Research Computing: How do you measure up? · What is the obstacle keeping you from the next...
Transcript of Research Computing: How do you measure up? · What is the obstacle keeping you from the next...
Research Computing: How do you measure up?
Leah Lang Director of Analy2cs Services
EDUCAUSE
Michele Norin Senior Vice President and Chief
Informa2on Officer Rutgers, The state University of New
Jersey
Agenda
§ EDUCAUSE Core Data Service § EDUCAUSE Benchmarking Service § Landscape of Research Computing § Research Computing Case Study § Challenges/Successes in Research
Computing
EDUCAUSE Core Data Service
FREE
BENCHMARKING SERVICE
IT financials, staffing, and services
Step 1: Complete the survey
Administration and Management of IT IT Support Services Educational Technology Services Research Computing Services Data Center services
Communications Infrastructure Enterprise Infrastructure and Services
Information Security Identity Management
Information Systems and Applications
CDS Update Newsle/er
813 Participating Institutions in 2015
Response Rate based on all eligible institutions
199
15%
27%
42%
30%
76%
55%
73%
0
50
100
150
200
AA BA MA Priv MA Pub DR Priv DR Pub Internet2 0%
10%
20%
30%
40%
50%
60%
70%
80%
90%
100%
CDS Reporting is powered by
Step 2: Access the data
Total central IT spending per institutional FTE vs. total central IT spending as a percentage of institutional expenses
Internet2 $1,035 3.5%
AA $637 5.2%
BA $1,281 4.4%
MA public $684 4.3%
MA private $987 4.5%
DR public $968 3.3%
DR private $1,620 3.6%
0.0%
4.2%
8.4%
$0 $917 $1,834
Med
ian total cen
tral IT sp
ending as a
pe
rcen
tage of ins3tu3
onal expen
ses (all
nonspe
cialized
U.S. m
edian = 4.2%
)
Median total central IT spending per ins3tu3onal FTE (students, faculty, and staff)
(all nonspecialized U.S. median = $917)
Central IT FTEs per 1,000 institutional FTEs
9.7
7.0
9.2
7.0
11.1
5.2
7.8
7.6
0 2 4 6 8 10 12
DR private
DR public
MA private
MA public
BA
AA
All nonspecialized U.S.
Internet2
Median central IT FTEs per 1,000 ins3tu3onal FTEs
CDS Benchmarking Report CDS Almanacs ECAR analysis of Core Data (accessible to ECAR subscribing institutions)
Step 3: Gain additional insight
EDUCAUSE Benchmarking Service
1) Build reports on demand with customized peer groups
I. Introducing the EDUCAUSE Benchmarking Service
2) Benchmark maturity and technology deployment
I. Introducing the EDUCAUSE Benchmarking Service
2015 EDUCAUSE Core Data Service
I. Introducing the EDUCAUSE Benchmarking Service
2015 EDUCAUSE Core Data Service
I. Introducing the EDUCAUSE Benchmarking Service
2016 Landscape of Research Computing
Research Computing spending as a percentage of central IT spending
2%
2%
1%
1%
1%
2%
1%
1%
0% 10%
DR private
DR public
MA private
MA public
BA
AA
All nonspecialized U.S.
Internet2
Research Com
pu2n
g
Median IT domain area spending as a percentage of central IT spending*
2015 EDUCAUSE Core Data Service *at ins=tu=ons with any research compu=ng expenses
Central IT domain area FTEs per 1,000 institutional FTEs
0.1
0.1
0.0
0.03
0.0
0.0
0.1
0.1
0 1 2
DR private
DR public
MA private
MA public
BA
AA
All nonspecialized U.S.
