All Tags
AWS
ai
algorithm-design
architecture
browser
cloud
cloud-efficiency
cloud-principles
cost-reduction
data-centric
data-compression
data-processing
deployment
design
documentation
edge-computing
email-sharing
energy-efficiency
energy-footprint
enterprise-optimization
green-ai
hardware
libraries
llm
locality
machine-learning
maintainability
management
measured
microservices
migration
mobile
model-optimization
model-training
multi-objective
network-traffic
parameter-tuning
performance
queries
rebuilding
scaling
services
storage-optimization
strategies
tabs
template
testing
workloads
Tactic: [Adopt Malleable Jobs]
Tactic sort:
Awesome Tactic
Type: Architectural Tactic
Category: resource-allocation
Title
[Adopt Malleable Jobs]
Description
Transition from fixed-size job scheduling to malleable jobs that can dynamically adjust their allocated resources (e.g., number of vCPUs or nodes) during execution based on system load and availability. Malleable jobs allow the system to expand or shrink job execution footprints in real time, thereby reducing energy waste, response time, and waiting time. This tactic is especially effective when paired with resource-aware monitoring and scheduling components like the Adaptive Batch Scheduler (ABS), which dynamically coordinates resource allocation to optimize performance and prevent overprovisioning or underutilization
Participant
Scientific software developers
Related software artifact
Slurm batch job scheduler, runtime resource monitor.
Context
HPC environments and scientific workflows using traditional batch systems (e.g., Slurm) where jobs are pre-assigned fixed resource allocations, often leading to underutilized or idle computational resources
Software feature
Job execution management, parallel task scheduling
Tactic intent
To reduce energy consumption and improve computational efficiency by dynamically adapting job sizes to current hardware resource availability
Target quality attribute
Energy efficiency
Other related quality attributes
< unknown >
Measured impact
waiting and response times, energy usage
Source
Stoico, Vincenzo and Voronovs, Dmitrijs and Malavolta, Ivano and Lago, Patricia, How Does Parallelism Impact the Energy Efficiency and Performance of High-Performance Scientific Software? The Case of Haddock (February 13, 2025). (DOI: http://dx.doi.org/10.2139/ssrn.5137167)Graphical representation
