• english
  • french
  • italian

Contact us

Contact us for more information about Jumbune, professional services and product support.

Send Message

Timely and cost-effective analytics over “Big Data” are now a key ingredient for success in many businesses, scientific and engineering disciplines, and government endeavors. The process of adjusting settings to record for memory, cores, and instances used by the system is termed tuning. Job Optimization process guarantees optimal performance and prevents resource bottlenecking. Effective changes are made to each property and settings, to ensure the correct usage of resources based on system specific setup. Several parameters along with different scheduling mechanisms have impact on the performance of the metrics. Our belief is that these parameters have to be efficiently measured for individual tasks and mapped with the scheduling policies for maximizing the performance.

Jumbune's 'Job Optimization' is a proprietary framework with an in-built cost based optimization algorithm that assists in development and tuning of the applications running on top of enterprise hadoop clusters. Recommends optimal configuration based on application, resource and cluster. It performs the balancing of cluster load and application behaviour together which results in fine tuning of the application's lifetime. It orchestrates the life-cycle of an application subject to actual workload, I/O, data size, behaviour until it finds optimal parameters that can be applied to it. The time bound optimization helps administrators, who are on strict deadlines to use the optimization framework in a fixed time frame. This makes sure that the optimization is performed during off-peak hours and it doesn't interfere with the normal job execution schedule on the cluster.

The optimization is iterative and is based on the application at hand and predicts the best configuration with respect to the application, cluster load and behavior.

There are multiple optimization profiles that aid in improved prediction for jobs of varying workload, I/O and data size. The time bound optimization helps administrators, who are on strict deadlines to use the optimization framework in a fixed time frame. This makes sure that the optimization is performed during off-peak hours and it doesn't interfere with the normal job execution schedule on the cluster.

During each passing iteration Jumbune's optimization framework makes intelligent changes to the configuration which are based on it's adaptive learning from previous iterations and stops once the optimizations have plateaued. The configuration can be exported as XML files or command line arguments. Jumbune also has an offline tuning mode that calculates optimal configuration based on the job's history. This feature the ability to improve HIVE/Pig jobs execution times. The user just has to mention the job id of the job and Jumbune would predict a better optimized configuration.

About Us
Reload

Jumbune, a product from Impetus Technologies, that helps to optimize and analyze Big Data applications running on enterprise clusters. It is built on open source and highly scalable with deep insights into performance of Hadoop applications and clusters.