Cluster Deployment
From AgileApps Support Wiki
Revision as of 18:49, 3 May 2011 by imported>Aeric (Text replace - 'Category:System Administration' to 'Category:Installation')
Prototype Deployment
The following diagram shows the kind of architecture that is typical for a production system:
The key points depicted in the diagram are:
- The platform is deployed across multiple servers.
- A Load Balancer distributes traffic across the web servers.
- Apache httpd is the front-end web server, and LongJump is the application server.
Learn more: Configuring Apache to Serve Static Content - Memcached servers reduce response time by caching data in memory.
Learn more: Configuring memcached - The suite of memcached servers are accessed by all LongJump servers, backend as well as front end. (Having one memcached server for each LongJump server is a common pattern, but is not strictly necessary.)
- The critical backend processes shown here (import, export, and scheduling, which uses quartz) are all being run from a single LongJump instance. But additional servers can be employed, as load demands.
Learn more: Managing Backend Services - Document storage (which includes pictures and image files) is managed separately from the database.
- The database is running on its own server, for added performance.
Learn more: Configuring MySQL and LongJump to Run on Separate Servers - The primary database instance and the replication instance are each running on separate servers, both for reliability and for performance of read-intensive operations.
Learn more: Using Replication with Different Master and Slave Storage Engines - Requests that access and update, whether coming from a user or a backend process, go to the primary database, while read-intensive operations (backups, reports, exports) are executed on the replicated database.
Learn more: Running Reports Using a Replicated Database Server.