About
In June of 1988 Dan Bikle graduated from Caltech with a BSEE.
A year before that he had started working at Oracle Corporation as an engineer.
A year after graduation, he left Oracle and started a consulting business to help early adopter customers implement ERP systems based on both Oracle Financials and Oracle Manufacturing which were both new and immature products. It was a perfect time to start that consulting business.
During the 1990s Dan worked at a wide variety of clients in the Bay Area who needed help migrating off of legacy systems (such as ManMan) to Oracle Applications.
In 1998 Dan started taking on clients who were building large database backed websites such as Lycos' MailCity.
The web started to dominate Dan's consulting business. In 2001 he finished his last Oracle Applications engagement and focused on database backed websites.
In 2003 Dan took on a full time job in the Sun IForce lab to connect with engineers and early adopter customers who wanted to use low-cost hardware to build clustered Oracle databases to back large websites. The timing was good. Oracle Corporation delivered cache-fusion in Oracle 9i (and enhanced in 10g) which made horizontal scalability possible in the database tier of an application.
Another big event was delivery of the low-cost 64 bit Opteron CPU from AMD. Clustered databases powered by low cost Opteron systems offered serious competition to expensive, vertically scaled machines from Sun, IBM, and HP. In 2005 Dan left Sun to take advantage of that opportunity.
Also in 2005, a large change was delivered to the world of web application development: Ruby on Rails (RoR).
Before RoR, Java based systems were the only good option available to an implementor of a large database backed website.
RoR hits the sweet spot of the constraint-configurabilty-tradeoff.
Because of this Dan has dedicated most of his recent efforts to building RoR based websites.
The RoR sweet spot is so sweet that a moderately proficient programmer can build a large feature rich database backed website by himself in 3 to 6 months.
This means that all of the communication and training based overhead associated with an equivalent team writing a Java based website can be swept aside.
RoR can deliver enormous development savings because all project based communication can reside in one developer's mind.
If you need to build a large database backed website, and you are willing to try RoR instead of J2EE, Dan can help.
Also, RoR can be deployed in a modular way. A legacy site can be migrated to RoR bit-by-bit while the two applications run simultaneously.
The end-user would see just one site while this is happening.
RoR plays well with Java. In fact you can run Java Classes inside of Ruby and vice-versa using JRuby.
Dan's methodology to building a large site is simple:
1. Obtain / Derive a list of functions the site will deliver
2. Obtain / Derive a list of Model classes for the application (Typically these are nouns which jump out from function descriptions)
3. Decide how the Model classes are related (One-Many, Many-Many, One-One)
4. Look for Data Structures in Models (Tree, Nested Set, List)
5. Code up the model classes
6. Code up the physical database structure and then run that code
7. Fill up the database with fixture data so the developer can "see" it
8. Build a web1.0 prototype which looks something like Craigslist (Also you could get a jump-start on step 12 here)
9. Iterate with the users over the prototype
10. Drape a pretty skin over the prototype and stop calling it a prototype
11. Iterate with the users over the site
12. Code up unit tests, functional tests, integration tests, and load tests.
13. Leverage off the testing code to setup monitoring
14. Swap out the development database with a clustered database infrastructure
15. Deploy the RoR code to a production infrastructure
16. Turn on monitoring, production level backups, and disaster recovery mechanisms
17. Run all the tests written in step 12.
18. Fix bottlenecks found by load tests.
19. Test disaster recovery
20. Go Live!
