2023- 2022
- 2021
- 2020
- 2019
- 2018
- 2017
- 2016
- 2015
- 2014
- 2013
- 2012
- 2011
- 2010
- 2009
2008
Polyglot 12: R- Air Compressor
- Stanford CS145
- Retina
- Electoral College
- Obama vs McCain
- Perpetual Motion
- Book Safe
- Cráneo de Cristal
- Back to School
- Artificial Neuron
- Search Appliance
- Orthogonal Viewer
- Porcupine
- Bitmaps in JavaScript
- My Brain
- Domino Logic
- Faster is Better
- Transfer
- Car Fire
- Helsinki
- Jessa Gamble
- Maker Faire 2008
- Downhill Walker
- Youngme / Nowme
- Python Parallel Port
- MobWrite 2
- Class Clown
- Mask
- Progress
- Angel Island
- Scapegoat
- Localhost
- Disneyland
- Technical Difficulties
- Wooden Brain
- 2007
- 2006
- 2005
- 2004
- 2003
- 2002
Search Appliance
13 September 2008
While working in Scotland at Digital Routes, we had a client that needed a search solution for a large number of confidential documents. The first implementation was a disaster: the programmer didn't know what a hash table was and created a buggy O(n3) mess. My rewrite was somewhat more successful, it continues to run stably and scalably six years later. But eventually the client's requirements changed. Now they need to index PDF and Word documents in real-time. Solution: Google's Mini Search Appliance.
![[Google Search Appliance box]](/news/2008/gsa-box.jpg)
For some inexplicable reason, Digital Routes' hosting provider is unable or unwilling to enter the IP, subnet mask and gateway information into the appliance's web-based setup. The work-around is to ship the new appliance to me in California so that I can configure it then ship it on to the host in Quebec.
![[Google Search Appliance server]](/news/2008/gsa-server.jpg)
The Google Mini is certainly a beautiful machine. The instructions are straight-forward, the cables are colour-coded, and the setup process is well thought out. Then I pushed the power button. Good grief that thing is loud. The cooling fan on that little box sounds like a jet engine. I was half expecting to see a stack of Mach diamonds appear out the back vent. Belatedly I found this note buried in the GSA documentation:
"If you install the search appliance in an office, place it in an area where any noise produced by the cooling fan in the search appliance will not be disturbing."
![[Broken images on Google Search Appliance]](/news/2008/gsa-404.gif)
The admin screens are also uncharacteristically clunky. The main menu has twelve broken image icons on it. Another page drops some (but not all) of your form data if any field is invalid. Using the back link to edit a previous screen takes 82 seconds to process. The user experience is certainly sub-optimal.
Since I happen to be a Google engineer, the GSA source code is available to me. So I've taken the opportunity to submit several patches (for the stuff I can fix) and file a bug report (for the stuff I can't). It's great to be at a company where if you see something wrong you can just reach in and fix it, even if it is not your department. I trust that future users of the appliance will have a better experience. In the mean time, be sure to order some ear plugs when ordering your Google Search Appliance.
This website is my personal blog. The opinions expressed here are my own, and not my employer's. This normally goes without saying, but this post is a bit more pointed than usual. Also note that this post only addresses the one-time network configuration, not the (more important) crawl and search results.
