Where Are Tomorrow's Embedded Developers? 245
An anonymous reader writes "In a similar vein to the previous discussion about the New York professors taking Java to task for damaging Computer Science education, Mike Anderson of the PTR group wonders why it's so hard to find good embedded developers these days. 'As for today's CS programs, it seems that long gone are the computer architecture classes, writing code in assembly language (or even C at this point) and engineering software economics. In fact, a large number of CS majors apparently believe that everything can be implemented in a virtual machine and that both memory and [CPU] cycles are infinite.'"
College Classes (Score:4, Insightful)
Used to, still could, but probably won't (Score:5, Insightful)
I think there are probably more than a few people out there that have the skills, but have been moved into other paths by market forces. They will come back if you pay them enough, but that is unlikely to happen, so they will probably stay where they are.
Re:College Classes (Score:2, Insightful)
here in Sweden it does not seem as there should be any bigger shortage of embedded developers than other types.
Still embedded development are a lot more fun than doing some ordinary desktop app, or even worse webapps
Eh. (Score:5, Insightful)
There's not really time in a 4 year degree (well, along with all the other crap that goes into it) to teach someone the kinds of things you need to know to be a good business application developer and to teach someone the kinds of things you need to know to be a good embedded applications developer.
A good embedded developer needs experience with languages that run "close to the metal" of the machine, needs to know how to manage memory, needs to know how the machine architecture works, needs to know how to perform optimizations within that world, etc.
A good business applications developer needs experience with languages that abstract or hide a lot of the above details in order to let them focus on business logic, needs to know a decent amount about databases, needs to know about software architecture and design patterns, needs to know about networking, generally needs to know something about UI design, etc.
Yes, there's some overlap.
Speaking as someone with a college education emphasizing the former and a career emphasizing the latter, I'm not convinced this is a terrible thing. There are a lot more business style applications that need writing in the world than embedded applications. That specialization and the need for it I don't see going away any time soon, but it's the exception rather than the rule, and I'm not convinced there's something holier about understanding the guts of the machine than in understanding how to design a complex system for extensibility, maintainability, high availability, or whatever best suits the project.
The problem isn't virtual machines (Score:5, Insightful)
how many embedded developers are needed? (Score:4, Insightful)
The relevant question is, how many embedded-system hackers are needed? If only .1% of job opportunities are for embedded-system hackers, then there really isn't much incentive for people to learn to hack embedded systems. If embedded hacking is a lucrative field with attractive opportunities, then hackers will follow. We saw it happen with other forms of hacking, we even saw it happen with web-page hacking. If there is a need for embedded hackers, it will be filled (guided by Adam Smith's Invisible Hand).
From EE, not CS? (Score:2, Insightful)
Re:how many embedded developers are needed? (Score:3, Insightful)
Programming today's desktops? (Score:3, Insightful)
Perhaps tomorrow's embedded developers are programming today's desktops.
The desktop machine I learned to program on was an Acorn RiscPC. Later in life, I helped write the embedded firmware in the Rio Karma portable MP3 player. Rio Karma had two 90MHz ARM CPUs, 16Mbytes of RAM, and 20Gbytes of disk. That would have been one kick-ass RiscPC. If anything, programming and optimisation techniques learned on late-80s and early-90s desktops are too "embedded" for a lot of today's embedded programming.
And to those advocating hiring EE graduates for embedded programming: unless your device isn't far north of a toaster on the embeddedness scale, please also hire people with broader or higher-level software expertise for system design and architecture. It's not the same skill as squeezing one more instruction out of a loop, you find people with one skill and not the other, and any medium- or large-scale software project needs both skills (whether in the same person, or through good collaboration).
People go on about the spectrum of computing, software/assembler/logic/gates/transistors; what some don't realise is that there's a spectrum even within software. Some really good, tight, expert coders just can't see the bigger picture. I've seen medium-scale software architected by the toaster contingent: it wasn't pretty.
Peter
Re:Wouldnt'cha know.. (Score:3, Insightful)
As it is, I'm looking at doing an English-teaching-in-Japan stint. I wholeheartedly agree with your sentiment.
Re:Used to, still could, but probably won't (Score:2, Insightful)
when comparing x86,PIC or (shudder) 8051 ASM to some more modern Python programming, I dont really want to go back anymore. Even though I maintain Python code with a curses GUI Yuck!
