Harry Mangalam, PhD

With the help of 2.5 other SysAdmins, I help run UC Irvine’s HPC compute cluster, the physical characteristics of which are described here in more detail. I do most of the BeeGFS-related work (5 separate filesystems ranging from 1 -6 subvolumes, totalling about 1.5PB, on ext4, XFS, & ZFS). I also respond to most of the biology-related software questions and installations (we support about 1000 custom-installed software packages via environment modules. Also the Python, Perl, R, and MATLAB installs and questions.

I’m comfortable coding and debugging bash, Perl, Python, C, R, tho I’ve been doing progressively original programming (tho see the section below for specific examples. As well, I’ve installed and maintained various web-based systems for tech support (Trac, Request Tracker) and the OwnCloud private cloud system.

I also particiate in UC-wide Research Computing discussions and Organizations including the Research IT Committee and the UCSD-led Pacific Rim Platform, to create structures that allow faster and more effective cross-campus support and collaboration. I was responsible for writing and updating the UC Irvine Cyberinfrastructure Plan and also largely structured and edited UCI’s Research CyberInfrastructure Center proposal

An increasing amount of modern research involves moving large chunks of data from one place to another; in this job, I’ve debugged many of the parameters of that movement, including internal to programs, on a system level, and on the networks, both LAN and WAN, including writing a parallel version of rsync and an improved mechanism to transfer data quickly if insecurely, tnc. Recently, I’ve been involved in large amounts of data across WANS, mostly to Google’s cloud storage, using both their gsutil and the rclone utility. This experience is currently being added to the data movement doc mentioned above.

In addition to the Systems Administration, I also do consulting with researchers at all levels and for all domains. Many such problems have to do with data movement, analytical workflows, getting applications to work (or to work together), debugging scheduler scripts, and occasional catalytic programming. By catalytic, I mean small amounts of programming that unblocks a key step in a workflow that is holding up an analysis. Also, catalytic implies not being consumed in the process.

A frequent task is finding software to support a research need. Being research, quite often there isn’t a packaged or commercial system available, so I have to knit together multiple pieces of software into a pipeline or script, usually addressed with Perl, Python, and R/Bioconductor. If it involves computational bottlenecks I have experience in several profiling tools (Oprofile, perf, HPC Toolkit, valgrind, etc), can program in C and am familiar with the GNU build toolchain.

I also assist in general infrastructure problems such as campus-wide software distribution & licensing (for which I initiated the 1st completely electronic software distribution system on campus (example here), and storage & backup issues, network problems, outreach to faculty to identify and resolve bottlenecks, representation to campus groups and UC-wide discussion & planning groups.

I also teach the Linux part of an introductory HPC/Bioinformatics course called BioLinux roughly once a month, which includes both lecture and tutorial. This has been expanded into teaching techniques in BigData handling.

Due to my ongoing relationship with local companies, our group has been the recipient of 7 racks of compute nodes (300 cores/rack, 3 racks currently in operation as part of HPC, 2 racks donated to ICS), 6 multicore (32-128 core) servers, several tape robots, some NVME SSDs, and a 396-port (72 ports active) QDR Infiniband switch. The approximate depreciated value of this equipment was about $250K.

Co-author & co-editor (with the RCI Vision Workshop) of the document A Vision For Research CyberInfrastructure that was recently funded from the UCI internal budget to provide about $1.4M for the establishment of the RCIC, hiring of several ongoing FTEs and significant hardware upgrades for Research Computing at UCI.

Co-PI (with Dana Roode, Allen Schiano) of a 2 yr, $400,000 NSF CyberInfrastructure grant to support an engineer to review and implement a number of networked storage technologies. (Sept, 2015 - Sept, 2017).

Co-PI (with Drs. Carl Cotman, Ken Longmuir, Tatsuya Suda, David Walker, all of UC Irvine) of a Pacific Bell CalREN grant: Medical Informatics Use of Asynchronous Transfer Mode (ATM) service across the LA basin (June '94-September '97)

Co-PI (with Dr. Thomas Cesario) of Irvine Health Foundation Grant to implement ATM-based telemedicine applications using the infrastructure of the above CalREN grant (July, 1995 - September, 1997). This resulted in one of the first telemedicine multicast video broadcasts via the MBONE Internet system in 1994.

Medical Research Council of Canada Post-doctoral Fellowship to study Drosophila genetics with Dr. John Thomas at the Salk Institute (1990-1991).