Computer System Engineering | Hacker News Reader

Computer System Engineering

ocw.mit.edu

323 pointshackitup735 comments

This was the most impactful class I took at MIT.

Of every class I took, the lessons from this class have helped me the most in the real world.

Algorithms, though heavily covered in lame interview tests, pale in significance to the benefit of understanding how to architect systems at a high-level.

And a knowledge of how existing computer infrastructure and platforms are architect-ed (DNS, Ethernet, etc.) provides a deep pool of successful, tested concepts for whatever new system is being built.

Highly recommend completing this course via OCW if you want to improve as an engineer.

I have recently found a strong renewed interest in the power of algorithms. Thinking about all of the sciences, computer science is still very much on the leading edge. There seems to be a lot of area left to explore with regard to applying algorithms correctly and devising entirely new ones.

For instance, the splay tree is arguably one of the more incredible data structures in all of computer science, and we didn't discover it until 1985. Think about all the undiscovered applications that may exist out there regarding such an algorithm. Contrast with how long mathematicians have had to play around with the ideas of Newton, et. al.

Another fun data structure is the skip list. This is probably the simplest probabilistic data structure you could hope to invent, and we didn't figure it out until 1989. 4 years after the splay tree. To me, this says "keep looking".

> Thinking about all of the sciences, computer science is still very much on the leading edge.

What do you mean by this?

He means there's a lot of low hanging fruit left.

There is as much "low hanging fruit" left as in mathematics, economics,control theory,analytical chemistry and so on. Like the proverbial "100 dollar bill in the sidewalk" joke, any truly impactful, low-hanging fruit would have been taken eons ago. PhD dissertations, postdoc applications and tenure case presentations would love to use them.

> left to explore with regard to applying algorithms correctly.

Yes, and even just a trade-off based on context for where it will be used.

I've realized that lots of libraries I use are generalized (by nature of being an OSS lib) so they aren't going to be hyper-refined for my particular use case.

Usually an OSS lib will be better than something you can write in a day but if something is part of your product's competitive advantage, then you will probably be able to make a more optimized version than an open source lib.

A commonly used algorithm that totally blew me away when I first encountered it was the Burrows-Wheeler transform for lossless string compression. [0] It rearranges characters in long input strings to increase compressibility by placing repeat characters closer together (you could do so with sorting too, but then lose the ability to recreate the original without associate metadata). Bloom Filters are a close second.

In terms of elegance, I believe it is a fight between CoDel [1] and Union-Find [2].

Speaking of Skip Lists, see also Fenwick Trees, not similar but you'd find them particularly interesting [3].

Also, on the topic of Splay Trees, because we are on Y Combinator, we mustn't go without mentioning Zippers [4].

[0] https://en.wikipedia.org/wiki/Burrows%E2%80%93Wheeler_transf...

[1] https://queue.acm.org/detail.cfm?id=2839461

[2] https://algs4.cs.princeton.edu/15uf/

[3] https://cp-algorithms.com/data_structures/fenwick.html

[4] https://stackoverflow.com/questions/380438/what-is-the-zippe...

There are two other versions of the course[0][1], from 2005 and 2009. This 2018 version seems to be the only one without video lectures.

Actually, looks like both versions are using the 2005 videos (but with different UI: one embeds, one links to YouTube).

[0]: https://dspace.mit.edu/bitstream/handle/1721.1/56581/6-033Sp...

[1]: https://dspace.mit.edu/bitstream/handle/1721.1/118791/6-033-...

They both seem to be missing Lectures 1,2, and 3. Not sure if that is intended, but it's a strange coincidence if it's not intended.

Secret knowledge?

Probably some copyright issues.

Ah dark magic beyond human understanding

This is the class that made my career. The idea that humanities sorts of skills – critical reading, analysis, discussion — could contribute to the work of systems engineering was a revelation. That I had the knack for it, and that it turns out to be the core work of engineering, I found later.

The THERAC paper, the X11 paper, the Kerberos sequence—still resonate 22 years later. But the big lesson wasn’t in any of the papers, but in the structure as it communicated that engineering was not a solitary enterprise.

Ever so often these courses, without video lectures, from prestigious universities come up here and gets upvoted to the front page. There will be some discussion by people who attended these courses in those universities.

What is the point of posting and upvoting these course pages here? The most charitable answer I can find is that it informs that there exists such a body of knowledge. The uncharitable one is that it lets people gloat about their alma maters here in the comments.

Any other explanations, anyone?

I personally avoid video lectures and prefer the written materials. I get that video works better for some people, but I do not understand how this can be the case: non interactive video is so much worse than text and diagrams.

Verbal communication contains a lot of information about relative importance of information

Unless you’re implying the way certain phrases are being spoken carries that information, text easily conveys relative importance as well.

I've been an enthusiastic follower of MOOCs since 2013 and it's amazing to me that there still isn't a better option than OCW for a serious, long-term learner.

