2020 · IEEE Internet of Things, GreenCom, CPSCom and SmartData
Cloud Platform Performance Evaluation Using Multi-level Execution Tracing
Evidence basis: metadata-only · Review status: catalog-reviewed; paper-author approval pending
system-tracing performance-analysis performance-engineering resource-analysis
cloud platforms multi-level execution tracing LTTng LTTng-UST OpenStack Nova Neutron QEMU KVM Open vSwitch Trace Compass live VM migration preemption VM interference
Core contribution: The paper correlates LTTng traces across OpenStack services, QEMU/KVM, network components, and host kernels to diagnose cloud-platform performance problems.
Problem and motivation
Cloud-service performance failures span application, virtualization, host-kernel, and network layers, so a single-layer view cannot reliably explain migration or resource-interference delays.
Method and contribution
The approach adds LTTng-UST Python probes for OpenStack Nova, traces QEMU/KVM and host kernels, adds Neutron/Open vSwitch tracepoints with packet identifiers, and synchronizes the streams in Trace Compass. VM activity is linked to the Nova instance, QEMU process, and kernel scheduling/preemption evidence.
Findings and evidence
The live-migration case records Nova, QEMU, controller, and source/destination host traces. Reported total migration times include 166.12 seconds in the low-interference case and 169.320 seconds with interference. Traces expose CPU preemption and co-located VM interference as causes of migration slowdown and allow packet and service behavior to be followed across layers.
Limitations and future directions
Limitations: The evaluation is a focused OpenStack/VM-migration case study and does not establish general cloud-wide overhead or portability across platforms. The conclusion presents the method as extensible rather than universally validated.
Future work: Apply the cross-layer method to other hard-to-detect cloud problems, including security flaws, using kernel, network, and application perspectives.
Sources and identifiers
- Published version published
When to cite this paper
Cite this paper when correlating OpenStack, virtualization, network, and host-kernel traces for cloud performance diagnosis.
- LTTng/LTTng-UST probes for Nova, QEMU/KVM, Neutron, Open vSwitch, and host kernels.
- Trace Compass synchronization of VM, service, packet, scheduling, and preemption evidence.
- Live-VM-migration analysis that attributes slowdown to CPU preemption and co-located VM interference.
Citation
@inproceedings{ezzatiJivan2020cloudplatform,
author = {Yves J. Bationo and Naser Ezzati-Jivan and Evan Galea and Michel R. Dagenais},
title = {Cloud Platform Performance Evaluation Using Multi-level Execution Tracing},
year = {2020},
booktitle = {IEEE Internet of Things, GreenCom, CPSCom and SmartData},
pages = {294-299},
publisher = {IEEE},
doi = {10.1109/ithings-greencom-cpscom-smartdata-cybermatics50389.2020.00063},
url = {https://doi.org/10.1109/ithings-greencom-cpscom-smartdata-cybermatics50389.2020.00063}
}Other citation formats for Word and reference managers
Bationo, Y. J., Ezzati-Jivan, N., Galea, E., & Dagenais, M. R. (2020). Cloud Platform Performance Evaluation Using Multi-level Execution Tracing. In IEEE Internet of Things, GreenCom, CPSCom and SmartData (pp. 294-299). https://doi.org/10.1109/ithings-greencom-cpscom-smartdata-cybermatics50389.2020.00063Y. J. Bationo, N. Ezzati-Jivan, E. Galea, and M. R. Dagenais, "Cloud Platform Performance Evaluation Using Multi-level Execution Tracing," in IEEE Internet of Things, GreenCom, CPSCom and SmartData, pp. 294-299, 2020, doi: 10.1109/ithings-greencom-cpscom-smartdata-cybermatics50389.2020.00063