PARALLEL DATA LAB 

28th Almost Annual Workshop & Retreat
November 7 - 9, 2022

INFO & Agenda

RETREAT AGENDA


Monday, November 7: Omni William Penn, Pittsburgh
8:30 am - 9:30 am Registration & Continental Breakfast
9:30 am - 10:00 am Welcome

Greg Ganger - Parallel Data Lab 2022 Retreat Welcome & Overview

10:00 am - 10:40 am New PDL Faculty Introduce Themselves

Akshitha Sriraman - Enabling Efficient, Sustainable, Equitable Web Systems

10:40 am - 12:00 pm Industry Poster Session
12:00 pm - 1:15 pm Lunch
1:15 pm - 3:30 pm Session I: Storage Device and System Interfaces

Thomas Kim - Redundancy and Availability for Arrays of Zoned Namespace SSDs

Sara McAllister - FairyWREN: A Superb Cache Co-optimized for Write-Limited Flash

Ankush Jain - CARP: An Adaptive Partitioner for Range Queries on Streaming Data

3:30 pm - 3:45 pm Break
3:45 pm - 5:00 pm Session II: Short Work-in-progress Talks + 2 More New PDL Faculty Introductions

Wan Shen Lim - Database Gyms

Juncheng Yang - C2DN: How to Harness Erasure Codes at the Edge for Efficient Content Delivery

Hojin Park - MACARON: Multi-cloud/region Aware Cache Auto-configuRatiON

More New PDL Faculty Introductions

Dimitrios Skarlatos - OS and Hardware Support for the Datacenter
Zhihao Jia
- Machine Learning Systems

5:00 pm - 6:00 pm Special Session
6:00 pm - 7:30 pm Dinner
7:30 pm - 8:30 pm Session III: Short Work-in-Progress Talks + Poster Previews

Ellango Jothimurugesan - Federated Learning Under Distributed Concept Drift

Michael Kuchnik - Validating Large Language Models with ReLM

Nj Mukherjee - Orchestrating Computation on Computational Storage Devices

PDL Students - 1-minute Intros to Non-speaker Poster Presenters

8:30 pm - 10:30 pm Reception/Poster Session I

Tuesday, November 8: Omni William Penn, Pittsburgh

8:15 am - 9:00 am Breakfast
9:00 am - 10:30 am Session IV: More Efficient ML Given Heterogeneous Clusters

Suhas J Subramanya - Apollo: Heterogeneity-aware, Goodput-optimized ML-cluster Scheduling

Daiyaan Arfeen - Exploiting Heterogeneity for Efficient Deep Learning

10:30 am - 10:45 am Break
10:45 am - 12:15 pm Session V: Redundancy and Data Placement in Cluster Storage

Francisco Maturana - Tiger: Disk-adaptive Redundancy Without Placement Constraints

Sanjith Athlur - Decoupling Data Stripes and Redundancy Groups

12:15 pm - 1:30 pm Lunch
1:30 pm - 2:30 pm Social activities or personal time
2:30 pm - 4:00 pm Session VI: ML System Efficiency

Pratik Fegade - CoRa: Ragged Tensor Compilation With Minimal Padding

Michael Kuchnik - Plumber: Diagnosing and Removing Performance Bottlenecks in Machine Learning Data Pipelines

4:00 pm - 4:15 pm Break
4:15 pm - 6:15 pm Session VII: Automate That Data System

Matt Butrovich - Tigger: A Database Proxy That Bounces With User-Bypass

Hojin Park - Mimir: Finding Cost-efficient Storage Configurations in the Public Cloud

Brian Schwedock - täkō: A Polymorphic Cache Hierarchy for General-Purpose Optimization of Data Movement

6:15 pm - 7:30 pm Dinner
7:30 pm - 8:30 pm Session VIII: Short Work-In-Progress Talks

Jiyu Hu - Rethinking Erasure-Coding Libraries in the Age of Optimized Machine Learning

Yiwei Zhao - PIM-tree: A Theoretically and Practically Efficient Index for Processing-In-Memory

