Gold is an open source allocation system that controls project usage on High Performance Computers.

1
Allocation Manager

• Gold is an open source allocation system that
  controls project usage on High Performance
  Computers.

• Ensures resources are used according to mission
  plan
• Tracks resource utilization
• Allows for insightful capacity planning
• Facilitates resource sharing between
  organizations
• Used in production at PNNL and other sites

2
Background

• SciDAC
• Scientific Discovery through Advanced Computing
  A DOE initiative to improve the impact of
  scientific computing.
• Scalable Systems Software Project (SSS)
• Research, develop and support an integrated suite
  of systems software and tools for the effective
  management and utilization of DOEs terascale
  computational resources. It is a 5 year project
  involving over a dozen sites.
• QBank
• Gold is based on a successful program, QBank,
  that has been used for years on government and
  university computing systems.

3
Allocation Architecture
Universe of Projects
Universe of Users
Account 1234 Name Chemistry
Account 1234 Name Chemistry
Universe of Machines
4
Capacity and Workload PlanningThe Brick Approach
100 Capacity
Now
- 1 Qtr
3 Qtr
2 Qtr
1 Qtr
- 3 Qtr
- 2 Qtr
4 Qtr
5
Reservations and Dynamic Charging
Accounting Manager (Gold)
0 Deposits are made in Account
1 A Job is Submitted
2 A Quote is Requested
3 A Reservation is Made
Scheduler (Maui)
4 The Job is allowed to Start
5 The Job Completes
6 The Reservation is Removed
and a Charge is Issued
Resource Manager (PBS, LL, LSF)
Deposit
Quote
Reservation
Reservation
6
Expiring Allocations
7
Other Gold Features
Credit and Debit Models
Guaranteed Quotes
CPU
Memory
Disk
QOS
Powerful Querying
Flexible Charging
8
Other Gold Features
Historical Journaling
Transparency Features
Customizable Accounting
Dynamically Customizable
Strong Security
9
Web-based GUI
10
Facilitates Resource Sharing Between Organizations
11
Status

• Status
• Currently in second beta release
• In production use at PNNL 11.8 TF system
• and at least one other site being evaluated at
  others
• General release targeting 3rd quarter of 2005
• Download
• http//www.emsl.pnl.gov/docs/mscf/gold
• For more information
• Scott.Jackson_at_pnl.gov
• http//www.emsl.pnl.gov/docs/mscf/gold
• http//sss.scl.ameslab.gov/home.shtml
• http//www.scidac.org/ScalableSystems

12
Additional Slides
13
Typical problems faced by HPC managers
14
Nested Accounts
University
25
25
25
Biology
Chemistry
Physics
50
40
Chem 101
Chem 201
40
40
Actinides
Workshop
15
Gold Welcome Screen
16
List Accounts
17
Modify Projects
18
Account Statement