Karmasphere Client
Regardless of the cloud type or Hadoop version used, a Hadoop MapReduce job must be created to communicate with the Hadoop implementation. This is done via a Hadoop Client, which serves as the Hadoop API for your job. Unfortunately, Hadoop clients are not interchangeable. Migrating from one Hadoop client to another requires that a MapReduce job be re-developed, debugged again and re-tested. The Karmasphere Client overcomes this problem, as well as other key issues, reducing development and administration time and complexity.
Big Data professionals are interested in putting their large data sets to work, rather than worrying about the details of working with the particular cloud implementation they have access to. The Karmasphere Client isolates the Big Data professional from the particular cloud implementation and simplifies the process of upgrading from one cloud (whether simply a version upgrade or across cloud types) to another.
There are three main classes of Hadoop cluster implementation:
|
Cloud Type |
Description |
| 1. Private Cloud | Local implementation of a Hadoop distribution (eg. direct from Apache, Cloudera, IBM, etc.) |
| 2. Public Cloud | Implementation of a Hadoop distribution on the infrastructure of a public provider (eg. Rackspace, GoGrid, SoftLayer, Amazon Web Services, etc.) |
| 3, Public Cloud Service | A service that provides a Hadoop implementation as a service (eg. Amazon Elastic MapReduce) |
The Karmaspehre client frees the Big Data professional from concerns about the particular Hadoop implementation being used as well as enabling the following benefits:
| Benefit | Karmasphere Client Support |
| Job portability | MapReduce jobs deployed with the Karmasphere Client are portable across multiple Hadoop implementations, including across different version types and vendors. |
| Operating System portability | Portable across Windows, Mac and Linux platforms. No other Hadoop client is completely portable across all these platforms. |
| Firewall Hopping | The Apache Hadoop JobClient cannot deploy MapReduce jobs to a JobTracker behind a firewall. Karmasphere Studio is now fully able to deploy, monitor and manage jobs on clusters that are protected by a firewall. |
| Fault Tolerant API | Allows the developer to build complex processes around the client without worrying about one part of the system having an adverse impact on another. |
| Synchronous and Asynchronous API | Asynchronous APIs make it easy for the developer to monitor and manage busy clusters without requiring multiple processes or threads. |
| Clean Object Oriented Design | Our easy to understand, intuitive, encapsulated and well-behaved design saves time and effort throughout your development and ongoing production environment. |
Big Data professionals who use the Karmasphere Client eliminate the need to re-develop a job when re-targeting to an upgraded Hadoop cluster, or simply migrating to a different Hadoop cluster. With the Karmasphere Client, the result is a robust deployment environment for MapReduce jobs both today and in the future. It makes it easy (and predictable) to maintain a business operation that is reliant on Hadoop.
Follow this link for more technical information about particular Hadoop Client versions.