Scalable Lisp applications

The focus of Franz's research and development efforts over the past few years has been storing billions of objects in AllegroCache and billions of RDF triples in AllegroGraph. In the process we had to invent and create a lot of new and interesting technology in Lisp. We implemented new types of B+trees, efficient marshalling, resourcing, new types of hash tables, efficient memory copies, new data type dependent sorts, cluster computing, and better support for memory mapped files. A number of Common Lisp function implementations were improved to be almost completely cons free. In the first part of my presentation I will discuss the scalability issues with AllegroGraph, and give an overview of the new technology and techniques that we had to add to make these products successful. In part two, I will argue that some of this technology should lead the way as part of a new standard for Common Lisp. For those interested, Duane Rettig will be holding a tutorial session that will be complimentary to my presentation detailing in-depth treatment of some of these new techniques.