Writing and Submitting a Successful XSEDE Proposal - Bridges Ken Hackworth

Transcription

1 March 30, 2016 Writing and Submitting a Successful XSEDE Proposal - Bridges Ken Hackworth

2 Overview References & Terms Guidelines for XSEDE Research (XRAC) Request Other Documents Review Criteria Overview of Proposal (Request) Types and AcDons XSEDE Awards (AllocaDons) XSEDE Systems (Resources) Procedures for submifng AllocaDon request Why requests are reduced? Metrics

3 Terminology Allocation Request Types Startup Development/tes,ng/ por,ng/benchmarking EducaDon Classroom, Training Research Produc,on work 3 Types of XSEDE Projects PI Principal Inves,gator XRAS XSEDE Resource Alloca,ons System XRAC XSEDE Resource Alloca,ons CommiDee SU Service Unit = 1 Core-hour

4 Traditional v. Community XRAC proposals are accepted in three general categories of research acdvides Single Principal InvesDgator Large research CollaboraDons Community Services (e.g., XSEDE Gateways) The general requirements for proposals of all types remain largely the same. Whether requesdng compute, storage, visualizadon, or extended collaboradve support or some combinadon

5 General Proposal Outline I. Research ObjecDves II. ComputaDonal methodology (ApplicaDons/Codes) III. ApplicaDon efficiencies IV. ComputaDonal Research Plan V. JusDficaDon for SUs(TB) requested VI. AddiDonal consideradons Note: Sec,ons III and IV are olen integrated.

6 I. Research Objectives TradiDonal proposals Describe the research acdvides to be pursued Community proposals Describe the classes of research acdvides that the proposed effort will support. Keep it short: You only need enough detail to support the methods and computadonal plan being proposed. TIP Reviewers don t want to read the proposal you submi^ed to NSF/NIH/etc., but they need to see if you have merit-reviewed (grant) funding.

7 II. Computational Methods (and Applications/ Codes used) Very similar between tradidonal and community proposals. For compute requests Describe the ApplicaDons and components you will use. Describe the methods/algorithms employed in your computadonal research Describe code development/features/advances home-grown codes. For storage requests Provide descripdon of data to be stored (organizadon, formats, collecdon mechanisms, permissions granted or received) Describe the amount and expected growth of data to be stored.

8 III. Application Efficiencies Very similar between tradidonal and community proposals. For compute requests Explain why you chose specific resources for your applicadons. Provide performance and scaling details on problems and test cases similar to those being pursued. (What is the appropriate scale for your problem?) Ideally, provide performance and scaling data collected by you for the specific resource(s) you are requesdng. Mandatory document! For storage requests Explain the efficiency of your storage algorithms and protocols. Describe and esdmate the expected costs of scaling to larger data sets and a larger number of clients.

9 IV. Computational Research Plan TradiDonal proposals Explicitly describe the problem cases you will examine BAD: a dozen or so important proteins under various condidons GOOD: 7 proteins [listed here; include sciendfic importance of these selecdons somewhere, too]. Each protein will require [X] number of runs, varying [x] parameters [listed here] [in very specific and sciendfically meaningful ways] Community proposals Explicitly describe the typical use-case(s) that the gateway supports and the type of runs that you expect users to make Describe how you will help ensure that the community will make sciendfically meaningful runs (if applicable) BAD: the gateway lets users run NAMD on XSEDE resources BETTER: users will run NAMD jobs on [biological systems like this] BETTER STILL: the gateway allows users to run NAMD jobs on up to 128 processors on problem sizes limited [in some fashion]

10 V. Justification of SUs, TBs TradiDonal, Research proposals If you ve done secdons II, III and IV well, this secdon should be a straighmorward math problem For each research problem, calculate the SUs required based on runs (base units) defined in IV and the Dmings in secdon III, broken out appropriately by resource Reasonable scaling esdmates from test-case Dming runs to fullscale producdon runs are acceptable. Clear presentadon here will allow reviewers to award Dme or storage in a radonal fashion Analogous calculadons should apply for storage requests

