Please RSVP so we could get an accurate headcount for food & drink which will be provided by Idealab.

Rough Agenda:

1. 7:00pm - 7:15pm Start to Eat & Mingle
2. 7:15pm - 7:25pm Introductions
3. 7:25pm - 7:30pm Identify those with open positions (no contingency recruiters)
4. 7:35pm - 7:37pm Speaker & Talk Introduction
5. 7:37pm+ The Talk!

Here is a synopsis of the talk that George Chang will be giving:

The Lunar Mapping and Modeling Project (LMMP) is tasked to aggregate lunar data, from the Apollo era to the latest instruments on the LRO spacecraft, into a central repository accessible by scientists and the general public. A critical function of this task is to provide users with the best solution for browsing the vast amounts of imagery available. The image files LMMP manages range from a few gigabytes to hundreds of gigabytes in size with new data arriving every day. Despite this ever-increasing amount of data, LMMP must make the data readily available in a timely manner for users to view and analyze. This is accomplished by tiling large images into smaller images using Hadoop, a distributed computing software platform implementation of the MapReduce framework, running on a small cluster of machines locally. Additionally, the software is implemented to use Amazon's Elastic Compute Cloud (EC2) facility. We also developed a hybrid solution to serve images to users by leveraging cloud storage using Amazon's Simple Storage Service (S3) for public data while keeping private information on our own data servers. By using Cloud Computing, we improve upon our local solution by reducing the need to manage our own hardware and computing infrastructure, thereby reducing costs. Further, by using a hybrid of local and cloud storage, we are able to provide data to our users more efficiently and securely. This talk examines the use of a distributed approach with Hadoop to tile images, an approach that provides significant improvements in image processing time, from hours to minutes. This talk describes the constraints imposed on the solution and the resulting techniques developed for the hybrid solution of a customized Hadoop infrastructure over local and cloud resources in managing this ever-growing data set. It examines the performance trade-offs of using the more plentiful resources of the cloud, such as those provided by S3, against the bandwidth limitations s uch use encounters with remote resources. We will outline some of the technologies employed, the reasons for their selection, the resulting performance metrics and the direction the project is headed based upon the demonstrated capabilities thus far.