Dismiss
InnovationQ will be updated on Sunday, Oct. 22, from 10am ET - noon. You may experience brief service interruptions during that time.
Browse Prior Art Database

Dynamic Load Balancing Algorithm for Fast Query Processing in NoSQL Database Management Systems

IP.com Disclosure Number: IPCOM000248177D
Publication Date: 2016-Nov-04
Document File: 5 page(s) / 128K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is a dynamic load-balancing algorithm for fast query processing in Not only Structured Query Language (NoSQL) DataBase Management Systems (DBMSs), which requires that the nodes handling load (data nodes) participate in the balancing of load. This dynamic load balancing algorithm for NoSQL DBMSs looks at the current status of all the data nodes, and then, based on that, decides how/when to re-distribute the load between data nodes to improve the performance.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 31% of the total text.

Page 01 of 5

Dynamic Load Balancing Algorithm for Fast Query Processing in NoSQL Database Management Systems

Load balancing is an important topic in Not only Structured Query Language (NoSQL) DataBase Management Systems (DBMS). Such databases are usually run on a Cloud environment, where each physical node can host many Virtual Machines (VMs) or be a bare metal node, or a local cluster that can have many physical nodes. In both environments, when the number of users and/or queries increases, the performance of the load balancer can highly affect the total system throughput ; that is, a good load balancer attempts give every node an equal loading and in particular avoid overloading one node while other nodes have unused capacity. Moreover, in many NoSQL DBMSs, input data is replicated between multiple data nodes for backup/fault recovery and performance reasons. However, having multiple copies of the same data, a smart load balancer can benefit from this feature to precisely balance the load.

While there work exists related to designing load-balancers for different systems [1, 2, 3] no decent work considers NoSQL DBMSs' features/requirements to design load balancers.

The novel solution is a dynamic load-balancing algorithm for fast query processing in NoSQL DBMSs.

Unlike current load balancers, this approach requires that the nodes handling load (data nodes) participate in the balancing of load. This is important because existing load balancers have data nodes participate only to the extent of providing some loading data that guides the load balancer in where to send future requests.

This solution introduces a dynamic load-balancing algorithm for NoSQL DBMSs that can look at the status of all the data nodes , and then, based on that, decide how/when to re-distribute the load between data nodes to improve the performance. This is important because all the load balancers that have been used for NoSQL DBMSs use simple algorithms such as Round Robin (RR) and do not make decisions based on the dynamic behavior of the system .

The load balancer considers:


1. The central processing unit (CPU) utilization on each node


2. The amount of communication between different nodes within the data center (Cloud environment) or local cluster


3. The storage performance on all data nodes for load balancing

This is in contrast with all previous techniques where either one or two of these features had been used to design the load balancer.

1


Page 02 of 5

The new load balancer accommodates the dynamic addition of new data nodes to the system at run time , if administrators determine that a specific node is very busy in terms of both computation and communication . This is a unique feature for NoSQL DBMSs that has not been addressed in the literature because of the cost of dynamically adding nodes in traditional systems .

The load balancer is a good place for auto-scaling, which is very common in cloud. As the load balancer is already doing the same monitoring, it can...