Browse Prior Art Database

Feedback Optimization for Distributed System Disclosure Number: IPCOM000248214D
Publication Date: 2016-Nov-10
Document File: 3 page(s) / 83K

Publishing Venue

The Prior Art Database


Currently, there is no method for the method to interaction with distributed system during the process. We introduce a method about Feedback Optimization for Distributed System. It would improce the performance for distributed system.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 68% of the total text.

Page 01 of 3

Feedback Optimization for Distributed System

Big Data is widely used currently.

In OLAP, we have many requirements to show top N grids or top N reports. Such as:

1. In database monitoring tool, tell me top slowest 10 SQLs including their corresponding metrics: elapsed time, rows read, number of executions, CPU time, I/O time, lock Wait time, sorts, etc.

2. In e-commerce website, tell me the most popular 50 merchants and their corresponding sales data: stocks, sold numbers, amount, like votes, unlike votes, category, etc.

The sample query behind the scenario likes below:
SELECT key, sum(c1) as sum_c1, max(c2), avg(c3), sum(c4)+sum(c5), ...
FROM table1
ORDER BY sum_c1

It is a burden to calculate so much useless aggregate functions and expressions.

We would like to introduce a new method that the job tracker would collect the 10th value repeatedly and publish the value to each task tracker. The task tracker would terminate the task if the final result must higher than the 10th value.

Advantages of Our Invention

By the invention, we gain:
 Less CPU consumption
 Cheaper access path
 Better performance
The following is the detail information:


Page 02 of 3

1. Identify the qualified job

The system should identify the qualified job. It could be checked by job tracker or by the programer.

2. Job tracker send the task to each task tracker


Page 03 of 3

The job tracker would send the task to each task tracker with special marker to i...