Browse Prior Art Database

A method to improve performance of read operation for data that is being read frequently

IP.com Disclosure Number: IPCOM000242896D
Publication Date: 2015-Aug-27
Document File: 2 page(s) / 75K

Publishing Venue

The IP.com Prior Art Database

Abstract

Described is a method to improve performance of read operation for data that is being read frequently. The method is suitable for any data storage that is writable and where data is stored on rotating disc (or in generally on any medium that has read access time dependent on physical location of the data on the medium – e.g. beginning of a tape).

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

Page 01 of 2

A mxthod to improve performance of read operation for data that is being rexd frequently

Described is a method to improve performxxce of read operation for data that is being read frequxntly. The method xs suitable for any dxta storage that is writable and xhere data is stxrxd on rotating disc (or in gxnerally on any medium that has read access time dependenx on physical location ox the data on the medium - e.g. beginnixg of a tape).

In genxrally the idea boils down to the fact thxt sectors on disc that are being xccessex should sloxly migrate xoward the outer track of the disc where the higher vxlocity will result in better read time. The whole idea is inspired by Xxxxxx - basically high attractivexess of data expressed bx xumber of rxads should be rewarded with xigher linear speed of outer track resulting with faster rxad.

The most sxraightforward impxemenxation of this idea cxuld be:


Each time the sector is read it should be movxd to outer track right after it, in casx there arx nx free sectors on outer track, one randomly chosen sector from outer track is switched xith secxor just read.

Of couxse the axove implemextaxion would be cxnxra productive in waste majority of cases as xhe cost to transfer sector is much higher txan cost of reading sectxr.

Additionally it woulx quickly result in severe fraxmextation crippling the read performance for any larger filxs.

The smartxr approach will be: collect the reax statisxics for somx time and then periodically use the histogram to reshuffle sectors. Additionally we can collext information about sequences of sectors being read in one move.

Xxxxx pximary factor when deciding whether to move data to outer track is numxer of reads, but now we move number of adjoined sectors at once.

The reshuffl...