Why Alacritech Changed Directions By Larry Boucher On January 10, 2011

While it’s not uncommon for companies that reinvent themselves to change their names, it was a course that we weren’t prepared to take with Alacritech. The reason is simple. We didn’t reinvent the company. Instead we changed focus. We’ve packaged our acceleration technology into a system as opposed to solely selling TCP/IP Offload network adapters as standalone products. There were a couple of reasons for this change in focus..

Without overly getting into the technical details, let me share why the previous path we were on wasn’t going to take us very far and the need for a course correction. While Alacritech continued to make advancements in Dynamic TCP/IP Offload Engine (TOE) technology, we focused exclusively on the Microsoft Windows market. The goal was to help customers get the most out of their Windows servers by reducing the processor and memory burden in processing TCP/IP. The benefit in doing so is that customers can get better performance and scale their servers to new heights. While we found customers that had a problem to be solved in the Windows space, they were few and far between. The success of VMware is a testimony that the average Windows customer is  not concerned with decreasing CPU utilization. Instead, their concern has been about getting better utilization from servers, i.e. increasing utilization. VMware and Alacritech were at polar opposites and it became clear given VMware’s growth that the average Windows server was drastically underutilized. Does this mean that our Dynamic TOE technology wasn’t worth anything?  This was and continues to be hardly the case. Let me explain why.

Dynamic TOE works very well when all of the CPU cores in a server are being utilized. We spent countless hours analyzing CPU performance in Windows-based servers and the simple fact is that CPU cycles go to waste. Windows servers are generally purchased for a single purpose, usually departmental in nature.  When another need arises, rather than centralizing the server and adding applications to it, another server was purchased for the new application.  When another department wished to deploy an application, rather than utilize another department’s server a new server was purchased for that same application.  As a result Windows servers proliferated, each running a single application.  This significant waste of resources has been substantially reduced by employing virtualization to aggregate many of these servers on a single server.  While this added more overhead than running multiple applications under Windows, it was deemed more secure, and provided more flexibility.  Because the individual servers were so significantly underutilized, the overhead of virtualization still allowed as many as twenty or more servers to be reduced to a single server.  When the server was utilized at this low level removing the TCP overhead had no effect on system performance.

The question became how to overcome this problem. We were absolutely convinced that Dynamic TOE would work extremely well if designed into a system where we controlled the variables. We elected to use parts of FreeBSD to design a purpose-built appliance that could take advantage of all of our underlying acceleration technology. And we built new technology on top of that, designed to accelerate NFS in addition to TCP. We could see a real customer problem that needed to be solved and realized that we had the underlying technology that could attack it. Construction of a purpose-built appliance designed to be a performance layer for NAS became our mission. This appliance, which we call the ANX 1500 is much more than DRAM, SSDs, and a UI.  The real differentiation that makes this product so flexible and scalable is the underlying Alacritech acceleration technology. The ANX 1500 works very efficiently, is very scalable, and takes a tremendous burden off NAS servers. Simply put, it accelerates NFS operations, delivering very high throughput and very low latency. One of our early adopters affectionately referred to the ANX as his “NFS Offload Engine”. It’s great when customers help with marketing. I encourage you to look over our redesigned web site and learn more about the ANX.

We’re excited about this change in focus. We sincerely appreciate the business and relationships forged with Alacritech’s older customers and look forward to building relationships with new ones. While we hope that in the future virtualization will bring the Windows server utilization up to the point where we can be of help, in the NFS world we can make a real difference today.