AgileApps Support Wiki Pre Release

Difference between revisions of "Cluster Deployment"

From AgileApps Support Wiki
imported>Aeric
imported>Aeric
Line 1: Line 1:
===Prototype Deployment===
<includeonly>===Prototype Deployment===</includeonly>
 
The following diagram shows the kind of architecture that is typical for a production system:
The following diagram shows the kind of architecture that is typical for a production system:



Revision as of 21:01, 2 October 2013

The following diagram shows the kind of architecture that is typical for a production system:

File:Prototype deployment.png

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 for the application server.
    Learn more: Installing and Configuring Apache for Use with the Platform
  • The Messaging server handles all social media interactions (e.g. Facebook and Twitter). Messages are exchanged between it and the LongJump server, to manage those interactions.
  • The HornetQ server maintains the store-and-forward message queue, for those communications.
  • Memcached servers reduce response time by caching data in memory.
    Learn more: Configuring memcached
  • The suite of memcached servers are accessed by all application servers, backend as well as front end.
  • The critical backend processes shown here (import, export, and scheduling, which uses quartz) are all being run from a single platform 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 to Run on a Separate Server
  • 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 and Storage Checks On a Replicated Database Server.