Browse Prior Art Database

Enhanced loop balancing mechanism on cascaded enclosure architectured storage

IP.com Disclosure Number: IPCOM000215126D
Publication Date: 2012-Feb-21
Document File: 7 page(s) / 124K

Publishing Venue

The IP.com Prior Art Database

Abstract

This article proposes a method which balances the I/O to the backend storage disks, thus allowing improved performance and reliability in storage services. The method takes various levels of logical or physical configuration into consideration when doing I/O balancing and ensures balance can be reached on all levels not only at initial storage install, but also at machine running time despite of random slot disk hot spare taken. Experiments show significant performance improvement when I/O is completely balanced against the biased.

This text was extracted from a PDF file.
This is the abbreviated version, containing approximately 47% of the total text.

Page 01 of 7

Enhanced loop balancing mechanism on cascaded enclosure architectured storage

In redundant design on storage products, each drive has two ports (addressed as port A and port B in this article) that can service I/O, one of which is primary route for I/O whereas the other the secondary. This article is to address an enhanced method for primary route selection for all drives so that not only I/O balance is ensured and dynamically adjusted at customer usage level, but also high RAS(reliability, availability and serviceability) is reached by preventing loss of access to customer data.

  Currently each storage product vendor implements their own I/O routing method wherever applicable. For example, some storage product ensures I/O route is balanced among all disks under a storage subsystem, i.e. the number of port A as primary route is equal to the number of port B as primary. Since the data unit that customers access is volume, which is most often evenly distributed among all disks in an array, current method does not make sure that customer's access to volume is balanced on two redundant routes to the drives. Also under traditional method, there is probability that multiple drives in the same array use the same port as primary, which will cause loss of access potentially when the primary port route is put into service mode. An enhanced I/O balancing method is needed on storage product to improve performance and reduce system level impact.

This article aims to suggest an I/O route balancing method at backend storage. This method dynamically monitors current balancing table of drives and adjusts the table. It takes into consideration the balancing at three levels, the array level, the enclosure level and the device adapter pair's loop level and generates an optimized balancing table for I/O service. The method can be applied to any storage architecture that involves cascaded disk enclosures and RAID concept as depicted in figure 1. The optimized balancing table should meet balancing requirements on various layers as below:


(1) Considering that customer I/O is evenly distributed on disks in an array, the disks in an array should be divided equally between two routes, also called RAID level of balancing.


(2) The disks in an enclosure should be equally balanced between two routes, also called enclosure level of balancing.


(3) The disks under the loop of a device adapter pair(DA pair) would better to be divided equally between two routes, also called loop level of balancing.

1


Page 02 of 7

Figure 1 Cascaded disk enclosure architecture and initial array position

Every time there is a device added or removed, loop algorithm depicted in figure 2 should be run. The algorithm depicted selects primary route for all disks in the cascaded enclosures in turn. The assignment rule being checked against follows the priority order of array followed by enclosure followed by loop. Whenever the rule is not met during checking on any level, the route having f...