CS 395T: Computer System Performance Analysis

Spring 2026



Given below is the tentative schedule for this course

Lecture # Date Topic Reading
1 Tue, Jan 14 Introduction Notes: When should we use which operation?
Optional:
How to read papers
L1 vs L2 norms and compressed sensing
2 Thu, Jan 16 Computational complexity in practice I Notes
3 Tue, Jan 21 Computational complexity in practice II Notes
4 Thu, Jan 23 Notions of fairness I Chapter 2.2 (α-fairness) and 2.4 (NUM) of R. Srikant and Lei Y. (alpha fairness and NUM)
5 Tue, Jan 28 Notions of fairness II Dominant Resource Fairness
6 Thu, Jan 30 Notions of fairness III FairCloud (fairness is hard)
7 Tue, Feb 04 Load balancing Valiant routing in a switch
Valiant routing in Google's datacenter
Optional:
Original paper by Valiant
8 Thu, Feb 06 Power of two choices Empirical blog
Survey of theory
Optional:
Sparrow scheduler
9 Tue, Feb 11 Overload control Overload control (Breakwater)
10 Thu, Feb 13 Process scheduling I Work stealing theoretical analysis
Optional:
Cilk programming model
Empirical analysis that concludes work stealing is best
11 Tue, Feb 18 Process scheduling II Decades of wasted cores in Linux
More bugs discovered through verification (section 5 only)
Optional:
Scheduler in Linux v4.6.8.1
12 Thu, Feb 20 Caching replacement policies I Tim Roughgarden's beyond worst-case lecture 3
Tim Roughgarden's beyond worst-case lecture 4
13 Tue, Feb 25 Cache replacement policies II Caching with delayed hits
14 Thu, Feb 27 Incorporating ML in systems I ML improves the average. Theory bounds the worst
15 Tue, Mar 04 Incorporating ML in systems II Traffic engineering by using ML to solve LPs faster
Optional:
Solving computationally hard problems with AlphaZero
16 Thu, Mar 06 Signal processing tasting menu I Rule: use sinusoids
Learn the rule to break it
17 Tue, Mar 11 Signal processing tasting menu II Fundamental Limits (read before 7.1.1 and skim the rest)
Use the limits
18 Tue, Mar 25 Congestion control I AIMD analysis
Delay based multi-bottleneck CC (pay attention to IIIB)
19 Thu, Mar 27 Guest lecture by Tegan Wilson Oblivious reconfigurable networks (sections 3, 4.3, 4.4, and 4.5 are optional)
Real-world version of the above idea in practice
20 Tue, Apr 01 Congestion control II/Performance Verifiaction I Everything is broken
Everything is unfair
21 Thu, Apr 03 Performance verification II How broken are optimizers?
22 Tue, Apr 08 Performance verification III How complex is my distributed system?
23 Thu, Apr 10 Hardware abstractions for performance I Hardware for sparse linear algebra
24 Tue, Apr 15 Abstractions that aid performance II PIFO
Approximate PIFO
25 Thu, Apr 17 Project presentations I --
26 Tue, Apr 22 Project presentations II --
27 Thu, Apr 24 TBD