Workloads Suited for Calxeda’s Architecture
Applications such as scalable analytics (e.g. Hadoop), static web hosting, lightweight web applications, applications written in Java or other interpretive languages are all good candidates. These applications workloads typically have a set of common characteristics such as:
- Ability to scale out efficiently with multiple processors
- Have relatively small memory requirements (< 4GB per process)
- Have relatively light computational requirements per thread
- Have high memory, I/O, and networking bandwidth requirements relative to CPU
- Typically run at low utilization on standard processors
Applications designed to scale out with a high degree of parallelism – like Hadoop, MapReduce or other “Big Data” applications – experience optimized performance and utilization.
Move your web-apps from underutilized x86-servers to a Calxeda-based cluster to realize instant energy savings.
For shared-nothing, stateless, middle-tier servers, running distributed caching applications – like Memcached – bring the elasticity of the cloud into your data center to scale on demand.
Some specific applications that exhibit these traits include:
- Map-Reduce Applications
- Simple database searches on large in-memory databases
- Video Servers
- Big-Data-oriented search systems
- HPC workloads that have high I/O requirements per computational load and that scale out well, such as Seismic processing.
- Financial Modeling for trading and risk analysis systems
There are probably many more we will uncover as we work with partners, universities, and leading software developers around the world. Let us know what you think! Send us your ideas! Become a Trailblazer and hop on board!