Faisal Nawab leads the EdgeLab at the University of California, Irvine (UCI). His work aims to build Edge-Cloud Data Management (ECDM) systems to support emerging Internet of Things (IoT) and edge applications.

Hightlights

🗞️ Learn more about our new book “Consensus in Data Management” HERE

🗞️ Learn more about AnyLog – an edge-cloud database that incorporates our research work HERE

Links

Google Scholar DBLP C.V.

Edge-Cloud Data Management (ECDM) Research Summary

In EdgeLab (started in 2018), we conduct research on the design principles of building edge-cloud data management systems.

What is Edge-Cloud Data Management (ECDM)?

ECDM refers to building data systems that operate across edge nodes (close to users). These edge devices can be as small as IoT devices and as big as clusters of machines in edge/micro data centers.

Why Edge-Cloud Data Management (ECDM)?

ECDM is essential to support emerging mobile, edge, and IoT applications that require fast response times and cannot afford the long wide-area latency to communicate with a data center.

What Research is needed to enable ECDM?

In EdgeLab, we tackle the unique challenges in building ECDM systems. This includes the following thrusts (1) Distributed Data Management: how to build distributed data management protocols for edge environments with a large number of nodes. (2) Decentralization: how to build protocols to manage the unpredictable/sporadic availability and untrust of edge nodes with decentralization and blockchain technologies. (3) Energy Efficiency: how to build data systems that optimize the energy consumption of edge devices.

Thrust 1: Distributed Data Management for ECDM

To enable efficient distributed coordination in ECDM we propose the following design principles:

(1) Hierarchy and localization: Build protocols that cluster nodes together and perform distributed consensus and coordination in a locality-aware manner

Sample publications: Dynamic Paxos published in SIGMOD’18 and Blockplane published in ICDE’19

(2) Indexing and transaction processing that span edge and cloud machines: propose data management foundations that allows managing the asymmetry between edge and cloud machines to utilize both in the same design

Sample publications: CooLSM published in ICDE’21 and Croesus published in ICDE’22

Thrust 2: Decentralization for ECDM

Edge nodes can be untrusted and unpredictable. We build protocols to tolerate untrust and unpredictability.

(1) Edge-Cloud byzantine protocols: we build protocols to enable using untrusted edge nodes and tolerate malicious/arbitrary behaviour using byzantine agreement

Sample publications: AnyLog published in CIDR’20, WedgeChain published in ICDE’21, TransEdge published in EDBT’23, and ServerlessBFT published in ICDE’23.

(2) Blockchain-based protocols for ECDM: we build protocols that leverage public blockchains for their immutability and trust characteristics to enable overcoming malicious acts from edge nodes

Sample publications: RollStore published in IEEE TKDE’24, WedgeBlock published in EDBT’23 and PeloPartition published in IEEE Blockchain’23

Thrust 3: Energy-efficient data systems for ECDM

Small edge nodes need to conserve energy. We build data structures and systems to allow energy-efficient data management on edge nodes

Memory-awareness: the main design principle we propose is to judiciously select memory segments when writing that would lead to the least memory consumption. We do this by mapping available memory segments based on their content using innovative data structures and machine learning methods.

Sample publications: Hamming Tree published in SIGMOD’23, Predict and Write published in ICDE’21, and E2NVM published in EDBT’23. A summary of the area and our work are in this paper and VLDB tutorial.

Our work in practice (AnyLog)

AnyLog is an ECDM system that targets enabling emerging IoT, mobile, and edge applications.

First paper introducing the design of AnyLog in CIDR 2020

Reflective paper about the progress of AnyLog in CIDR 2024

A high-level blog post about AnyLog