Last thing I worked on was a GPRS/GPS combo on an ARM device, which was quite fun.
Why such a long prelude? Well, I just cant find serious employment on the field of embedded devices in Eastern Europe. Most guys who need a developer are of the type with "A good Idea" and hiring a dumbass to do the work; you just cant feed a family from those guys payments. So, I'll just stick to my Python stuff and keep integrating some Brainfuck code, just so they cant fire me anytime soon.
And at night I dream of indirect addressing, C pointers and interrupt vectors...
Re:College Classes (Score:5, Insightful)
A number of my friends took the hardware capstone where you did even more of this type of work. It still goes on, perhaps you just aren't looking in the right place? It seems the CE and EE folks take more embedded classes than CS students, so that is where you should look. CS is a huge field, and not everyone is going to follow the story's author's preference of study.
Re:College Classes (Score:5, Insightful)
I haven't seen a speck of java (or even c++) in my ECE (Electrical & computer engineering) courses, but that's all I've used in my CS courses. Furthermore, I have a lot of friends in CS that know very little about what the actual computers are doing on a bitwise level. (I had to help one of them work on bit masking last weekend.)
If you're looking for embedded people don't look at CS, look at CE.
Re:Programming today's desktops? (Score:3, Insightful)
Amen!
I more or less drifted into embedded systems work from a desktop/server background. I frequently point out to people that there's a lot of confusion about what an "embedded" system is. Some people use the term to refer to 8-bit mircocontrollers that have a whopping 4K of memory and no persistent storage. Others use the term to refer to 32-bit devices with 2-4 MB of memory and 256 MB of flash for storage. There's yet another camp that's really talking about dedicated-purpose desktop systems in a compact form factor.
The point is... there used to be a pretty well defined spectrum of what "embedded" meant. Economies of scale and new technologies have started introducing gaps into that spectrum, so there's not a single continuous spread of processors, memory and storage that qualifies as "embedded". When you can take a desktop system from 1995 and shrink it down into something like a Hammer board or a Gumstix... it may be small, it may be compact, it may be used for a dedicated purpose; but practically, you're developing on something that has massively superior capabilities when compared to the traditional idea of an embedded system. You are, in fact, developing for a desktop system - albeit a desktop system in a small package with some unusual constraints.
Re:Wouldnt'cha know.. (Score:4, Insightful)
This is a hard lesson to learn in technical/engineering jobs: be extremely careful what job you first take, because it will go a long way in determining what you do for the rest of your career. Employers always pigeon-hole people based on their previous experience. So if your last job was a
Re:CS != Programming (Score:3, Insightful)
Re:College Classes (Score:5, Insightful)
Of course!
"a large number of CS majors apparently believe that [...] both memory and [CPU] cycles are infinite.'"
That's the difference between a computer ENGINEER and a computer SCIENTIST. After all, complete turing machines *do* have infinite memory and cpu cycles.
The industry killed the golden goose. (Score:5, Insightful)
So if it doesn't have any now, then it really can't look elsewhere to blame anyone.
I started out as a R&D engineer working with video game technology, but essentially, it was all embedded work... I lived and breathed machine code and logic - to me software and hardware were one in the same, a symphony of technology with blurred distinction between the two. I remember sitting down with six spare GAL16v8s and a couple of low-power walkie talkies and a spare afternoon and built myself a radio modem for fun. That was the sort of work I used to do.
But there weren't many people like me - Assembly programmers were hard to find, even back in the 80s and most engineers fresh out of university just didn't know how to write real-time code in assembly language properly - didn't know how to write fault-tolerant code or build a spinlock as the starting point for your application. Didn't understand the necessity of understanding how many cycles an instruction took or how to watch for errors by measuring the duty cycle of the interrupt pin with a logic probe.
So the people who employed hardware designers (back then, if you knew machine code, you usually had a hand in the design of the system as well) found that it was difficult to get replacement engineers. As a result, they couldn't employ similar salary replacements and as old engineers got tired of being mistreated and poorly paid, they simply left and went off to do something different.
Industry responded to the lack of engineering by eliminating the need for the machine code engineers - they moved away from the embedded design with assembly to embedded design with C or even to outsourcing the product they needed, and the hardware got designed by dedicated hardware engineers.
Once again, any real skill in the area was lost as employers wouldn't pay for experience and the best engineers realised they would never be paid what they were worth, so left to do something else.
