Suggestion required to achieve caching functionality in WCF

I’ve a WCF service which is used to get some data from my database. Size of database is very large approximately 2 GB. So I cache this data. I want when service host this data should be cached so I firstly used

InstanceContextMode = InstanceContextMode.Single

This Service Behviors allows me that I can simply write caching code in service constructor, as constructor will only be invoked when servicehost.open (); method call. Whenever client will call this service through proxy constructor will not invoked. This works very fine. Later I realized that this InstanceContextMode has performance issue when 1000 users call this service at a time, because only single instance of this service serves the all requests. To get maximum performance I changed my settings to

InstanceContext开发者_运维百科Mode = InstanceContextMode.PerCall with ConcurrencyMode = ConcurrencyMode.Multiple

Now I want to get the same caching feature that is when service host data would be cached. Please help me to solve this problem.

Please do let me know either through CustomBehaviors I can achieve this?

Regards, Rizwan

Combination of InstanceContextMode.PerCall with ConcurrencyMode.Multiple doesn't make sense. PerCall instancing creates new service instance for each request. Multiple concurrency mode allow service instance to handle multiple parallel requests. How can service instance which is created only to handle single request (PerCall) be used to handle parallel requests (Multiple)?

I think the problem which occured by calling your singleton service (I believe that you also used ConcurrencyMode.Multiple) with 1000 concurrent users is simply based on the fact that server doesn't have performance to deal with so many concurrent users or the service code / caching is not optimized. This will not be solved by using PerCall instancing.

Anyway if you want to share some cache among multiple PerCall instances of the service you have to create some well known object representing your cache - use singleton pattern or service locator. This object will handle caching and concurrent access from service instances.