finding the cost of data compression

1
votes

What would be more costly to operate, a compressed website or uncompressed? Consider the example of Wikipedia:

Compressed size: 280GB
Uncompressed size: 5TB

Consider we have all data in a database, then to serve the web pages to the users you would need:

If you store the data in compressed form:

  • Query the corresponding record from the database
  • Uncompress the record
  • Serve the data to the user's browser

If you store the data in uncompressed form:

  • Query the corresponding record from the database
  • Serve the data to the user's browser

Compression adds one step more into the process. If we assume the browser doesn't support compression and leave out bandwidth costs, how does this impact on the operation in terms of hardware costs? To uncompress the record you would need more RAM, and more CPU instructions being executed. If you have many requests you would need to have more application servers to uncompress the data, your operation costs rise.

Is there any (case) study about compression costs I could read?

1
webcompression
I found this wonderful article that explains very well the cost of compression: devcentral.f5.com/weblogs/macvittie/archive/2009/05/27/… - Nulik
Is your data static? Pre-compression of data should also be possible and would not impact real-time performance. - Toymakerii

1 Answers

0
votes

Generally databases already solved this compressed/uncompressed problem, and allow you to configure the compression as needed, depending on how much you would sacrifice modification of data. Generally looking over large solutions - storing compressed pre-cached responses is what happens in RAM-based cache solutions, this is most cost-effective.

In Wikipedia architecture, "Text is compressed and only revisions between articles are stored".