Advanced Systems Lab

Aus VISki
Wechseln zu: Navigation, Suche

Overview

Structure

This course is based on the question whether system A or system B is better for you. It talks about throughput and response time of systems and how you can influence those (e.g. by partitioning your data).

In a second part, the course will talk about a methodology to run experiments on systems, what we actually measure and how to interpret this data. This includes a short detour into statistics (which really is not that big, if you heard about standard deviation before you should be fine).

In the third and final part the course will give a short introduction to queuing theory.

Project

Accompanying the lecture is a big threetwo-part-project done in groups of two. In the first part you're asked to implement a systemand to run experiments on this system and in the second part you're asked to model said system. This project is very time-consuming, especially the testing part. In it, you're asked to run some experiments on your system on multiple machines for which you can either use the local ETH dyrad cluster (scheduled times for every group) or the Amazon E3 cloud. Each team gets a 100$ voucher for Amazon Cloud services which should easily be enough to run all tests (even if you do a huge number of long running tests you should not use up more than about half the voucher).

Exam

The exam is very close almost unrelated to the sample exam questions (updated w.r.t. exam 2012).

Exam 2nd of Feb. 2012

Total of 15 Questions, 86 points. No partial points given for questions, and if they don't understand your solution (even if the result is correct) no points will be awarded.

They did not 100% enforce this policy though (in the question where you had to fill out a sign table (8 points), you still got 7 points if you made just one small calculation error)

Summaries

Exercises

Exams

Additional Material

Literature