.: Workshop Program

10:15 - 10:20  DaMoN 2014 Welcome

10:20 - 11:30  Keynote (details, slides)

  • The Picosecond is Dead; Long Live the Picojoule

    Christos Kozyrakis (Stanford University)

11:30 - 12:20  Paper Session I: Energy

  • Dynamic Fine-Grained Scheduling for Energy-Efficient Main-Memory Queries (slides)

    Iraklis Psaroudakis (EPFL), Thomas Kissinger (TU Dresden), Danica Porobic (EPFL), Thomas Ilsche (TU Dresden), Erietta Liarou (EPFL), Pinar Tözün (EPFL), Anastasia Ailamaki (EPFL), Wolfgang Lehner (TU Dresden)

  • Heterogeneity-Conscious Parallel Query Execution: Getting a better mileage while driving faster! (slides)

    Tobias Mühlbauer (TU München), Wolf Rödiger (TU München), Robert Seilbeck (TU München), Alfons Kemper (TU München), Thomas Neumann (TU München)

12:20 - 13:45  Lunch

13:45 - 15:00  Paper Session II: Indexing/Cracking

  • Main Memory Adaptive Indexing for Multi-core Systems (slides)

    Victor Alvarez (Saarland University), Felix Martin Schuhknecht (Saarland University), Jens Dittrich (Saarland University), Stefan Richter (Saarland University)

  • Database Cracking: Fancy Scan, not Poor Man’s Sort! (slides)

    Holger Pirk (CWI), Eleni Petraki (CWI), Stratos Idreos (Harvard), Stefan Manegold (CWI), Martin Kersten (CWI)

  • Online Bit Flip Detection for In-Memory B-Trees on Unreliable Hardware (slides)

    Till Kolditz (TU Dresden), Thomas Kissinger (TU Dresden), Benjamin Schlegel (TU Dresden), Dirk Habich (TU Dresden), Wolfgang Lehner (TU Dresden)

15:00 - 15:30  Coffee Break

15:30 - 16:45  Paper Session III: Bitmaps/New Memories

  • Vectorized Bloom Filters for Advanced SIMD Processors (slides)

    Orestis Polychroniou (Columbia University), Kenneth Ross (Columbia University)

  • HICAMP Bitmap: Space-Efficient Updatable Bitmap Index for In-Memory Databases (slides)

    Bo Wang (Stanford University), Heiner Litz (Stanford University), David Cheriton (Stanford University)

  • SOFORT: A Hybrid SCM-DRAM Storage Engine for Fast Data Recovery (slides)

    Ismail Oukid (GWT-TUD), Daniel Booss (SAP), Wolfgang Lehner (TU Dresden), Peter Bumbulis (SAP), Thomas Willhalm (Intel)

16:45 - 17:15  Coffee Break

17:15 - 18:15  Panel (details)

  • Database Machines 2.0: Doomed to fail (again) or not?

    Panelists: Peter Boncz (CWI), Ryan Johnson (University of Toronto), Christos Kozyrakis (Stanford University), Andy Pavlo (Carnegie Mellon University ), Eric Sedlar (Oracle Labs)

    Coordinator: Ippokratis Pandis (Cloudera)

18:15 - 18:30  Closing Remarks

.: Keynote

The Picosecond is Dead; Long Live the Picojoule

Christos Kozyrakis, Stanford University, http://mast.stanford.edu

Abstract

For decades, CMOS technology provided exponential improvements in transistor density and energy consumption, allowing hardware architects to focus on removing picoseconds from processor clock cycles and adding megabytes to on-chip caches. Unfortunately, we are now in a phase where transistor cost and energy consumption are barely scaling. Consequently, the new name of the game is accounting for and optimizing every picojoule the hardware consumes. This talk will describe the challenges and opportunities in designing high performance, yet energy efficient systems. Specifically, we will discuss hardware and software specialization, memory systems for data intensive applications, and raising utilization in cloud deployments. While these approaches represent a non-trivial departure from the way we design and use systems today, combined they can provide improvements equivalent to a few decades of Moore's law scaling.

You can download the keynote slides here.

Short Bio

Christos Kozyrakis is an Associate Professor of Electrical Engineering & Computer Science at Stanford University. He works on architecture, runtime management, system software, and programming models for systems ranging from cellphones to warehouse-scale datacenters. His past research includes energy-efficient data-parallel architectures, transactional memory technology, and practical hardware support for robust security abstractions. He is currently working on hardware and software techniques for resource efficient cloud computing. Christos received a BS degree from the University of Crete (Greece) and a PhD degree from the University of California at Berkeley (USA), both in Computer Science. He is a senior member of the ACM and the IEEE.

.: Panel

Database Machines 2.0: Doomed to fail (again) or not?

One can argue that now, nearly three decades after database machines quickly emerged as a trend and failed even faster, it is the most appropriate time for the community to rethink about database machines. With Moore’s Law still intact but Dennard scaling failing we see the rise of dark silicon. Aggressive microarchitectural techniques yield diminishing returns and the economies of scale, from the various cloud providers and high-end database vendors, make the case for customized hardware for database processing. At the same time, the so-called Big Data management takes place primarily on low-end (and unreliable) commodity hardware with the focus of the community to have swifted to scaling out database processing.

Why did database machines not fare very well in the 80s and 90s? What are your thoughts about database machines now and in the future? Is there a war between database machines and commodity scale out databases? Can they co-exist? Which technologies seem to be the most disruptive? We look forward to an open and lively discussion.

Panelists:

Peter Boncz, CWI
Ryan Johnson, University of Toronto
Christos Kozyrakis, Stanford University
Andy Pavlo, Carnegie Mellon University
Eric Sedlar, Oracle Labs

Coordinator:

Ippokratis Pandis, Cloudera