March 2014 Summaries

The DataStax Sandbox is a virtual machine (VM) image containing Linux-based DataStax software, including DSE, OpsCenter, DevCenter, and other utilities. It's designed to simplify getting started with DataStax software without installation or configuration. Users can download the Sandbox for free and use it indefinitely. The Sandbox is not intended for production installs and has a few prerequisites, such as 8GB of RAM. Suggestions for improvement are welcome at [email protected].

This series of posts provides advice on choosing the right Elastic Compute Cloud (EC2) instances for hosting Cassandra on Amazon Web Services (AWS). It covers selecting instance types, scaling out your cluster, using an Amazon Machine Image (AMI), handling disaster planning and recovery, and best practices. The recommended minimum node size is m1.large, with larger nodes like m1.xlarge or i2.2xlarge for more data and lower latencies. Scaling horizontally on commodity machines is preferred over scaling upwards due to Cassandra's fault-tolerance and linear scalability. The article also discusses testing methods to estimate cluster throughput and the uncertainty of Cassandra's infinite scalability, but its linear scalability has been confirmed for clusters between 100 and 1,000 nodes.

AWS offers a variety of EC2 instances with different strengths and purposes. General purpose machines like m* include the m1.* and m3.*. Memory-optimized machines like M2 are suitable for memory-intensive applications, while compute-optimized types such as c1.*, c3.*, cc2.*, and cr1.* are designed for calculation-intensive tasks. Graphical application instances include g2* and cg1.*. High storage instances (hs1.*) and high I/O instances (i2.* and hi1.*) are also available. The t1.* instance type is a low-cost option for testing applications. EC2 instances can be launched as PV or HVM, with some offering Enhanced Networking support when started as HVM inside a VPC. Choosing the right EC2 instance depends on factors such as cost and performance requirements, with m1.* being a good starting point for Cassandra clusters. As cluster size grows, consider moving to larger instances like m1.*.xlarge to improve performance and reduce costs.