Browse Prior Art Database

Improved Multipath Routing Algorithm Based on Operation Type

IP.com Disclosure Number: IPCOM000247894D
Publication Date: 2016-Oct-10
Document File: 3 page(s) / 57K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is an improved multipath routing algorithm that selects a storage subsystem input/output (I/O) interface based on operation type, segregating reads from writes in order to improve application performance.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 51% of the total text.

Page 01 of 3

Improved Multipath Routing Algorithm Based on Operation Type

Multipath routing algorithms choose the path on which to send an input/output (I/O) from among the multiple paths between a host and a storage subsystem. A path consists of a unique source interface on the host and target interface

on the storage subsystem, along with zero or more intermediate 'hops' which are introduced by Storage Area Network (SAN) switches. Multipath routing algorithms are typically implemented in a host operating system (e.g., Linux*). Conventional multipath algorithms include round-robin, which alternates among the possible paths and the shortest

queue, which sends an I/O down the path with the smallest number of outstanding I/Os.

Existing multipath routing algorithms do not discern between read and write commands, forcing storage subsystem interfaces to process a mixture of both. In certain scenarios, such as Redundant Array of Independent Disk (RAID) rebuilds, write operations can take longer than reads to process. In other scenarios, the proportion of write operations can predominate over reads. Both cases may lead to delays in read operations being processed. This can lead to a degradation in application performance, since applications may stall, waiting for data to be read from the storage

subsystem.

The novel contribution is an improved multipath routing algorithm that selects a storage subsystem I/O interface based on operation type, segregating reads from writes in order to improve application performance.

This approach allows the user to designate a proportion of storage subsystem I/O interfaces to specifically process read or write operations. For example, assuming there are four interfaces, the user can specify that three be dedicated to processing read operations, and one be dedicated to processing write operations. In this scenario, the system prefers reads to writes.

When th...