Coursera and edX each had some great courses at their peak but the vast majority were lower-level, they came from a wide variety of institutions and most were time-restricted. As a result, even for subjects like mathematics or computer science there just wasn't a path for an ambitious learner to piece together the full material from a bachelor's degree. These days, even automated graders are behind the paywall and the promise of an education for all is more distant than ever.

OCW is truly an invaluable resource worth supporting and sharing!

Some courses can be of amazing quality, like 6.824

This course (at least the videos) are on YouTube for the latest iteration. I found the YouTube interface a bit more familiar for sampling the course.

https://youtube.com/playlist?list=PLrw6a1wE39_tb2fErI4-WkMbs...

Especially if you're willing/able to pay some money there's still a lot on edX etc., both individual courses and the various MicroMasters programs.

But, yes, totally free content is somewhat limited at this point and, generally, MOOCs have to be seen as a disappointment to anyone who saw them as a radical free substitute for a university education.

Perhaps interestingly, when OCW launched, MIT presented it as very much not a substitute for an MIT education but rather as raw materials for educators to assemble their own courses. Which I'm sure was at least in part a matter of internal politics but still. (And, of course, ubiquitous video of lectures wasn't a thing at the time anyway.)

Many if not most MOOCs certainly lack depth, maybe the problem is in M part? Sometimes they feel like an ad

Many of them are fairly faithful representations of the lecture components of courses from "elite" universities. There are certainly many MOOC courses I would struggle with. But, discussion boards are mostly a tire fire (in part, because of the M part as you suggest), problem sets/assignments are limited even if they're not behind a paywall, there are no recitation sessions, etc. So even if someone is motivated, which can be hard in the absence of meaningful certifications, they're just more limited than taking an actual university course.

Is it possible to piece together a full undergrad of coursework, self-study and then pass a subject GRE?

That would be a more recognizable credential than a "micromasters".

No one (except a grad school) cares what you get on a GRE test. GREs aren't that different from SATs.

Even if that were true, which I don't believe, it's still highly useful. E.g. if someone passes a GRE Physics test, they may be able to enter a graduate program with an entirely unrelated undergraduate degree instead of spending 100k getting another.

You may be different, but I'd have increased confidence in someone's general comfort with mathematical problem-solving if they posted good scores on a test covering this: https://www.ets.org/gre/subject/about/content/physics/

For computer science specifically, that's unfortunately not possible because the GRE for CS was discontinued in April 2013.

That's really a shame for ambitious learners trying to achieve more with less time, wealth or pedigree.

It completely makes sense from the university's perspective, though.

The MIT on edX courses for Probability [0], Single Variable Calculus [1], and differential equations [2] are of the absolute highest quality.

In general, I'd if there's a course by MIT on edX for the topic you wish to learn, I'd check it out.

[0] https://www.edx.org/course/probability-the-science-of-uncert...

[1] https://www.edx.org/xseries/mitx-18.01x-single-variable-calc...

[2] https://www.edx.org/xseries/mitx-18.03x-differential-equatio...

Thanks! Looking for a free version of [2], do you know where else one could find it?

Totally agree, for me online education is a bummer because it has so much potential but most of the offerings instead of being great inherited the WORST of both worlds, old-brick-and-mortar education and web technologies. So to take a course I have to first sign-up for the platform, then sign-up for the specific course, then the material is arbitrarily split in micro-segments where you have to watch the professor droning on like in person education, then you have to answer some inane quizzes. The firs section is the SAME for all the courses, introduce yourself and blah blah. At the end you realize that the actual content was like 50% of a traditional college course. Do you want some credit? You have two options, you may pay 50-100 bucks for a useless certificate or the second option and the most insulting one, you take the course from a BIG NAME UNIVERSITY and you are requested to pay a high fee but the credit given is from a third-rate-municipal college for their "general requirements" course.

In the case of OCW you are presented with THE SAME EXACT course the kids at MIT are taking and many many courses have full lecture notes, quizzes, and so on. You could argue that the experience is not the same but at least the content is, you don't have to sign up for anything, the institution is actually world-class and the cost is 0.

i love how you put it. all of this micro organization and "filing" for ease actually makes it harder to follow. Give me a 2 hour youtube videos with stamps and i'll find what im looking for

This seems a lot different than the degree I did in Oz back in the 90's, which was a 6 year course (every 6 months was mandatory work experience): Computer Systems Engineering @UTS. Their current equivalent is this sub major:

https://handbook.uts.edu.au/directory/smj03034.html

Interesting is how different the two courses are. I learnt a bunch of hardware design and build (cpu, embedded, electronics) as well as software design and build.

Looks like your course focused on real-time and embedded systems, but used the same name. MIT's course is focusing more on general-purpose servers.

I expect MIT also teaches RT/Embedded also but probably calls it something along those lines.

Both skillsets are necessary, and IMHO not enough programmers are exposed to real-time systems in school. It's a very different style of coding and all modern desktop systems would work better if their designers had a better understanding of real-time principles.

This looks very interesting. What are (or could be) some prerequisites to taking this course?

Reply on news.ycombinator.com