Spring semester of the 2014-2015 year was the final term in order to complete the project. Here is a week by week break down of the work done this final semester.
The first week of the semester was spent in planning. Coming off of a month of solid development during RIT's intersession period required for some housekeeping and re-organization to occur. During this time, project management was renewed and revamped to help ensure the successful completion of the project.
Week 2 was spent re-initializing several cluster nodes and developing all of the new network interfacing required to support the project. Two more storage nodes were added, as well as an additional increase in pool storage. Overall, 10TB was added to the project. Additionally, all Intruder nodes received new RAM, boosting their stock to 8-12GB apiece.
Virtualization was investigated in order to increase the productivity of the cluster. Additionally, network testing restarted to ensure that networking is indeed the bottleneck of the program.
Heavy programming.
The skeleton of the program was written - in shell. Adding user directory checking, performance monitoring, and starting to foray into reverse engineering the Sony 3D LUT's
Spent the week reverse engineering Sony RAW Viewer to take their LUT's. LUT's were ingested and converted to PS6 formats and committed to repositories. This was then used to generate all of the GraphicsMagick HALD LUT's. This continued through Week 7.
See Week 6.
More of the initial thesis program written - optimized some of the C code. Also all directory input was sanitized from student tests and directory structure of the NFS shares was tested. Shares holds well, networking is still bottleneck.
Virtualization kicked off. Researching, and implementing bare metal hypervisors for CentOS hosts on Crawler and Whisper (Part of the Fortress bare metal class). VMWare ESXi chosen.
More storage and RAM added. Cluster size grows by 8 cores. Re-optimizing code and now restructuring function calls within header program. Imaging science tests begin as well - full image pipeline completed and represented with HALD LUT's.
ESXi virtualized hosts generated and migrated. Cloned templates now on /SLURM/host and 4 guests per host is running well. Must update thesis program to handle guests as physical hosts as seen by the program.
Updating the code one function at a time. Completed a full pool status test for all nodes.
Final testing, ensuring the project can work at scale
DigitalOcean VM hosting and full scale enterprise testing.
Final prep for presenting and paper writing.