Browse Prior Art Database

Enable high load scalability with cluster adaptor

IP.com Disclosure Number: IPCOM000246496D
Publication Date: 2016-Jun-13
Document File: 3 page(s) / 75K

Publishing Venue

The IP.com Prior Art Database

Abstract

There have been some ways to adopt normal software in the clusters to provide high availability. However in practice load balance draws more concerns, which is missing in the cluster of normal software. We present a new system of realising load balance. Our system uses a simple agent to convert the normal software to cluster-managed software with load metrics information. The cluster manager can monitor the normal software just like the cluster-specific software. We suggest to use a container or a virtual machine (VM) to contain the normal software. Our system can be based on open source, easier to implement than the expensive commercial software.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 55% of the total text.

Page 01 of 3

Enable high load scalability with cluster adaptor

Non-scalable applications are found in abundance. For examples, many desktop applications reside on single machines. Normally they only service one single user at a time. There are also many non-clustered applications working on servers and servicing multiple clients at the same time. Due to the limitation of hardware and software of individual server, the scale is only for very limited concurrent users. Simple methods of managing the existing non-scaled systems for high load scalability is highly desirable, especially for reusing the legacy software and hardware.

    Our suggested clustered architecture includes a new adaptor for each non-clustered (old) application. There are no changes to old software and the generic clustered servers. The new adaptor work between the generic cluster server and the old software. The old software directly talks to the adaptor which attaches to the cluster member agent for the communication with cluster controller or workload manager. The old software is transparent to the clustered server. The clustered server only see the adaptor as a normal application in clustered members. The old software can only see the adaptor and work for it, as a normal client. With the presented idea, the codes of both the old software and the clustered server/platform can be reused. The change complexity is contained within the new adaptor. The overall change cost and risks can be contained within the adaptor as

well. Furthermore the adaptor can potentially be reused for various applications,

which will further drive down the cost.

    The presented invention is based on the generic application server clusters for easy implementation. The following is a typical clusters


Page 02 of 3

    In above figure, the Cluster Domain Manager controls two cluster nodes. The clustered application instances run inside the cluster nodes. Cluster Domain Manager controls the work of clustered application via the agent. The agents not only pass the commands of Cluster Domain Manager to the applications, but also report the status of the nodes to the Cluster Domain Manager. Any clients send service request to HTTP Server and HTTP Server routs the request to a clustered application on a node. The response of the service is sent back to the client via HTTP Server as well. Aim for the load balance, HTTP Server is the key for the scalability of the clustered application.

    Unlike the clustered application, the normal non-clustered applications are not purposely built for the specific cluster environment. For example, some legacy software were implemented by C and run in DOS environment. It can not be directly f...