11 V. Justification of SUs, TBs Community (gateway-type) proposals The first big trick: CalculaDng SUs when you don t know the precise runs to be made a priori. In Year 2 and beyond Start with an esdmate of total usage based on prior year s usage pa^erns and esdmate for coming year s usage pa^erns. From this informadon, along with data from secdons IV and III, you can come up with a tabuladon of SU esdmates. Year 1 requires bootstrapping Pick conservadve values (and jusdfy them) for the size of the community and runs to be made, and calculate SUs. TIP Start modestly. If you have ~0 users, don t expect the reviewers to believe that you will get thousands (or even hundreds) next year. Analogous calculadons for TBs of storage needed

12 VI. Additional Review Considerations! Ability to complete the work plan described (more significant for larger requests) Sufficient merit-reviewed funding Staff, both number and experience! Local compudng environment! Special Needs! Other access to HPC resources (e.g., Campus centers, DOE centers, etc.)

13 VI. Additional Considerations Community (gateway) proposals these components can provide key details: Community Support and Management Plan Describe the gateway interface in terms of how it helps community burn SUs or access TBs. Describe plans for growing the user community, graduadng users to Research allocadon awards, reguladng gateway hogs Progress report The actual user community and usage pa^erns Manuscripts thanking this service, or list ardcles referencing XSEDE. Local compudng environment Other HPC resources

14 Renewals require a Progress Report For Research Project Renewal and Supplement Requests Summary of Scien,fic Discoveries Accomplishments of Computa,on Plan Usage Achievements of the ComputaDons (more detail than summary). Specify the number of publica,ons, conferences, reports that result from XSEDE support. Contribu,ons to other research efforts. (experimental/computadonal/instrumental, etc.).

15 Other Documents: Required: CVs for PIs and Co-PIs (2 pages) Adding publicadons in the XRAS submission form pulling from portal profile OpDonal: Special Requirements References (If they won t fit in Main Doc.) Other

16 Review Criteria Appropriateness of Methodology ApplicaDons, methods, algorithms and techniques reasonably described and modvated? Data usage and access methods clearly described? Appropriateness of Research Plan Are research objecdves explained and how they will be achieved? Do proposed computadons list simuladon parameters that are needed to obtain results? Details of human resources devoted to accomplish these tasks? Efficient Use of Resources Are resources selected being used efficiently as possible and in accordance with the recommended use guidelines of the resources? Is performance and scaling date presented for the resource requested? Are there details of the efforts to improve the opdmizadon and/or parallelizadon of the applicadons?

17 XSEDE Projects An XSEDE Project is like a bank account for allocadons. It is permanent, only one per PI. It holds a year s worth of allocadon (on 1 or more systems) PI s request an allocadon renewal each year thereaqer. An AllocaDon awarded to a New Request creates an XSEDE Project. A PI s ComputaDonal Projects evolve over the years. ComputaDonal Projects begin, end and extend. In subsequent years successful Renewal Requests provide allocadons for new ComputaDonal Projects under the same XSEDE Project. Your XSEDE Project remains the same. A Renewal Requests is just like New Request, but must contain a Progress Report of last year s ComputaDonal Projects and list of publicadons from past year s allocadon.

18 Eligibility Principal invesdgator (PI) must be a researcher or educator at a U.S.-based insdtudon, including federal research labs or commercial organizadons, (Commercial requests must guarantee that their results are publically available, and work must be in collaboradon with an open science organizadon.) A postdoctoral researcher is eligible to be a PI. A qualified advisor may apply for an allocadon for his or her class; but a high school, undergraduate or graduate student may not be a PI.

