<html>
<head>
<meta http-equiv="Content-Type" content="text/html; charset=us-ascii">
</head>
<body>
<div class="" style="word-wrap:break-word; line-break:after-white-space">This is an announcement of Meng Wang's Candidacy Exam.<br class="">
===============================================<br class="">
Candidate: Meng Wang<br class="">
<br class="">
Date: Thursday, October 27, 2022<br class="">
<br class="">
Time: 10 am CST<br class="">
<br class="">
Remote Location: <a href="https://uchicago.zoom.us/j/9571908284?pwd=dTF4S1psWnB6M1RaNkhGanQwZTgvZz09" class="">https://uchicago.zoom.us/j/9571908284?pwd=dTF4S1psWnB6M1RaNkhGanQwZTgvZz09</a><br class="">
<br class="">
Location: JCL 298<br class="">
<br class="">
Title: Scalable Multi-level Erasure Coding in Hierarchical Data Centers<br class="">
<br class="">
Abstract: Storage systems require data redundancy strategies to protect data from drive failures. As the sheer size, scale, complexity, and layering of distributed mass-capacity storage continue their unabated growth, both the frequency of drive failures and
the time to rebuild a failed drive are growing. Therefore, data durability approaches must continue to evolve. Existing storage systems mostly adopt single-level erasure coding (SLEC) to protect data, either using network-only SLEC or local-only SLEC. However,
both SLEC approaches have their limitations, as network-only SLEC introduces heavy network traffic overhead, and local-only SLEC cannot tolerate rack failures. Moreover, the typical ways to improve the durability of SLEC introduce either higher capacity overhead
or slower encoding throughput.<br class="">
<br class="">
Accordingly, we herein propose to study multi-level erasure coding (MLEC), which is a hybrid between the two existing SLEC approaches. We envision that MLEC will provide a better trade-off between durability, capacity overhead, encoding/decoding CPU overhead,
repair over- head (network traffic and rebuild IO), and update overhead. In this project, we propose to (1). study the detailed design of MLEC, (2). build a mathematical model to compute durability for both SLEC and MLEC, (3). develop a simulator to simulate
the failures and repairs of drives in large hierarchical data centers with different erasure approaches, (4). evaluate and compare the durability and overheads of SLEC and MLEC and other erasure approaches under constant failures and correlated failure bursts.<br class="">
<br class="">
Advisors: Haryadi Gunawi<br class="">
<br class="">
Committee Members: Haryadi Gunawi, Junchen Jiang, and John Bent
<div class=""><br class="">
</div>
<div class=""></div>
</div>
<div class="" style="word-wrap:break-word; line-break:after-white-space">
<div class=""></div>
<div class=""><br class="">
<br class="">
</div>
</div>
</body>
</html>