Tuesday, 5/27/14 Adam Manzanares, Research Staff Member, HGST

Bio: Adam Manzanares received the BS degree in computer science from the New Mexico Institute of Mining  and Technology, in 2006, and the PhD degree in computer science from Auburn University, Alabama, in 2010. From 2010 to 2012 he was a postdoctoral researcher at the Los Alamos National Laboratory working on storage systems for high performance computing platforms.  After this he was an Assistant Professor at California State University-Chico. Currently he is employed by HGST in San Jose, California.

Title: System Software Updates For Shingled Magnetic Recording  

Abstract: Shingled Magnetic Recording (SMR) technologies aim to increase the areal density of hard disk drives (HDD's), This increase is achieved by narrowing data tracks of the HDD, which demands smaller magnetic poles on write heads to write data exclusively to a particular track. The poles can only be narrowed to a point, until the magnetic field generated by the poles is no longer strong enough to flip data bits on a track. To overcome this limitation, SMR drives write using a magnetic field hat overlaps many tracks. These overlapping tracks demand large contiguous writes to achieve reasonable performance, but also require changes to systems software. To limit the impact of shingling, SMR drives are broken into several zones, such that each zone can be written to without damaging any other zone. In this talk I will discuss the impact that the SMR properties have on systems software and our efforts in this area.

Background Information:

http://www.cs.cmu.edu/~./garth/papers/05_feldman_022-030_final.pdf

http://www.t10.org/cgi-bin/ac.pl?t=f&f=zbc-r01a.pdf   [ Section 4 & 5 ]

Thursday, 4/24/14 Dick Sites, Senior Staff Engineer, Google

Bio: Dick Sites is a Senior Staff Engineer at Google, where he has worked for 10 years. He previously worked at Adobe Systems, Digital Equipment Corporation, Hewlett-Packard, Burroughs, and IBM. His accomplishments include co-architecting the DEC Alpha computers, advancing the art of binary translation for computer executables, adding electronic book encryption to Adobe Acrobat, decoding image metadata for Photoshop, and building various computer performance monitoring and tracing tools at the above companies. He also taught Computer Science for four years at UC/San Diego in the 1970s. His work at Google has included web-page encoding detection, text conversion to Unicode UTF-8, and language detection. Most recently he has been working on disk servers and on CPU and network performance analysis. Dr. Sites holds a PhD degree in Computer Science from Stanford and a BS degree in Mathematics from MIT. He also attended the Master's program in Computer Science at UNC 1969-70. He holds 37 patents and is a member of the U.S. National Academy of Engineering.

Title: Identifying Dark Latency

Abstract: In astronomy and cosmology, dark matter is hypothetical matter that is undetectable by its emitted radiation, but whose presence can be inferred from gravitational effects on visible matter. [Wikipedia]

By analogy, dark latency in software is latency that is undetectable directly, but whose presence can be inferred from overall application delays. We bring four tools to bear to observe dark latency in some Google web services, then identify and fix several root causes, in low-level Google libraries and in the TCP stack.