19 Overview: Research Request portal.xsede.org "Allocations"Submit/Review Request ** Web forms: InvesDgator, Grants, Resource Request, Requires Main Doc. = proposal (pdf upload) & CV Reviewed by experts in same Field of Science 2.5 months from deadline to award availability Details: Alloca,on Size: Unlimited Reviewed: Quarterly Deadlines: 15 th of October, January, April, July Awards Begin: 1 st of January, April, July, October

20 Overview: Startup/Education Requests portal.xsede.org "Allocations"Submit/Review Request ** Web forms: InvesDgator, Resource Request, Requires only an abstract and CV Reviewed by a XSEDE Staff (Startup AllocaDons Commi^ee) 2 weeks from submission to award availability For code devel / performance eval / small-scaling computa,ons / classroom & training instruc,on Details: Request limit: 200,000 SUs total or combina,on of all resources requested Reviewed: within 2 weeks of submission Deadlines: None Awards Begin: within 2 weeks of submission

21 Proposal Document(s) CV (s) required for all requests. Abstract for startup/educadon request (in forms, or as a PDF document) Proposal Main Document for Research request (renewals/supplements) Key to a successful review: Adhere to page limits! JusDfy allocadon request. Page Limit Proposal Document 3 Progress report 10 New or Renewal 15 Over 10 Million SUs Pg. limit: DOES NOT INCLUDE FIGURES & TABLES.

22 The Award = Allocation One per PI (generally) 1-year duradon Unused SUs are forfeited at the end of an award period Progress report required for renewal requests. Add users to a grant via XSEDE User Portal Advance 4 quarters = 1 yr allocation period Submission Review Award Time to renew

23 The Resources: Compute HPC Systems: (Bridges(PSC), Comet(SDSC), Gordon(SDSC), Stampede(TACC), SuperMIC(LSU) and Wrangler(TACC), Xstream(Stanford)) Large Memory Systems: (Bridges(PSC) and Comet(SDSC)) VMs: (Jetstream(IU/TACC)) Advanced VIS Systems: (Maverick(TACC)) HTC Systems: (OSG) Storage Systems: (Data Oasis(SDSC), Jetstream(IU/TACC), Pylon(PSC), Ranch(TACC)) *Ranch is the only storage resources that can be requested without compute resources.

24 The Resources: Extended Collaborative Support(ECS) Dedicated, but limited, XSEDE staff assistance 5 QuesDons which are part of resource request secdon of applicadon Reviewers rate need for ECS

25 Login at portal.xsede.org

26 Example Form: New Project

27 Example Form: Getting Info in order

28 Example Form: PI entry page, populating with Portal information

29 Example proposal submission: Personnel/Roles

30 Example proposal submission: Title/Abstract/FOS

31 Example proposal submission: Resource request

32 Example proposal submission: Resources request

33 Example proposal submission: Resource Request

34 Example proposal submission: ECS

35 Example proposal submission: Uploading documents

36 Example proposal submission: Uploading documents

37 Example proposal submission: Uploading documents

38 Example proposal submission: Supporting Grants

39 Example proposal submission: Supporting Grants

40 Example proposal submission: Supporting Grants

41 Example proposal submission: Publications

42 Example proposal submission: Saving and Final Submission

43 Example proposal submission: Successful submission

44 Pending Request

45 Notification of successful submission

46 Requesting an Advance

47 Why a Research Request may be reduced or rejected: Poor or no Scaling/Code performance informadon Scaling/Code performance informadon not from resource being requested Non-disclosure of access to other resource(s) Disclosure of access to other HPC resource(s) but no details of difference of research MulDple requests from same research lab Page limits Poor jusdficadon for resources Lack of descripdon of research team Low producdvity from prior XSEDE Research award with no details in progress report

48 Example Proposals

49 Interesting Facts ~800 research requests per year ~1,200 other requests ~4.0B SUs requested(3.7b are research requests) ~1.4B SUs awarded(1.0b are research awards)

50 Questions? Asking for Help