Overlapping processing
When using IIS 5.0, if a Web site running in low isolation mode or medium isolation mode fails, the only way to restart the site is to restart the entire IIS. Doing so will cause IIS to suddenly stop service, so all requests that arrive during the restart process will be invalidated.
IIS 6.0introduces a revolutionary concept, that is, the concept of overlapping processing. Based on the concept of overlapping processing, even if an application pool is reclaimed, all requests that arrive later can continue to be served. IIS 7. 0 still supports this concept.
If an application pool is recycled, the existing worker process does not quit immediately, but starts a second worker process, and once the second process starts successfully, Http.sys sends all new requests to the new worker process. Close and exit when the existing worker process has processed all the requests. Because Http.sys can queue incoming requests before sending them to a new worker process, there is no loss of page requests during recycling of the application pool.
Although there is no loss of page requests or invalidation of page requests during recycling of the application pool, there may be adverse effects on the recycling process because all data saved in the work process will be lost during recycling of the application pool. By default, ASP.NET saves session-state data and in-process cached data (we call it InProc data). The validity time of this data is exactly the same as the lifetime of the worker process, so the data must be recreated during the recycling of the application pool. Therefore, you must consider saving session state data out of the process. Session-state data can be saved in StateServer, SQLServer, or other external session-state storage. In addition, load performance problems may occur when starting a new worker process. At this point, all aspects of IIS and ASP.NET must be loaded into the worker process, so overall loading time is relatively long, often taking a few seconds. As a result, the first page that runs after the application pool is recycled often takes longer to run properly than the normal operation of the application pool.
