Virtualize When It Works

Virtualization is justifiably popular as a great way to reduce server sprawl and minimize datacenter operating costs. Virtualization delivers much better utilization of your servers, sharing them among many applications and users. Virtualization is a powerful enabler of cloud computing, which has many extremely appealing economic and operational advantages of its own (such as elasticity and sheer ease of deployment without having your own hardware).

Virtualization works well when each of the VMs (virtual machines) running an application consumes just a small part of the available hardware resources (processor cycles, memory, I/O operations, and network bandwidth). But some applications "pin the needle" on CPU, memory and I/O usage. Such applications — which often include high-performance SQL databases and high-performance caching software — are painful to run in a VM. Relying on VMs to do this heavy lifting can really hurt. Users sometimes have to shard their data by a factor of 10x or even 100x compared to what they could get from a single properly-equipped dedicated server.

VMs are just too slow to be the most economical choice for database and caching servers meant for production use in performance-sensitive applications. You can end up with poor performance at a much higher cost. High-performance apps require special handling if you want the lowest total cost of ownership.

Reduce Your Datacenter Footprint by a Factor of Ten

Use SchoonerSQL™ and Membrain™ to get dramatic server consolidation with superb performance and highest availability. Each of these Schooner software products efficiently balances the use of all resources on standard x86 servers to maximum utilization and therefore maximum efficiency.

For example, the high performance of SchoonerSQL enables consolidation of multiple traditional MySQL masters or slaves onto a single Schooner-powered server, as shown in Figures 1 and 2. Its high throughput enables one server to assume the workload of many of the same master or slave nodes that don't use SchoonerSQL. It also provides headroom, reducing the need for application changes for sharding.

In short, virtualization is great, but sometimes it's not the right solution to the most acute scaling problems for your SQL database, software cache, or NoSQL data store. Schooner is. Because Schooner lets you handle way bigger loads with way fewer nodes.

Figure 1: Consolidation of multiple masters into a single Schooner-powered x86 server

Figure 2: Consolidation of multiple slaves into a single Schooner-powered x86 server