CSE Seminar: X-SSD: A Storage System with Native Support for Database Logging and Replication

Speaker: Professor Alberto Lerner, University of Friborug, Switzerland

Abstract: SSD specialization is currently a market reality. There are different enough models of off-the-shelf SSDs that can collectively cover, for instance, a wide range of performance profiles. On a different, orthogonal axis, there are also devices geared towards enterprise or consumer use, depending on the robustness features they present. Lately, we have started to see yet another axis of specialization: functional specialization. Vendors now offer KV-SSDs and Zoned Name Spaces (ZNS) devices that have different semantics and interfaces than a traditional block device's.

In the first part of this talk, Professor Alberto Lerner will present a taxonomy that focuses on this new functional axis of specialization. The taxonomy helps identify the different techniques that go into building devices from different categories. It also reveals that there is still ample space to explore functional specialization. To illustrate this possibility, he will present the X-SSD (read "cross-SSD") in the second half of the talk. The X-SSD is designed to take a low-latency workload---e.g., a transaction log workload from a main-memory database---and provide it something akin to a unique SLA. The device can, for instance, "protect" the workload from interferences from other workloads, even more demanding ones, and can optionally replicate the log across other X-SSDs (hence its name). This device exemplifies an area of the taxonomy we call co-designed devices, which we believe to be particularly promising.

Bio: Alberto Lerner is a senior researcher at the eXascale Infolab at the University of Fribourg, Switzerland. His interests include systems that closely combine hardware and software to realize untapped performance and/or functionality. Previously, he spent years in the industry consulting for large, data-hungry verticals such as finance and advertisement. He had also been part of the teams behind a few different database engines: IBM's DB2, working on robustness aspects of the query optimizer; Google's Bigtable, on elasticity aspects; and MongoDB, on general architecture. Alberto received his Ph.D. from ENST - Paris (now ParisTech), having done his thesis research work at INRIA/Rocquencourt and NYU. He has also done post-doctoral work at IBM Research (both at T.J. Watson and Almaden Research labs).


Host: Shel Finkelstein

Join us on Zoom: https://ucsc.zoom.us/j/95906558311?pwd=WG5COENjS0dBaUQwUUNkcXB4VGlmQT09

Wednesday, March 1 at 9:20am to 10:25am

Virtual Event

Recent Activity