This repo contains source code and other artifacts related to the paper WASL: Harmonizing Uncoordinated Adaptive Modules in Multi-Tenant Cloud Systems. WASL is a rate-adaptation based technique for runtime cross-layer coordination in multi-tenant clouds to mitigate performance interference arising due to multiple colocated adaptive applications.
DOI Reference: Zenodo
TailBench applications have been used for evaluation.
Tailbench details are here.
TailBenchMod : Updates to standard TailBench suite used for experiments
Bare-metal hardware resources are needed for Energy Monitoring and CPU Frequency Scaling for this system.
Could VMs are often not bare-metal instances; those will not work for this system.
Many Intel processors have active mode enabled by default for intel_pstate driver.
For managing CPU frequencies in this work, passive mode is needed; hardware managed P_State (HWP) support has to be disabled.
echo "passive" > /sys/devices/system/cpu/intel_pstate/status
Further information about CPU performance scaling driver can be referenced here.
Please review the setup instructions available here carefully.
/ : Overall Runtime System
|-- apto-tailbench-apps : Wrapper/Profiler for Application/System
|-- apto-tailbench-apps/scripts : Helper scripts for profiling/parsing
|-- apto : Processing and Activation,
coordination with the Adaptation Module
|-- OptimizingController : Adaptation Module (local)
|-- PoleAdaptation : WASL-based Multi-Module Adaptation (global)
|-- Plots : Scripts related to some results
|-- Infra : Experimented Set-up on an AWS instanceEach folder has its own Readme file.
The following paper can be cited:
@inproceedings{DBLP:conf/wasl/PervaizDKSH26,
author = {Ahsan Pervaiz and
Anwesha Das and
Vedant Kodagi and
Muhammad Husni Santriaji and
Henry Hoffmann},
title = {{WASL:} Harmonizing Uncoordinated Adaptive Modules in Multi-Tenant
Cloud Systems},
booktitle = {Proceedings of the 17th {ACM/SPEC} International Conference on Performance
Engineering, {ICPE} 2026, Florence, Italy, May 4-8, 2026},
pages = {301--313},
publisher = {{ACM}},
year = {2026}
}