Tianyu Zhang - Efficient Fault Tolerance for Recommendation Model Training via Erasure Coding

Phil Gibbons - Two Topics You MAY Find Interesting (AI-in-the-Wild & Processing-in-Memory)

8:30 pm - 10:00 pm Reception/Poster Session II

Wednesday, November 9: Omni William Penn, Pittsburgh
8:45 am - 9:15 am Breakfast
9:15 am - 10:45 am Session IX: Wake Up! With Architectural Assists

Ziqi Wang - Memento: Architectural Support for Ephemeral Memory Management

Mohammad Bakhshalipour - Bridging Robotics and Computer Architecture

10:45 am - 11:00 am Break
11:00 am - 12:30 pm Session X: Can't Get Enough Caching!

Daniel Wong - Baleen: ML-driven Flash Caching

Juncheng Yang - Segcache: Memory-efficient and High-throughput DRAM Cache for Small Objects

12:30 pm - 12:45 pm Grab lunch and back to meeting room
12:45 pm - 2:00 pm Lunch and Industry Feedback Session


Accommodations

The entire PDL Retreat will occur at the Omni William Penn Hotel in downtown Pittsburgh. Industry guests should book their room for all nights needed, using this link or by calling +1-800-THE-OMNI (1-800-843-633) and referencing "PDL Retreat" to access the special rate for the room block.

Maps

The Omni William Penn Hotel is centrally located in downtown Pittsburgh, PA. The hotel's physical address is 530 William Penn Place, Pittsburgh, Pennsylvania, 15219.

Parking

Retreat parking will be at the Omni William Penn parking facility. There are several options if you are driving your own vehicle.

There is a charge for both self- and valet-parking:

Self-parking: Self-parking is offered in the Mellon Square parking garage located across from the hotel. There are no in/out privileges for this lot.The current self-parking rate is $20.00 per vehicle, per night weekdays and $6.00 on weekends. (Prices may vary).

Valet-parking: Valet parking is available to overnight guests daily. The current overnight valet rate is $38.00, including in/out priveleges. Please note that valet can park only standard size vehicles; no oversize vehicles can be accommodated.

Transportation

There are several options available for getting to and from the retreat hotel and the airport. Please make your own transportation arrangements from the airport to your hotel.

Uber Pittsburgh - learn more

Uber is a mobile application that connects you with a driver at the push of a button.Drivers arrive curbside in just minutes, and you can track the arrival of your ride. Payment is seamlessly billed to your credit card, PayPal account, or Google Wallet at the end of your trip—no need to tip.

You can download the Uber app using the following links:
iTunes App Store, Google Play Store, or Blackberry App World.

Car Services

Airport Sedan Service(Larry Waite)
Advance reservations are required, ~$60 each way
Contact Larry Waite, 412-401-LIMO (5466)

Gateway Limo
Rates: $60 from the airport to Downtown, Oakland, Shadyside, and Squirrel Hill (sedan). The rate for a van, a 14 passenger vehicle, is $78.00 per hour. Mention CMU when making a reservation.
Call the office at 412-782-5800 or 1-800-390-1222 (phones answered 24/7) for reservations.
Email: amydemeri@comcast.net

Harper's Transportation
Harper1084@verizon.net - 412-531-1940

Classy Cab – SUV cabs
http://www.classycab.com
Phone 412-322-5080
Fax 412-322-5085

Rental Car available at the airport from all the major companies.

Taxi Service (approximately $50 each way)

  • Yellow Cab: 412-321-8100
  • Cab Service: 412-855-4484

Supershuttle - book online

  • Shared van - $35.55 each way, $9 per additional passenger
  • Private van - $115.55 each way
  • Executive sedan - $73 each way


CONTACTS:


    Director, Parallel Data Lab
    VOICE: (412) 268-1297


    Executive Director, Parallel Data Lab
    VOICE: (412) 268-5485


    PDL Administrative Manager
    VOICE: (412) 268-6716