Kryder's Law, a variation of Moore's Law, describes the trend "magnetic disk areal storage density doubles annually" [1]. In other words, you don't want to know how much I paid for a 40MB Seagate MFM drive in 1989, but today 1000GB drives rule the day for much less money. This increase in capacity follows a predictable trend.
Conversely to Kryder's Law and Moore's Law (which basically describes computers becoming predictably faster), Wirth's Law states software becomes larger, more complex, and slower [2]: in the end the win on one end is washed out by the loss of the other.
Let's compare these laws against OpenOffice.org to see which law wins. We'll measure the installed disk usage of OpenOffice.org for Linux in English as built by Sun Microsystems. The size of OpenOffice.org installation over time fits a linear equation with R2 = 0.858 and an expoential curve with R2 = 0.876: that means it is predictable like Kryder's Law.
Growth in the 1.x version series was slow, but the 2.x versions made up for it. The chart looks odd around versions 1.1.5 and 2.0.0 because they were released 36 days apart. The 3.0.0 beta currently also has a oddly sharp curve that will look more natural when its data point moves over to September 2008.
OpenOffice.org's disk storage usage comprises several parts. The vast majority of the space is in these subdirectories:
| Subdirectory name | Description |
|---|---|
| help | Documentation |
| preset | Configuration |
| program | OpenOffice.org executable code including a functional copy of Python |
| share | Configuration, dictionaries, icons, templates, fonts, XSLT filters, etc. |
The proportional disk usage of these directories has changed over time. The largest growing directory share has ballooned 513% from 23MB in 2002 to 141MB in 2008. In version 2.4.0, the many preinstalled dictionaries consume 83MB (59%) of the share directory. Meanwhile, the program directory has slowly increased by 52% over the same period.
One component of computing performance that has not grown as fast as others is hard disk seek times, so it's an overall win to store large indexes on a disk to speed access to large files [3]. OpenOffice.org uses this trick: indexes consume 7% of the dictionary directory.
Back to the Laws. Has the OpenOffice.org installation size indeed followed the general trend of growth in PC disk capacity? This chart, which compares the hard drive capacity to OpenOffice.org installations, indicates the answer is no. In 2002, OpenOffice.org version 1.0.1 consumes 0.45% of a then-common 40GB drive. In 2008, version 2.4.0 consumes 0.15% of a 250GB drive—common on new systems today—and even less of the monster 1TB drives available. OpenOffice.org has grown slower than predicted.
Conclusion
Though slower than Kryder's Law, OpenOffice.org's growth is still fairly predictable. The increase is positive and means a wider variety of features, more comprehensive documentation, and otherwise more of what customers want. Even today's larger OpenOffice.org installation occupies a lower percentage of a hard disk drive of 5 years ago.
OpenOffice.org performance
This article begins a series on OpenOffice.org performance. Subscribe so you don't miss anything. :)
Comments
When you talk to MS, they will tell you all about their MS Office "Family," which includes Access, Outlook, MS Publisher, Visio, and on and on...
Taking both points into consideration, I fail to see how the analysis serves as a realistic look at the situation.
Just some thoughts...
However, what I think we're seeing is a fairly tightly written program that isn't expanding that much itself, but has a large expansion as extra fonts and filters are added to it, the little extras to allow people to easily import files from other sources. Every time MS change MS Office, a whole new set of filters need to be incorporated into OO so we can open the new MS documents.
I like the OO percentage graph, but would also like to see it on an actual MB growth chart to see how the size of the help and the program has grown in reality.
justwally: On Linux OOo 2.4.0, OpenOffice.org Base is just 8,930,631 bytes. Writer is even smaller. Most of the bulk is in core packages. All these packages (Base, Writer, Calc, Math, etc.) are included in all the charts in these two articles because they are installed by default.
The hardware of the REGULAR user (ie., the majority of people) doesn't keep up with any of these laws.
This "research" is useless until it factors in the performance experience on a regular computer.
Still, this an excellent article, can you post the data?
The disk usage data is available here: Size of OpenOffice.org installed.
I'm thinking that what I was missing is contained/will be contained in additional blog entries here. It appears that this was the beginning of a series, and that is what left me scratching my head.
Am I reading that correctly?
Yes, this the first in a series of what may be about 10 performance articles. I recently posted "Is OpenOffice.org Getting Faster." Feel free to ask any questions anytime.
Andrew