システム・ソフトウェア
System Software / Operating Systems

Outline

• The World is Changing...What's Coming?
• So What Do We Need to Do?
• So What Do We Need to Study?

Final Projects and Grades!

I will submit my grades on July 28. Any work submitted by 17:00 on July 27 is guaranteed to be graded. Any work after that, will not!

You need to schedule an appointment with me, either face-to-face or via Skype or Polycom, for sometime between tomorrow (July 20) and 17:00 on July 27, to present your final project.

I expect a 3-5 written report (英語でも日本語でも）, and your data must present a message! Error bars and linear fits, as appropriate. What did you learn?

SOSP 2007

The Symposium on Operating Systems Principles is the premiere conference on OS research. Held once every two years in some beautiful location, it's highly competitive and prestigious. Here is a list of the sessions from 2007's SOSP:

• WEB MEETS OPERATING SYSTEMS
• CONCURRENCY
• BYZANTINE FAULT TOLERANCE
• SOFTWARE ROBUSTNESS
• DISTRIBUTED SYSTEMS
• SYSTEM MAINTENANCE
• ENERGY
• STORAGE
• OPERATING SYSTEM SECURITY

There were three best paper awards:

• Chong et al. (Cornell), Secure Web Applications via Automatic Partitioning (slides)
• Kotla et al. (UT Austin), Zyzzyva: Speculative Byzantine Fault Tolerance (slides)
• Aguilera et al. (HP Labs), Sinfonia: A New Paradigm for Building Scalable Distributed Systems (slides)

SOSP 2009

Here is a list of the sessions from 2009's SOSP:

• Scalability
• Device Drivers
• Debugging
• I/O
• Parallel Debugging
• Kernels
• Clusters
• Security

There were three best paper awards:

• David G. Andersen and company (CMU), FAWN: A Fast Array of Wimpy Nodes. See also the slides.
• Dobrescu et al. (EPFL, Lancaster and Intel), RouteBricks: Exploiting Parallelism to Scale Software Routers (slides)
• Klein et al., (NICTA), seL4: Formal Verification of an OS Kernel (slides)

Personally, I am more excited about:

• Park et al., Do you have to reproduce the bug on the first replay attempt? PRES: Probabilistic Replay with Execution Sketching on Multiprocessors, SOSP 2009
• Glerum et al., Debugging in the (Very) Large: Ten Years of Implementation and Experience, from Microsoft, on the Windows Error Reporting (WER) system, SOSP 2009
• Chou et al., An empirical study of operating system errors, SOSP 2001.
• Miller et al., An empirical study of the reliability of UNIX utilities, CACM 1990 (a classic; many bugs were fixed as a result of this, but are we doing any better as a community?)
• King et al., Debugging operating systems with time-traveling virtual machines, USENIX ATC 2005
• Larus, static analysis at Microsoft?

SOSP 2011

Papers are available here. Personally, I think 2011 looks like it was a less exciting, original conference than 2009, though of course the work is of extremely high quality.

Review

What are the basic roles of an operating system?

The Changing World

Any thought about systems today must consider the following facts:

• We have entered the Late Moore's Law Period.
• Everything is parallel, distributed, and asynchronous.
• Everything is mobile and/or ubiquitous.
• It's a post-Microsoft, Google world.
• Organizing and finding data is now harder than processing it.
• Virtualization is everywhere.

Moving in a straight line is easy, changing directions is hard!

Some Themes

Here are some ideas I consider to be important, looking forward:

• Speculation and Delay-Tolerant Computation
• Boundary-less Systems
• Move the data to the computation, or the computation to the data?
• Unpredictability in systems

How do we

• design,
• build,
• program,
• debug,
• operate and maintain,
• and use

Note that I am not condoning any "give up" attitude, or treating a computer as some mystical artifact that cannot be analyzed rigorously and scientifically!!!

Next Lecture

None! We're done!

• The Symposium on Operating System Principles

その他 Additional Information

• Rodney Van Meterのホームページ (home page)
• 村井研究室 Murai Lab