Browse Prior Art Database

System and method to integrate load balance mechanism of heterogeneous big data analysis platform

IP.com Disclosure Number: IPCOM000235543D
Publication Date: 2014-Mar-07
Document File: 4 page(s) / 110K

Publishing Venue

The IP.com Prior Art Database

Abstract

The disclosure delivered a practical mechanism and system to make online analyzing telecom or likely big data and cooperate online and offline computation work; and it enables online analyzing have concurrency, accuracy, real-time response abilities. It consists of two parts: 1) pattern based online volume data analyzing supporting tier, including a) Subscriber classification; b) online incremental analyzing updating mechanism; c) pattern based online offline cluster resource management; and 2) pattern based online offline cluster resource management

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 58% of the total text.

Page 01 of 4

System and method to integrate load balance mechanism of heterogeneous big data analysis platform

Background


Great effort has been done on analyzing Telecom data such as CDR data, 2G/3G data, Wi-Fi, and etc in both industry and academic area. Telecom data always has the following characteristics:
-Normal high concurrency with random much higher concurrency
-straight forward insights are limited
-Since person's behavior is changing with time, incremental analysis can't base on a constant basic pre-processing result

Problems and Key Findings


Big data analytics face these common challenges:
Waving capacity of incoming data
The top and bottom of incoming data is hard to predict
Mixture of various response time requirement
Depending on how the end user make use of the analytics system
Complexity of updating analytics model online

In telecom domain

In telecom domain,

,, subscribers can be categorized by clustering the mobile data

subscribers can be categorized by clustering the mobile data


By mining mobile data (Position and Social network), we are able to find these subscriber groups below: Daily-grinder
etc...


- These groups intrinsically holds constant capacity trend


- These groups are of explanable meanings that could help fix analytics algorithm and response requirement

Discovered Groups of Telecom Subscribers:

1



Page 02 of 4

Problem statement


Current times most of telecom data analyzing are offline analyzing because of 2 reasons:


1)the data volume

2



Page 03 of 4


2)normal...