2. Increased Pressure on Design
Environment
• Increased complexity in
chip design
• More simulations
requiring more compute
power
• Designers must do more
with less
• Production schedules
more important than ever
before
3. Understanding your Environment
using LSF
• Job submission and completion times
• Number of processes initiated by a user
• Number of CPU cycles used by a job
• Memory, swap, and I/O metrics used by a job
• Resource usage by user, group, or tool
4. Case Study 1
Parallel Simulation
• Design group were running simulations on a single server
platform (SUN E-450) assigned to their group
• Complete simulation took about 5 days
• LSF Base and Batch software installed
• Simulation suite took 24 hours to execute using 50 desktop
workstations which were idle at night
• Net savings of 4 days per simulation run
• Designers spend less time waiting for results
5. Case Study 2
Accurate Capacity Planning
• Two design groups competing for a SUN Enterprise Class
host (E-6500) to perform formal verification runs
• Request to purchase an additional SUN E-3500 was made
by one of the design groups
• Analysis revealed that the existing Enterprise Class host
had sufficient compute (CPU) power to run both
simulations
• A simple memory upgrade eliminated the requirement to
purchase new Enterprise Class host
6. LSF Reliability and Management
• No single point of failure in LSF cluster
• Failure of an individual CPU has little impact on design
schedules
• Failed jobs are automatically re-started on the next
available host
• Job management tasks are transparent to user
• Simple to use for developers. Only a few minutes of
training required
• Over 300 users managed with 1 full-time resource