Self-installing memory cache - does it exist?

Question

I've read about various cross-machine caching mechanisms (Redis, Velocity, nMemCached, etc...). They all seem to require a central machine to manage the cache.

Is there such a thing as a cache engine that self installs - eg if caching does not exist on the current subnet, it creates a node. If it does exist, it joins the machine to the caching pool?

Context: I have an app that deploys to around 100 users within the same subnet via ClickOnce. Each of these users access a resource via the WAN (across country and in some cases across the ocean) that performs very CPU-intensive computations and takes significant time to complete.

As a result, the app feels sluggish. I've done what I could to alleviate that by throwing long-lived queries onto separate threads. But that only takes you so far. I've added local caching (via a SQL Compact DB) which works pretty good, but most users access the similar information and together they exert a bit of pressure on the computation server. I think I can take it to the next level if I am able to ship an in-memory cache with my app that is able to seemlessly work with other machines to create a network wide caching mechanism.

Answer 1

If you don't want a master machine, or you don't want to rely on a specific network layout/installation, you could consider Peer to Peer as an option. WCF has native peer to peer support. Here is a link that looks somewhat relevant to your need: How To Design State Sharing In A Peer Network

Answer 2

You're the one who knows what will be the best, but having a "server app" that would coordinate the whole lot might be a good thing :

1. User1 asks Server "I need X".
2. Server tells User1 "Well, ask for it to DataBase"
3. User2 asks Server "I need X."
4. Server tells User2 "User1 got it."
5. ...
6. User1 tells Server "I don't want X anymore."

You could also make some type of data "uncacheable" due to their volatile nature, or to avoid clogging one of the users' connection. Sure the server will get a lot of requests, but if you'd have to compare this with a broadcast across the network solution. If I didn't understand your problematic correctly, just write a comment, disregard this, and I'll remove the answer not to mislead SO Users.