Then the industry got around this constraint by using really powerful embedded devices - basically a complete PC ready to run whatever PC programmers could write for it.
That's where we're at now. The skills left the industry because the industry wouldn't pay what they were worth... If you can make more money at another job (in my case at the time, selling PCs and Journalism) then why would you keep on developing hardware for a company that doesn't want to pay what you're worth?
I'm seeing the same thing now in Network Analysis... The world is full of network technicians (and I include many people who consider themselves engineers in that description, but don't really know how to actually measure things or understand the technology they work on) but has very few network engineers.
The solution for me? I got smart and moved to management.
I'm a lousy manager ( really, I suck at it ) but I try hard and for once, my contribution is recognised by the company I work for financially... And I have a family to look after.
Would I even go back to engineering or even embedded engineering?
I would love to go back, I really would. I can sit in front of circuits all day and build something and I enjoy every second of it, but I can't afford to do company critical work that won't feed my family or pay my bills.
So unless the industry is prepared to pay for skilled people, and by pay I mean pay them more than they would get being a manager or an accountant or even a journalist, then they will leave.
The other embedded engineers I know all did the same thing... One works on an offshore oilrig, another as a miner, one went to a call centre. One even opened a grocery business. These are all smart people and although they all miss working with electronics and embedded designs, they have families to feed too.
GrpA.
Re:College Classes (Score:3, Insightful)
I've yet to meet a CS grad who properly understands the difference between TCP, UDP and IP. I haven't met one who knows anything about algorithm analysis (or big O notation; I think they were all out getting a lot of Big-O's instead of studying).
If you've met anyone that has a CS degree and gives you a blank stare when you mention big-O notation, then they either failed out or graduated from Rafael Alonzo's I.T.F. Technical Computer Institute.
And, let's be realistic here. Concepts such as TCP, UDP, etc are advanced programming topics and are either taught as electives, or in post-graduate classes.
Computer Science involves a lot of math classes, and then programming classes that implement those mathematical ideas in code. Yes, the school will offer classes like HTML and AJAX, but these are electives used to broaden a student's understanding of the buzzword technology du jour (in an effort to make those students more marketable after graduation). Your post just makes you sound ignorant about what is actually offered in a *good* CS department.
Re:College Classes (Score:5, Insightful)
Re:College Classes (Score:3, Insightful)
And these guys were from KTH
Had to spend several hours teaching them about the realities of quite a few embedded projects. Such as the fact that the hardware we were targeting had a limitation of 256kB of RAM, of which 192kB needed to be available for data+processing, so no, you couldn't squeeze SSH+auto-patcher and some other stuff in there. Also, these were supposed to stand outside, in severe temperatures, so there were limits on the hardware itself. And, they were to be placed all over the country, to collect and pre-process weather and climate data. So yeah, we HAD to get it done right the first time... unless they wanted to travel around and handle reflashing manually?
Re:College Classes (Score:3, Insightful)
A starting language shall not be too unforgiving, and there is a small advantage with Java, and it is that it's very C-like. At least as long as you don't use a lot of built-in classes. If a student is limited to the "java.lang" package and isn't allowed to use anything else from the Java library it can actually provide a good learning base. But I assume that many teachers aren't into this... The really great disadvantage with Java is that the programmer relies completely on the garbage collector.
The "not to worry too much about memory or speed" is a REALLY stupid opinion. Hardware is good today in the upper-end systems, but even if the speed is doubled every 18 months or so the optimization of a program can improve speed a lot more. It's more a question of several magnitudes than percent if you find the right spot to optimize.
Learning C is really a good thing - it's small, efficient and fast. Of course - you can do stupid things in C too, but it will bite back pretty fast!
A really good course would be if students were to build their own computers and program them. Not too hard considering the components available today compared to the components available in the 80's when I built my first computer from a Z80, 2716, 6116 and a lot of TTL chips and a few other chips. I still have it, but I'm not sure if I can make it work anymore. Programmed completely in assembly.
Today it would have been either some kind of 8051 (like the nRF24E1 [nordicsemi.com]) or an ARM. Neither are too bad, and the poison of choice depends more on power consumption versus performance. The only disadvantage with modern processors are that they aren't usually DIL packaged, which means that they are a lot trickier to use when throwing together something just for learning.
Re:College Classes (Score:1, Insightful)
HA! (Score:2, Insightful)