Proposal

No annual report

Introduction

Proposal ID 2015-089
Submitted April 22, 2015
Owner rinnan
Department Forest Resources, College of
Category Remote Computing
Funding Status Fully Funded
Metric Score 3.44

Contacts

Primary
  • Name
  • Title
  • Email
  • Phone
  • Mailbox
  • Shane Kraus
  • Sr. Computing Specialist Lead
  • skrause@uw.edu
  • 206-616-5414
  • Box 352100
Budget
  • Name
  • Title
  • Email
  • Phone
  • Mailbox
  • Wendy Star
  • Administrator
  • wkstar@uw.edu
  • 206-685-2047
  • Box 352100
Dean
  • Name
  • Title
  • Email
  • Phone
  • Mailbox
  • Lisa Graumlich
  • Dean
  • envdean@uw.edu
  • 206-221-0908
  • Box 355355

Descriptions

Abstract

Students in the Quantitative Ecology and Research Management Interdisciplinary Graduate Program (QERM) conduct an enormous amount of data analysis and computationally demanding simulation studies as part of their research. To date, students in the QERM program have had limited access to computing resources, and have instead relied on limited access to servers across campus, which has led to a fractured and uncoordinated research environment. This proposal requests the purchase of a high-performance server for computing and simulation studies for the QERM students and the students in the core departments supporting the program (the School of Environmental and Forest Sciences [SEFS] and the School of Aquatic and Fisheries Sciences [SAFS]). Having a dedicated server for students in the QERM program and its affiliated departments will help to centralize computing resources, establish resource equity for the students, facilitate collaboration, and result in higher quality research and more research opportunities for students.

Category Justification

Remote Computing – This proposal is for a centralized computing system that can be operated remotely and simultaneously by several individuals.

Background

The Quantitative Ecology and Resource Management Interdisciplinary Graduate Program trains students to apply statistics, mathematics, and decision science to a diverse range of problems in terrestrial and marine ecology and natural resource management. QERM students enjoy strong working relationships with a wide variety of UW departments, governmental and non-governmental organizations, and other educational establishments, and their research is at the cutting edge of science and policy.

More and more, environmental science is turning to “big data” for answers, relying on analysis of large datasets to discern meaningful patterns and information, and computationally expensive simulation studies to make substantive and accurate predictions. A large portion of research conducted by QERM students involves these types of data manipulation and analysis.

Although it has operated successfully as an independent program for more than 25 years, the QERM program is fairly limited in its own resources, and relies a great deal on the infrastructure of affiliated departments. QERM students presently must rely on a piecemeal patchwork of computing resources spread across campus, each with their own set of regulations, guidelines, and different available software.

This proposal will create a centralized server for the use of QERM students and students in affiliated departments (SEFS and SAFS). This will provide a more stable and reliable computing resource that is uniquely tailored to the needs of our program.

Benefits to Students and the University

Student research: A program-specific server will provide reliable, location-independent access to needed software. It will greatly increase the efficiency of research and make some research more analytically tractable because of the access to computing power and the possibility of parallelization of processes. It will also facilitate student collaboration by providing a centralized workspace accessible to everyone affiliated with the QERM program, regardless of department affiliation.

Program: Having our own server will help the QERM program to encourage interdisciplinary collaboration. It will be easier for students to conduct research with professors outside of traditional QERM-associated departments without concern for computing support. More professors may also be interested in working with QERM students knowing that we offer our own computing resources. Such a server will also be seen as a benefit for recruitment of prospective students.

Departmental Endorsements

I am pleased to offer my strongest possible endorsement for the application for Student Technology Funds to enhance the computing capacities for the Quantitative Ecology and Resource Management Program (QERM). QERM is a unique interdisciplinary graduate program that focuses on the application of cutting edge quantitative tools to address ecological and natural resource management questions. As such, QERM students frequently find themselves in need of high-capacity computing, as more and more quantitative methods involve either complex models, "big data", or a combination of the two. It is not uncommon for students to pare down the scope of some of their thesis or dissertation work for lack of dedicated access to computer facilities that will enable them to conduct a fuller, more comprehensive analysis.

