Browse Prior Art Database

A Method of Balancing Workload I/O and Space Utilization of File Systems in Tandem

IP.com Disclosure Number: IPCOM000243718D
Publication Date: 2015-Oct-15
Document File: 2 page(s) / 39K

Publishing Venue

The IP.com Prior Art Database

Abstract

Disclosed is an algorithm, applicable to a file system, that directly translates an internal number into a file name, avoiding excessive width and depth to minimize lookups and maximize performance.

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

Page 01 of 2

A Method of Balancing Workload I/O and Space Utilization of File Systems in Tandem

Some applications require the ability to store thousands of files in a file system. In many cases, a database or lookup table is required to translate an internal representation of the file into the actual file name in the directory. For example, internally, a file may be assigned ID #1034 and, externally, this file is stored in /abc/def/xyz.dat. In order to retrieve the contents of this file, the application needs a translation between 1034 and the file name. This can be avoided if the user builds the ID number into the file name. For example, instead of calling the file /abc/def/xyz.dat, maybe it is named /abc/def/1034.dat.

The problem here is that there are now thousands of files in a single directory. On most operating systems, when that

many objects are saved to a single directory, the lookup times to find a file grow dramatically, thus reducing performance. To alleviate this, the files must be distributed across multiple directories to improve lookup times. However, the user does not know to which directory 1034 belongs, without having a lookup table that points to the directory.

The objective is to eliminate this lookup; therefore, again, the user can turn to the name of the file and build directories based on that. For example, 1034 may be stored in a 64-bit number internally as 0x000000000000040A. The user can create a series of subdirectories 00/00/00/00/00/00/04/0A.dat. This allows a user to easily identify a unique directory based on the file name, thus avoiding thousands of files in a single directory. However, now every single file becomes buried seven directories deep.

A hybrid algorithm is needed that allows the user to avoid excessive width in the file system (i.e. number of files in a single directory) while also avoiding excessive depth (i.e. number of directories deep to reach file).

The novel contribution is an algorithm that directly translates an internal number into a file name, avoiding excessive

widt...