The idea behind the Image Set concept is to provide the PicoCluster user with a completely configured environment so that they can get up and running quickly.
Installing an application or framework on a single machine can be difficult, particularly if one is not a systems engineer or administrator. Even something relatively simple like the Apache Web Server or Tomcat Application Server can be hard to make work if you're not very familiar with them.
The difficulty is multiplied when trying to install and work with distributed or parallel applications or frameworks. Not only must the application or framework be installed on multiple machines, it must be configured in such a way that all of the machines can communicate with and coordinate with each other.
Not only is basic networking required, but each application or framework implements their coordination and communication layers in very different ways. Some of them are natively distributed and required a specific understanding of how they operate in order to configure successfully.
Some of them rely on more general coordination frameworks like Zookeeper. That means that the coordinator needs to be set up and working properly first, then the application or framework itself needs to be setup and configured with the coordinator.
Still others rely on distributed file systems like HDFS, GPFS, GlusterFS, Lustre, and many others. Applications like Hadoop that rely on this architecture typically are layered frameworks. This adds even more complication because many layers need to be installed configured, and tested before you can even get started developing or testing.
By now you have some idea of the complexity of distributed and parallel systems. Now try doing this kind of setup and configuration for dozens of such systems.
Because of the form factor of PicoCluster, we can create pre-configured clusters for any of these applications or frameworks The PicoCluster owner can simply pop in the memory cards, turn PicoCluster on, and start working with. This entire layer of complication can be done away with.
There is no other system in existence that can do this. Think of the value in the time savings alone in being able to purchase an entirely pre-configured system!
This is just a starter list of the Application Image Sets we'll be creating. Let us know which ones you're looking for.
- Apache Hadoop
- Apache Spark
- Apache HBase
Be sure and contribute to PicoCluster Forums to help build a strong PicoCluster community!
- PicoCluster LLC