Search NASASearch

Engineering topics

Lifka, Dave

Publications and source records attributed to Lifka, Dave.

A proof-of-concept implementation of persistence in a hierarchical storage system

The concept of providing transparent access to a collection of files in a mass storage system is a familiar one. The goal of this project was to investigate the feasibility of providing similar access to a collection of persistent, complex objects. We describe an architecture for interfacing a persistent store of complex objects to a hierarchical storage system. Persistent object stores support the uniform creation, storage, and access of complex objects, regardless of their lifetimes. In other words, a mechanism is provided so that persistent objects outlive the processes which create them and can be accessed in a uniform manner by other processes. We validated this architecture by implementing a proof-of-concept system and testing the system on two stores of data. These tests indicate that this architecture supports the creation, storage, and access of very large persistent object stores.

Grossman, Robert

Analyzing high energy physics data using database computing: Preliminary report

A proof of concept system is described for analyzing high energy physics (HEP) data using data base computing. The system is designed to scale up to the size required for HEP experiments at the Superconducting SuperCollider (SSC) lab. These experiments will require collecting and analyzing approximately 10 to 100 million 'events' per year during proton colliding beam collisions. Each 'event' consists of a set of vectors with a total length of approx. one megabyte. This represents an increase of approx. 2 to 3 orders of magnitude in the amount of data accumulated by present HEP experiments. The system is called the HEPDBC System (High Energy Physics Database Computing System). At present, the Mark 0 HEPDBC System is completed, and can produce analysis of HEP experimental data approx. an order of magnitude faster than current production software on data sets of approx. 1 GB. The Mark 1 HEPDBC System is currently undergoing testing and is designed to analyze data sets 10 to 100 times larger.

Baden, Andrew