The centralized server proposed here will greatly enhance the opportunities for graduate student research in QERM. As an interdisciplinary program housed within the graduate school, we have limited access to technology improvement funding that programs in other administrative units might enjoy. We also have a modest budget with little discretionary funds that could be used to invest in technology. Consequently, this investment via the STF will provide a vast improvement in the student’s access to much needed technology. Moreover, this access will help us recruit top-ranked applicants who will be looking enroll in a program that provides them with the resources they need.

-- Tim Essington, Professor and Associate Director, Aquatic & Fishery Sciences and Director, Quantitative Ecology and Resource Management Program

Computational resources are often a limiting factor for graduate students in the QERM program and its affiliated departments. These students often have to borrow server time and many are unable to run the analyses they want due to a lack of resources. A server dedicated primarily to students in the QERM program and its supporting departments would give these students a needed resource, allow them to complete their research in a timely manner, and to tackle bigger and more ambitious research problems. Although I have a server that I allow students in other research groups to use when it is available, it is often being used to capacity by my own graduate students and postdocs. It would be wonderful to have a dedicated server that students from QERM and SEFS and SAFS in general could use for their research. I fully support the proposed server purchase.

-- Joshua Lawler, Associate Professor, School of Environmental & Forest Sciences and the Quantitative Resource Management Program

Access to a dedicated, high-performance computational resource for the Graduate School’s QERM student population is long overdue. The modeling experiments conducted by these students often require a controlled computing environment: the ability to control and constraint certain background processes and applications in order to make CPU measurements possible. This is a fundamental capability that is needed when one is to construct computer models that run beyond polynomial time and performance measurements are highly sensitive to what the operating system and other users are doing. This capability is not even remotely possible on a shared computing or data center to which all of us have access at the University of Washington. For these reasons, I highly recommend this initiative for funding.

-- Sándor F. Tóth, Associate Professor, School of Environmental & Forest Sciences and the Quantitative Resource Management Program

I strongly support the proposal for QERM to purchase and maintain and powerful server for simulation running. I see this as essential for the continued growth and extension of the QERM program.

My students commonly use simulation in their individual research, which can be incredibly CPU intensive and often takes weeks to run an analysis. The simulations typically consist of a statistical model fitted to data generated from a known, controlled source. A single model is quick to run, but to understand the impact of stochasticity they run the same model for many replicates, each with different stochastic errors. As such, CPU time is often a limitation in the studies, restricting the number of scenarios or replicates that can be run. Because of these long run times it is inconvenient or impossible to run them on personal computers or laptops. A dedicated server with remote access would allow my students to develop, test, and run their code with ease. Further, due to their independent nature, replicates can easily be distributed across multiple CPU cores in the same session. We therefore would prioritize CPU cores over other hardware such as RAM or hard disk. My students would certainly make good use of a centralized, remote-accessible server.

-- Trevor A. Branch, Assistant Professor, School of Aquatic & Fisheries Sciences and the Quantitative Resource Management Program

To my knowledge, QERM is the only quantitative PhD program on campus that does not have a dedicated computer cluster. This is clearly a suboptimal arrangement, because QERM students routinely need to run computationally intensive numerical experiments during their PhD research. The current solution to this problem is for students to ask their advisors or committee members for access to a computer cluster in the home department of the advisor or committee member. I received such a request at least once when serving on a QERM PhD dissertation committee. Not only this solution is difficult to implement in practice, because the departments are protective of their computing facilities, but in some cases the solution is infeasible. For example, if a QERM student works with Anthropology faculty, getting access to a computer cluster becomes difficult, because, as far as I know, UW Department of Anthropology has no such resource. Gaining dedicated QERM cluster will be a huge boon for the program and will give QERM students more freedom in choosing their advisors and will allow them to tackle more ambitious research problems. I support this proposal highly.

-- Vladimir Minin, Associate Professor, Statistics, Biology, and the Quantitative Resource Management Program

Installation Timeline

Equipment will be purchased and installed as soon as funding becomes available.

Resources Provided by Department

SEFS Technical Services will provide installation, storage, maintenance, and management of all equipment and services. SEFS Technical Services staff will also help prioritize access and provide technical support.

Access Restrictions (if any)

QERM students will have priority access to the server, but access will be granted to students in SEFS and SAFS as well as other students from across campus as availability permits.

