M.I.T. DEPARTMENT OF EECS

6.033 - Computer System Engineering Handout 5 - February 13, 2002

Assignment 2: February 19 through March 3

For Lecture: Tuesday, February 19 (Monday Schedule)

Read Chapter 2, section B of the 6.033 class notes.

For Lecture: Wednesday, February 20

Read Chapter 2, sections C and D of the 6.033 class notes.

For Recitation: Thursday, February 21

Read Ritchie and Thompson's paper, The UNIX Time-Sharing System (reading #6), and then do hands-on 2(UNIX). The paper is in your reading package, and it is also available online.

When reading the paper, keep this question in mind (You do not need to answer this question, but it may help with the Hands-on assignment).

A useful feature of the UNIX system that the "UNIX Time-Sharing System" paper highlighted was the ability to create filters and pipelines of filters (see also handson-assignment). However, this is not a feature that really exists outside of the UNIX world. Subsequently, early versions of the Mac and Windows did not allow for any useful interaction between different programs. Why do you think the designers of the Mac and of Windows ignored what was considered to be an important feature or lesson of UNIX? Do you think one or the other set of designers made a bad choice? (Hint: Consider the goals and uses of each of the systems)

For Lecture: Monday, February 25

Read Chapter 3, sections B, C and D (skip section A, we'll read that for Wednesday).

For Recitation: Tuesday, February 26

Read the paper by Pai, et al., Flash: An Efficient and Portable Web Server(reading #7), and answer the following question in a one-pager: The paper is in your reading package, and it is also available online.
The main advantage claimed by the creators of the AMPED architecture is that AMPED offers superior performance on workloads consisting of requests for both cached and non-cached data. However, in section 6.2 of the Flash paper, the authors observe that the SPED version of Flash is faster than the AMPED version on data sets that fit entirely in the cache. In your one-page paper, please address the following question: Given the current trends in pricing for RAM, do you think that the AMPED architecture is likely to have overall better performance than SPED in the future? For example, if RAM becomes cheap enough that servers commonly have tens of Gigabytes of cache, why not just use SPED?

For Lecture: Wednesday, February 27, 2002

Read Chapter 3 sections A and E of the 6.033 class notes.

For Recitation: Thursday, February 28, 2002

Read the paper by Mogul, et al., Eliminating Receive Livelock in an Interrupt-driven Kernel(reading #8) section 1 through 5, and do hands-on assignment 3 on caching. The paper is in your reading package, and it is also available online


Go to 6.033 Home Page Questions or Comments: 6.033-tas@mit.edu