Browse Prior Art Database

An adaptive system and method for disk management using iterative shrinking mechanism in an HA environement

IP.com Disclosure Number: IPCOM000237729D
Publication Date: 2014-Jul-08
Document File: 5 page(s) / 240K

Publishing Venue

The IP.com Prior Art Database

Abstract

The core idea of this disclosure is to perform front-end load balancing based on the content of incoming message, instead of backend server status. The traffic will first be routed to the master server of the appliance cluster (e.g. virtual IP owner). The master server perform quick parse of incoming headers to determine two things : 1) the content-type of incoming message, 2) the content-length of incoming message. Since the master server keep track of the persistence store status of each appliance in the cluster, it knows which appliance is approaching its archive threshold. In order to have appliance trigger the archive process as quick as possible , the master appliance will route the new transaction to appliance that approaching the threshold.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 52% of the total text.

Page 01 of 5

An adaptive system and method for disk management using iterative shrinking mechanism in an HA environement

    More and more enterprise customers use appliances to perform their business operation nowadays. Most appliances are equipped with persistent storage (e.g. database or file system) to maintain application/transaction data. So, when lots of traffic comes in, lots of transactional data will also be stored in persistent storage.

    In current IT environment, most customers uses cluster to handle the incoming traffic. A common problem is that when the persistent storage is full, its very likely that the persistent storage of appliances are also full because of HA architecture. As a result, the whole system could provide very limited capacity to process new transactions before most appliance are busy archiving their data at the same time.

    Even load balancer is introduced in the cluster, it still can't solve the problem. There are several algorithms that load balancer can use to choose the backend servers. They are first alive, hash, least connections, round robin and so on. They all based on which backend is alive or is the least busy, but none of them are based on how much disk space used in terms of archive purpose.

    Most of the current solution will redirect the packet to the machine with lower loading or send the packet to all machines equally, through this solution will let all machines get equally disk space. Once one machine is reaching the archiving limit, the others will have almost the...