Internet2
Research Com
pu2n
g
Median central IT domain area FTEs per 1,000 ins3tu3onal FTEs*
2015 EDUCAUSE Core Data Service *at ins=tu=ons with any research compu=ng expenses
Organizational unit primarily responsible for research computing services
2015 EDUCAUSE Core Data Service
0% 50% 100%
Specialized scien2fic
Ins2tu2onal grid compu2ng
Cyberinfrastructure
Mgmt of research servers
IaaS cloud service
High-‐throughput compu2ng
Data mgmt/storage/cura2on
High-‐performance compu2ng
Data centers for acad. units
High-‐performance wide area
High-‐performance local area
Primarily central IT Shared Other unit System Outsourced (non-‐comm) Outsource (comm)
Internet2
Organizational unit primarily responsible for research computing services
2015 EDUCAUSE Core Data Service
0% 50% 100%
Specialized scien2fic
Ins2tu2onal grid compu2ng
Cyberinfrastructure
Mgmt of research servers
IaaS cloud service
High-‐throughput compu2ng
Data mgmt/storage/cura2on
High-‐performance compu2ng
Data centers for acad. units
High-‐performance wide area
High-‐performance local area
Primarily central IT Shared Other unit System Outsourced (non-‐comm) Outsource (comm)
0% 50% 100% Internet2
Research Compu3ng Maturity Index by Carnegie Class
2015 EDUCAUSE Core Data Service
5. Op3mized 4. Managed 3. Defined 2. Repeatable 1. Absent/ad hoc
3.2
3.2
3.2
3.0
3.0 3.2
3.4
3.2
1.0
5.0 US Non-‐specialized
AA
BA
MA Pub
MA Priv
DR Pub
DR Priv
Internet2
Research Compu3ng Maturity Index by Carnegie Class
5. Op3mized 4. Managed 3. Defined 2. Repeatable 1. Absent/ad hoc 2015 EDUCAUSE Core Data Service
3.2
3.2
3.2
3.0
3.0 3.2
3.4
3.2
1.0
5.0 US Non-‐specialized
AA
BA
MA Pub
MA Priv
DR Pub
DR Priv
Internet2
Research Compu3ng Maturity (All Non-‐specialized US Ins3tu3ons)
Central IT Emphasis/Priority
Ins2tu2onal Emphasis/Priority
Pervasiveness
Technical Infrastructure Physical/Facili2es Infrastructure
Centraliza2on
5. Op3mized 4. Managed 3. Defined 2. Repeatable 1. Absent/ad hoc 2015 EDUCAUSE Core Data Service
Support 3.2
2.9
4.0
3.2 3.7
2.9
3.1
1.0
5.0
US Non-‐specialized
Internet2
Central IT Emphasis/Priority
§ Types of items § Research computing is a part of the central
IT strategic plan § Central IT staff are written into grants § Central IT is proactive about research
computing needs § Recommendations
§ Collaborate with researchers on plans § Share successes in partnering with central
IT to build awareness
Role of central IT in research grant awards
2015 EDUCAUSE Core Data Service
0% 25% 50%
Grants had staff in central IT as PI and a faculty member as SI.
Grants had staff in central IT as PI and a faculty member as co-‐PI.
Grants had a faculty member PI and staff in central IT as SI.
Grants had a faculty member PI and staff in central IT as co-‐PIs.
Grants to faculty members had explicit subcontracts or budgetary alloca2ons for
central IT
Internet2 All non-‐specialized US
Institutional Emphasis/Priority
§ Types of items § Research computing is a strategic,
institutional priority § Functional funding model § Adequate number of trained staff § Policies in place
§ Recommendations § Partner with senior leaders to develop a
strategic staffing and funding model § Assess faculty satisfaction with services
Funding models for research computing
2015 EDUCAUSE Core Data Service
0% 50% 100%
Specialized scien2fic apparatus
High-‐performance compu2ng
High-‐throughput compu2ng
Ins2tu2onal grid compu2ng
High-‐performance local area
Cyberinfrastructure
Data centers acad. unit
Mgmt of research servers
High-‐performance wide area
Data mgmt/storage/cura2on
Opera2ng budget Capital budget Student IT fee Chargeback (dept) Chargeback (grants) Other
Funding models for research computing
2015 EDUCAUSE Core Data Service
0% 50% 100% Internet2
0% 50% 100%
Specialized scien2fic apparatus
High-‐performance compu2ng
High-‐throughput compu2ng
Ins2tu2onal grid compu2ng
High-‐performance local area
Cyberinfrastructure
Data centers acad. unit
Mgmt of research servers
High-‐performance wide area
Data mgmt/storage/cura2on
Opera2ng budget Capital budget Student IT fee Chargeback (dept) Chargeback (grants) Other
Pervasiveness
§ Types of items § HPC in place § Researchers place a high priority on research
