What goes forward never back?
What’s only coming, or has just past?
What goes loopy when you take crack?
What dilates when you go fast?
– some time series guy
$ git push origin hamster
– same guy
Professor: Prof. Stan Zdonik
TA: Franco Solleza
Mondays, 15.00 - 17.20, First class is on Monday, January 25, 2021
Zoom Link: https://brown.zoom.us/j/92550023931
Questions? Email Franco (firstname.lastname@example.org)
If you are requesting an override code, please do so on CAB. However, also email me and Prof. Zdonik. We’ll be sending them out all at once.
Please fill out this form just to be sure we get your request covered.
Yes. Just as it’s all just bits and bytes, it’s all just time series also. This class will be about modern Time Series Management Systems (TSMS) and the current research around this area. It will focus on three topics that give you an overview of the landscape:
Time Series Storage: We’ll read about storage engines used in industry and and proposed in academia. The fundamental problems of time series storage: how to store so much data, how long to store the data, and how much of the data store? We’ll learn about how different storage engines address the problem and compare their approaches, benefits, and drawbacks.
Time Series Analysis: Given some short time series, can we find the most similar segment(s) of a larger time series? We’ll take a look at the algorithms and data structures that allow us to ask these questions. We’ll identify different approaches to this problem and critique the current state of the art.
Observability: Modern microservices infrastructure are extremely complex. Observability is the process of giving visibility into the workings of this infrastructure. In observability, there are three main types of data: metrics, logs, and traces. The common dimension for all of these data is time. We will discuss the use of metrics and traces in Observability, what’s been done in academia and industry, and what possible lines of research there are in the future.
If you want extra resources or to schedule a meeting, email Franco!
You’ll also be working on projects that involve modern or even novel techniques to address a research question. We provide a systems project,
Systems Project: In this project, you will implement a TSDB in Rust. It will support
writing entries into persistent storage, reading and filtering these entries similar to a
SELECT with a
WHERE clause, and simple aggregations. We will benchmark your project
using synthetic data we will make available. Details can be found
Analytics Project: In this project, you will implement a time series nearest-neighbor classifier. This requires finding a distance measure like Euclidean Distance or DTW, a representation, or both, with which you find nearest neighbors of a specific timeseries. We will benchmark your project using the UCR/UEA time series archive. Details can be found here.
The schedule below follows the readings for every week. Although these readings are the specifics we want to cover during the class, feel free to read more material directly related to but not included in the readings.
Reading Summaries for the day’s are to be submitted before class. These summaries are pretty free-form but generally speaking we want you to compare and contrast the approaches in each of the papers.
Submit your summaries here
During each class, two students will present on the paper readings for the week.