Browse Prior Art Database

A Method and System for Uniformly Distributing Incoming Tasks to Multiple Hosts

IP.com Disclosure Number: IPCOM000236064D
Publication Date: 2014-Apr-04
Document File: 3 page(s) / 83K

Publishing Venue

The IP.com Prior Art Database

Related People

Madhukar Mulpuri: INVENTOR

Abstract

A method and system is disclosed for uniformly distributing multiple incoming tasks to multiple hosts, by allocating the multiple incoming tasks to the multiple hosts arranged in a random order. Further, the method and system utilizes a distributor for allocating in round robin fashion the multiple incoming tasks to the multiple hosts arranged in a random order. The method and system also involves using the distributor for breaking the multiple incoming tasks into multiple uniform sized chunks before allocation.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 53% of the total text.

A Method and System for Uniformly Distributing Incoming Tasks to Multiple Hosts

Abstract

A method and system is disclosed for uniformly distributing multiple incoming tasks to multiple hosts, by allocating the multiple incoming tasks to the multiple hosts arranged in a random order.  Further, the method and system utilizes a distributor for allocating in round robin fashion the multiple incoming tasks to the multiple hosts arranged in a random order.  The method and system also involves using the distributor for breaking the multiple incoming tasks into multiple uniform sized chunks before allocation.

Description

Currently, in a parallel processing system, a process for allocating tasks to multiple hosts relies primarily on a distributor which queries the current state of the system.  Each of the multiple hosts updates a storage system with a load on the host.  Subsequently, when an incoming task is received at the distributor, the storage system is queried for identifying the load on each host.  Depending on the total number of incoming tasks received, and the load on each host, the distributor allocates the incoming tasks to the multiple hosts.  An overview of the process is illustrated below in fig. 1.

Figure 1

One of the disadvantages of the aforementioned process for allocating tasks is that the storage system needs to be queried for each incoming task.  Further, as the frequency of the incoming task increases the storage system will be queried a lot.  If frequency increases too much, then the storage system will become a bottleneck for the system performance.  In effect the simplest way to overcome the bottleneck on the storage system is to remove the storage system from the distributor.  This can be done by distributing the chunk of incoming tasks to the hosts in a round robin fashion.  This has an issue that some hosts can be overloaded based number of tasks in the chunk.  Consider the total number of incoming tasks to be tN and the total number of hosts to be tH.  In accordance with the scenario, the first set of hosts i.e. tN mod tH, is likely to be overloaded.  For example if tN is 23 and tH is 10, then the first set of hosts 1 to 3 will be overloaded.  

Disclosed is a method and system for uniformly distributing multiple incoming tasks to multiple hosts uniformly, by allocating the multiple incoming tasks in a round robin fashion to the multiple hosts arranged in a random order. 

In an embodiment, the method and system utilizes the distributor for uniformly distributing the multiple incoming tasks.  Accordingly, the distributor is configured to...