| Highwinds Software Product Documentation: Typhoon, Twister, Tornado, Hurricane, and Cyclone | ||
|---|---|---|
| <<< Previous | Chapter 13. What's New in the Tornado Overview System | Next >>> |
The new Overview system is divided into three components:
An Indexing System
Raw Header Storage
Raw Header Indexing
Each component has a file type associated with it:
o files (Overview Index)
h files (Headers)
track file
Overview files contain vital indexing information about each article including its location on disk. o files are created as needed in the directories given by the OverviewPaths directory.
Chunk files contain actual overview data. h files are created in the directories given by the HeaderPaths directory. They start out at 64k, and can grow to 2GB. Data is stored in a format that is ready to stream to a client so they are self caching.
The numbers are in the format hXXYY where XX is the chunk size, and YY is the file series.
For example:
h0101 means chunk size 1, file 1 h0102 chunk size 1, file 2 h0103 chunk size 1, file 3 h0201 chunk size 2, file 1 |
Lower chunk sizes fill in before upper ones, and the 3 files are utilized more or less evenly. If all three of a class fill then a new one will be created. Splitting these files up into as many different file systems as possible is desirable and simple with the OverviewPath directive.
Files can be moved (so long as the new OverviewPath contains the new path), but cannot be misplaced ( i.e. A file system mount failure) or the results will be total overview loss!
The track file knits the h* files together, it knows where all the headers are for a given group and is very important. If the track file is lost, all overview information is lost.
| <<< Previous | Home | Next >>> |
| Faster Design | Up | I/O Optimization |