High Performance Clusters for Drupal, Joomla, Mambo and WordPress sites
Zubr Communications Managed Clusters are specifically designed to allow the peak performance of high traffic websites running under complex content management systems. Our clusters are specially tuned for the unique challanges presented by Drupal, Joomla, Mambo and even large Wordpress installations allowing the customers truly push the hardware to the limit without breaking the bank.
"Standard" Multi-server Cluster Architecture
Eventually, a CMS driven site, be that site powered by Drupal, WordPress, Joomla, Mambo, or any other content management system, becomes too heavy for a single server. That's when it gets moved to the "clusters." Call the sales person, add a new server(s) or create some virtual instances and and Presto! everyone is happy. The shiny new architecture gets added to the presentations, lots of people applaud and things just work...
That is, That is, until a random entry on the site becomes a viral link Facebook's Like, or ends up featured on a site like Slashdot, Digg, Drudge Report. Basically, any site that actually has traffic:
| CPU | RAM | |
|---|---|---|
| Web server | 2x Dual Core | 16 GB |
| Database server | 2x Dual Core | 32 GB |
| Caching server | 1x Dual Core | 8 GB |
All servers are connected via dual gigabit ethernet to a non-blocking gigabit switch.
The graph to the left shows the output of a cluster running a standard Drupal configuration on a multi-server cluster with a story from that site linked at the top of the Drudge Report. It peaked at 42 Mbit/sec, limited by the cluster's performance.
Zubr Communications Managed Clusters
The real cause of the problem is the non-scalable, if not to say brain-dead, design of the content management and portal systems (especially Drupal, Joomla and Mambo). But complaining about the inefficiencies of these systems won't solve the problem. What does solve the problem is the proper design of the cluster, tuning, and a set
of clever patches to mitigate the disastrous choices of software developers -- as the live traffic of the same site running on Zubr Communications Managed Cluster demonstrates.
| CPU | RAM | |
|---|---|---|
| Web server | 2x Dual Core | 8 GB |
| Database server | 2x Dual Core | 16 GB |
All servers are connected via single gigabit ethernet to a non-blocking gigabit switch.
As the graph clearly illustrates Zubr Communications Managed Cluster had no issues pushing over 220 Mbit/sec. That's two hundred twenty Mbit/sec of dynamically generated Drupal pages - nearly five times the performance on half of the hardware. Even better, the aggregate traffic was limited not by the cluster but only by the remote browsers.
In fact, with Zubr Communications know-how the same server hardware that in simulated tests was peaking and crashing at 44 Mbit/sec of dynamic content delivers over 880 Mbit/sec - that's twenty times the performance!
So, unless you enjoy wasting money on non-scalable solutions designed by incompetent service providers, contact our Sales Consultants.