computing § A variety of academic disciplines are
supported § Recommendations
§ Establish partnerships across disciplines to leverage shared resources
Research Computing Technology Deployment
2015 EDUCAUSE Core Data Service
0% 10% 20% 30% 40% 50% 60%
Special apparatus Local sodware Cloud service Visualiza2on
Revision control Ins2tu2onal grid
Research databases Cloud storage Data archival
Specialized sodware Consul2ng | training
High throughput compu2ng Data storage
High performance compu2ng High Performance LAN High Performance WAN
Internet2 All non-‐specialized US
Technical Infrastructure
§ Types of items § Adequate bandwidth and storage capacity § Adequate processing capacity for current and
future needs § Recommendations
§ Review infrastructure status and gaps § Plan for increased capacity needs
TeraFLOPS capacity for HPC
2015 EDUCAUSE Core Data Service
0 10 20 30 40 50 60 70 80
Agencies
Acad unit
Admin unit
Central IT
Internet2 All non-‐specialized US
Physical/facilities Infrastructure
§ Types of items § Adequate floorspace, power, and cooling for
research computing § Recommendations
§ Review infrastructure capacity and status § Develop a plan for improvement
Data Center Cooling
2015 EDUCAUSE Core Data Service
0% 10% 20% 30% 40% 50% 60% 70% 80% 90%
Wet pipe
Dry pipe
Cleaning agent
Chillers
Generators
Internet2 All non-‐specialized US
Centralization
§ Types of items § Centrally managed and planned research
computing § Researchers are not solely responsible for
obtaining research computing needs § Recommendations
§ Increase collaboration efforts between researchers and central IT
§ Identify cost efficiencies gained by centralizing
Research Computing Organization
2015 EDUCAUSE Core Data Service
0% 5% 10% 15% 20% 25% 30% 35% 40%
Other
None (no plan)
None (planned)
Minimal (no plan)
Minimal (planned)
Independent (no plan)
Independent (planned)
Integrated
Internet2 All non-‐specialized US
Support
§ Types of items § Adequate wait time for consulting and other
requests § Recommendations
§ Assess wait times § Determine strategies to decrease wait times
Organizational unit primarily responsible for research computing practices
2015 EDUCAUSE Core Data Service
0% 50% 100%
Sta2s2cal consul2ng
Help preparing applica2ons
Data mgmt plans (NSF)
Experience sharing
Visualiza2on
Sodware development
Federal resources
Info sec plans
Other tech aspects
Cloud resources
HPC configura2on
Storage solu2ons
Primarily central IT Shared Other unit System Outsourced (non-‐comm) Outsource (comm) Not applicable
Organizational unit primarily responsible for research computing practices
2015 EDUCAUSE Core Data Service
0% 50% 100%
Sta2s2cal consul2ng
Help preparing applica2ons
Data mgmt plans (NSF)
Experience sharing
Visualiza2on
Sodware development
Federal resources
Info sec plans
Other tech aspects
Cloud resources
HPC configura2on
Storage solu2ons
Primarily central IT Shared Other unit System Outsourced (non-‐comm) Outsource (comm) Not applicable
0% 50% 100% Internet2
Research Computing Case Study
Research Compu3ng Maturity (All Non-‐specialized US Ins3tu3ons)
Central IT Emphasis/Priority
Ins2tu2onal Emphasis/Priority
Pervasiveness
Technical Infrastructure Physical/Facili2es Infrastructure
Centraliza2on
5. Op3mized 4. Managed 3. Defined 2. Repeatable 1. Absent/ad hoc 2015 EDUCAUSE Core Data Service
Support 3.2
2.9
4.0
3.2 3.7
2.9
3.1
1.0
5.0
US Non-‐specialized
Internet2
Research Compu3ng Maturity (All Non-‐specialized US Ins3tu3ons)
Central IT Emphasis/Priority
Ins2tu2onal Emphasis/Priority
Pervasiveness
Technical Infrastructure Physical/Facili2es Infrastructure
Centraliza2on
5. Op3mized 4. Managed 3. Defined 2. Repeatable 1. Absent/ad hoc 2015 EDUCAUSE Core Data Service
Support
1.0
5.0
US Non-‐specialized Internet2 Rutgers Arizona
“Moving the Needle” on Research Computing
Discussion Questions
§ What is the obstacle keeping you from the next maturity level for this dimension? What can help you overcome that obstacle?
§ Where you have had success in this dimension? What were the keys to that success? What helped you get there?
§ What are you going to do next week and in the next 90 days? What actionable steps will you take?