HTCondor Week 2015

University of Wisconsin
Madison, Wisconsin
May 19–22, 2015

picture of Madison

Wednesday, May 20

 8:00 am  9:00 am Breakfast and Registration
vegetarian quiche, whole wheat toast, muffins, fruit platter, orange juice, coffee, tea
Session Moderators: John "TJ" Knoeller, Brian Lin
 9:00 am  9:30 am Welcome
Miron Livny
Center for High Throughput Computing
 9:40 am 10:10 am What's new in HTCondor? What's coming up?
Overview of upcoming enhancements in HTCondor v8.4.x and beyond. Note the opening video from the talk can also be found on YouTube at
Todd Tannenbaum
Center for High Throughput Computing
10:15 am 10:45 am Break
coffee, tea, soda, cookies, fresh fruit, yogurt, granola
Session Moderators: Aaron Moate, Mátyás Selmeci
10:45 am 11:05 am Batch Services at CERN - Present Status and Future Evolution
Helge Meinhard
11:10 am 11:30 am Implementing an HTCondor service at CERN
At CERN our batch system runs job submissions from physicists around the world, looking to analyze the LHC data. Currently, this represents around 65,000 cores running 400,000 jobs a day, at any one time over 60,000 jobs are executing. This has now become a matter of scaling, and when compared to other systems, HTCondor looks to be our best option to take us to 50,000 worker nodes and beyond. Therefore, CERN IT plans to move to HTCondor as their primary batch management software in the next few years. In this presentation we report about the status of the project and our future plans.
Iain Steers
11:35 am 11:55 am CMS Experience Provisioning Cloud Resources with GlideinWMS
Anthony Tiradani
Fermi National Accelerator Laboratory
12:00 pm  1:10 pm Lunch
prime rib, chicken roulade, green salad with vinagrette, eggplant and mozzeralla napoleon with roasted tomato sauce, bread, brownies, bars
Data Management
Session Moderators: Tim Theisen, Lauren Michael
 1:10 pm  1:30 pm Efficiently Sharing Common Data
We examine the pros and cons of combining two technologies: Using BitTorrent for data transfer, and caching data on execute nodes using a pluggable cache replacement policy.
Zach Miller
Center for High Throughput Computing
 1:35 pm  1:55 pm Data caching with HTCondor: The HTCondor CacheD
We present the HTCondor CacheD, a caching daemon that optimizes stage-in time for large shared input data through through novel transfer methods and proactive replication.
Derek Weitzel
University of Nebraska – Lincoln
 2:00 pm  2:20 pm HTCondor Input File Transfer via Squid Caching Proxy
Carl Vuosalo
Center for High Throughput Computing
 2:25 pm  2:45 pm Scientific Workflows - How Pegasus can enhance your DAGMan experience
Karan Vahi
USC Information Sciences Institute
 2:50 pm  3:25 pm Break
coffee, tea, soda, snack mix, peanut butter filled pretzels, tortilla chips, salsa, guacamole, vegetable and dip platter
Session Moderators: Jaime Frey, Kent Wenger
 3:25 pm  3:45 pm "No Idle Cores": Automatic Load Balancing in our ATLAS Pool
In our ATLAS pool, multicore jobs coexist with single core jobs of many different varieties, and ensuring that all types of jobs have access to the resources they require is a challenging problem. I will discuss how we partially solved the problem a combination of Hierarchical Group Quotas and Partitionable Slots, and then present work we did to allow automatic balancing between groups based on demand from PANDA, the ATLAS workload-management system -- balancing the effects of single vs. multicore jobs and queues with differing priorities.
William Strecker-Kellogg
Brookhaven National Laboratory
 3:50 pm  4:10 pm Scientific Computing on emerging infrastructures using HTCondor
Sanjay Padhi
University of California, San Diego
 4:15 pm  4:35 pm One Pool To Rule Them All: The CMS HTCondor/glideinWMS Global Pool
The CMS experiment at CERN runs all distributed computing now through a single global HTCondor/glideInWMS pool. I will discuss the reasons for doing this, how it is constructed and our experience running with it.
David Mason
Fermi National Accelerator Laboratory
 4:40 pm  5:00 pm Adopting HTCondor at Raytheon
Raytheon Company, with 2014 sales of $23 billion and 61,000 employees worldwide, is a technology and innovation leader specializing in defense, security, and civil markets throughout the world. Today's extensive High Performance Computing needs require Raytheon's Information Technology organization to continuously push the envelope to create leading-edge technical processing solutions for complex systems and program challenges. Beginning in late 2012, Michael Pelletier took on such a challenge, a 750+-thread compute cluster which was straining the limits of the existing infrastructure, as well as the capabilities of the IT support team and its engineering partners. By adopting HTCondor, and with the help of the CHTC development team, he was able to transform the performance of the existing hardware, and leverage HTCondor's unique capabilities simplify workflow and multiply the productivity of the modeling, simulation, and analysis tasks across multiple Raytheon programs. He will discuss Raytheon's adoption of HTCondor and the innovations and improvements which flowed from it.
Michael V. Pelletier
Raytheon Integrated Defense Systems - Information Technology

Reception sponsored by Cycle Computing

Courtesy of Cycle Computing we will have a reception following the Wednesday session. It will be at Lucky's Bar and Grille, 1421 Regent Street from 6pm to 7pm. Drinks and appetizers will be served. It is a short walk from the meeting location. Walking directions

Specific talks and times are subject to change.