Student Endorsements

I have regularly used a remote server for my research for the last three years, and it is an invaluable resource. In that time, I have experienced difficulty sharing my work with other students in my program and vice versa, since our professors are associated with different departments and we don’t have the same access rights. A server for our program would greatly facilitate student collaboration, and would make my workflow much smoother.

-- Scott Rinnan, QERM graduate student

Given the amount of computation that happens in QERM-related research, a centralized supercomputer seems more than justified. I could see myself using this frequently, primarily for solving optimization models. They can easily take days or weeks to solve on even high-powered standalone computers. Access to a supercomputer has the potential to speed up our research by hours or even days per model. Multiply that gain by the number of models that we solve and we're looking at weeks of waiting compressed to days.

-- Nick Kullman, QERM graduate student

Access to a centralized supercomputer would be great for my work. I currently am in the process of running a stock assessment model and a simulation study, both of which run much faster in parallel and which I can't run simultaneously on my desktop. I also have run Bayesian models in parallel for my dissertation work and had to use the SAFS supercomputer, which is in a computer lab. It would be great to be able to use that much power from my office or home.

-- Christine Stawitz, QERM graduate student

I use a server all the time and think it's a great idea to get one for QERM. I use it to run simulations for stock assessment model testing. In this framework we run many replicates of many scenarios, ending in tens of thousands of model runs. The ability to remote connect in is an absolute must, but the IT in SAFS is hesitant to allow that for their servers.

-- Cole Monnahan, QERM graduate student

Having access to a centralized supercomputer would greatly improve the progression of my work. I often use a program (MATLAB) on my personal computer to run long computations that can take a week or longer, preventing me from running smaller computations in the meantime and draining speed for other tasks on my computer. Being able to run such jobs remotely would boost my productivity greatly by both speeding up lengthy calculations and allowing me to multitask as I run mixtures of small and large jobs on different computers.

-- Austin Phillips, QERM graduate student

I think this is an excellent idea. Most QERM students have projects that involve some amount of intensive computing, while some have projects with a substantial amount of computing requirements. The project I am currently working on is very computationally demanding. I am performing simulations to assess Bayesian model fits, which can take weeks to run. The time requirement for these types of projects could be greatly reduced if resources were available to allow parallelization of the processes across multiple processor cores. Therefore, I think that a powerful, multi-core server available to QERM students would be a very valuable resource to the program.

-- Jim Faulkner, QERM graduate student

This sounds like a great idea! I've been doing some simulation work (mostly MCMC) that takes on the order of days per model for the simplest models, with a number of iterations running in the tens of thousands. Due to the complexity of some of these models, I'd really like to run them for hundreds of thousands of iterations, but the computing power of my office machine can't do that in a reasonable amount of time. Having a supercomputer resource would be invaluable for speeding up this work.

-- Will Chen, QERM graduate student

A simulation server devoted to the QERM department is a necessary step toward keeping QERM's departmental research at the cutting edge of modeling, optimization, and statistical analysis. Given the computationally heavy nature of many QERM research projects (mine certainly included), and the clear and present inevitability of increasingly large data sets, a devoted, high-performance server would be a highly used and greatly appreciated asset to the QERM program.

-- Melissa Muradian, QERM graduate student

A significant proportion of the QERM student community is involved in fitting complex ecological models that may require days, weeks and possibly months to fit on typical desktop platforms. Using personal desktops for this task can be rather inconvenient and can carry the risk of accidental data loss. Moreover, many QERM students, myself included, are consumers of climatological data. This space-time data can oftentimes be terabytes in size and access to a server can be a prerequisite to manipulating these large file types. Hence, a QERM server would be an excellent tool for QERM students in both ecological and climatological research.

-- Harry Podschwit, QERM graduate student

Items

Group Funded Item Unit price Quantity Subtotal
Hardware

PowerEdge R730 Server

$14,376.00 3 $43,128.00
Description

High performance terminal servers

Justification

Needed for advanced analytic computations. 3 servers will allow load balancing and multiple simultaneous users to get full capacity from the equipment.

None

Tax

$4,314.00 1 $4,314.00
Description

Tax

Justification

Tax

Total requested: $47,442.00

Total funded: $47,442.00

Comments

No comments