Gregory L Davoll

Title: Gregory L. Davoll and Adaptive Data Compression

Introduction

Gregory L. Davoll is an inventor associated with Los Gatos, California, and International Business Machines Corporation. His patent work includes methods for preparing data-compression dictionaries tailored to database characteristics. He is credited with 2 patents.

Latest Patents

Davoll’s listed patent, “Method and System for Adaptively Building a Static Ziv-Lempel Dictionary,” describes creating a static dictionary for hardware-based compression of database records. The dictionary is built before compression begins, avoiding the initial inefficiencies that can occur when the Ziv-Lempel method learns while compressing data.

The invention adapts a fixed-size parse tree to the database through either of two approaches. One approach builds an oversized tree and prunes it to the desired size by removing nodes with low use counts and least-recently-used status. The other builds a tree to the target size, then recycles selected nodes as database samples are examined, choosing candidates based on use counts and recency.

After adaptation, the parse tree is converted into a static dictionary and stored in memory for use by a hardware-based compression process. The method also initializes the tree with NULL and DEFAULT sequences from database definitions before examining the data.

Career Highlights

Davoll’s listed patent activity reflects work on data compression and database systems at International Business Machines Corporation. The described invention focuses on preparing a reusable dictionary in advance, with the aim of improving compression efficiency for database records.

Collaborations

Davoll’s listed coworkers include Craig Alan Friske and Mohamed H. El-Ruby. The available information identifies them as coworkers but does not specify their roles in the patent.

Conclusion

Gregory L. Davoll’s patent work centers on adaptively constructing static Ziv-Lempel dictionaries for hardware-based database compression, using database-aware tree-building and node-selection techniques.