Cheriton School of Computer Science PhD student Brad Glasbergen, recent CS undergraduate Fangyu Wu and Professor Khuzaima Daudjee have received the Best Demonstration Award at the 2021 ACM SIGMOD/PODS Conference for Dendrite, a proof-of-concept system they developed that allows data systems to adapt easily to changes in workload and system behaviour.
This year’s ACM Special Interest Group on Management of Data conference was hosted virtually from June 20 to 25, 2021 in Xi’an, Shaanxi, China.
L
to
R:
PhD
student
Brad
Glasbergen,
recent
CS
graduate
Fangyu
Wu,
and
Professor
Khuzaima
Daudjee
Brad
Glasbergen
studies
distributed
systems,
with
a
primary
focus
on
leveraging
machine
learning
techniques
to
improve
performance
through
understanding
system
behaviour
for
data
storage
and
management.
Fangyu
Wu
(BCS
’20)
is
a
recent
graduate
from
the
Cheriton
School
of
Computer
Science
who
works
on
performance
analysis
of
complex
data
systems.
Cheriton
School
of
Computer
Science
Professor
Khuzaima
Daudjee
is
interested
in
systems-oriented
research,
and
works
at
the
intersection
of
distributed
systems
and
data
management,
particularly
on
building
large-scale
systems,
storage
and
infrastructure
in
the
cloud
and
on
modern
hardware.
Data systems need to respond to fluctuations in client workloads, as variability places stress upon them, ties up system resources and causes performance to degrade. Typically, a system administrator resolves workload issues to deliver high performance by vigilantly configuring and monitoring the system. To help reduce the burden on system administrators, adaptive systems are emerging that extract important features and metrics from a system to determine how the system is performing and to adjust storage and processing techniques adaptively to optimize for a given workload.
However, many popular data systems do not support robust adaption capabilities. Furthermore, enhancing existing systems to support such adaption is challenging. It requires significant design and development effort, as each self-driving data system uses a custom monitoring and response approach.
Dendrite: Bolt-on adaptivity for data systems
Dendrite, the award-winning system developed by the Data Systems and Systems & Networks Groups’ researchers, addresses these challenges by integrating easily with any data system, eliminating the need for a large design and development effort. It also democratizes self-driving systems by providing a widely applicable framework to extract metrics, declaring adaption rules intuitively to determine if a reaction is warranted and, if so, to specify appropriate responses.
“Dendrite automatically extracts metrics from application debug logs and the operating system, allowing system developers and administrators to specify adaption rules using these metrics,” said Professor Daudjee. “If a rule’s conditions are satisfied, Dendrite executes a user-defined function configured by a system administrator as a response. Dendrite is linked directly into debug logging libraries and does not rely on any system-specific functionality, so any system using such a library can automatically use Dendrite’s functions to deploy self-driving components.”
Watch the research team’s demonstration video presented at SIGMOD 2021
Two for two
This is the second time that Brad and his supervisor Professor Khuzaima Daudjee received a Best Demo Award at the ACM SIGMOD conference. Last year, PhD students Brad Glasbergen and Michael Abebe along with Professors Khuzaima Daudjee, Daniel Vogel and Jian Zhao received the Best Demo Award at ACM SIGMOD 2020 for Sentinel, a tool to understand data system behaviour and performance.
To learn more about this research, please see Brad Glasbergen, Fangyu Wu, Khuzaima Daudjee. Dendrite: Bolt-on Adaptivity for Data Systems. Proceedings of the 2021 International Conference on Management of Data (SIGMOD ’21), June 20–25, 2021, Virtual Event, China.