News for hackers. Stuff that matters.
Re:We Replaced an SSD with Storage Class Memory by lrem on Monday August 31, 2020 @05:02AM (#24329631)
Back in the day when I interviewed for Google I had this beautiful question. The interviewer fished for a basic distributed key-value store. I just kept coming up with single machine solution to his numbers. "No, I really can have that storage+bandwidth, here's the part number."

I'm still wondering if that interview costed me a level.

Re:We Replaced an SSD with Storage Class Memory by lmilcin on Monday August 31, 2020 @05:59AM (#24329873)
Being interviewer myself I can say interviewers tend to be stubborn. They have an excellent question (at least they think they have) that they have spend a lot of time thinking through and discussing with other candidates. Your answer causes him/her to miss the opportunity to have a real discussion and understand your knowledge of the topic.

It is better to play along and only mention the real solution at the end, to finish on a high note.

Re:We Replaced an SSD with Storage Class Memory by thechao on Monday August 31, 2020 @08:22AM (#24330694)
Alternately, the interviewer should recognize that their question is flawed & move on. I’ve had candidates nail a 30 minute discussion question with a 2 minute nonobvious answer. Shit happens: you’re not the smartest person in the room that day.
Re:We Replaced an SSD with Storage Class Memory by dijit on Monday August 31, 2020 @08:28AM (#24330741)
It should go without saying that as an interviewer; I am never the smartest person in the room. I am a person with different experiences and I need to find if you’re a fit- not if you’re unintelligent. If you got to me, you’re not unintelligent.
Re:We Replaced an SSD with Storage Class Memory by lmilcin on Monday August 31, 2020 @12:09PM (#24332821)
Tell this to my candidates:)

I have interviewed couple candidates that were offended by the fact that I am asking hard questions that I know answers to.

This resulted in remarks like "I have never seen this question on the Internet" or "So what is the answer to this question anyway?" or "If I knew questions would be this hard I would not bother to come".

Somebody explain to people that it doesn't matter how hard the questions are but how the candidates compare. I am fully prepared that the candidates study questions that are available on the Internet, I want to see how they deal with something that requires a bit more than couple hours of effort and rote memorization.

Re:We Replaced an SSD with Storage Class Memory by mushi on Monday August 31, 2020 @11:14PM (#24338191)
Is this because you were interviewing candidates who were incapable of saying “I don’t know?” That’s a good filter during any interview.
Re:We Replaced an SSD with Storage Class Memory by AnIdiotOnTheNet on Monday August 31, 2020 @09:47AM (#24331393)
Most interviewers though aren't actually any good at it and very often don't like not being the smartest person in the room.
Re:We Replaced an SSD with Storage Class Memory by lmilcin on Monday August 31, 2020 @12:15PM (#24332889)
Whether you are interviewer or interviewee, putting your ego before the goal of the interview is just going to waste the time of both parties.

I can't hire a person that can't restrain their ego for the duration of the interview and an interviewer that is focusing on anything else than figuring out if the candidate is right person for the job is causing disservice to everybody.

Re:We Replaced an SSD with Storage Class Memory by sukilot on Monday August 31, 2020 @12:54PM (#24333358)
The thing to remember is that in some companies the interviewer doesn't care about hiring and is only their because they were forced to.
Re:We Replaced an SSD with Storage Class Memory by dhosek on Monday August 31, 2020 @10:23AM (#24331727)
I guess it depends on where you're doing the interviewing. I've had surprisingly unintelligent people get to me in the interviewing stages. It amazes me how many developers, experienced even, who can't write a simple recursive function. And we're not even looking at, "can you come up with a recursive function as a solution to this problem?" we're looking at "write a recursive function to calculate the factorial" as the question.
Re:We Replaced an SSD with Storage Class Memory by joefourier on Monday August 31, 2020 @11:18AM (#24332327)
How often do recursive functions come up in your daily programming life? For me they're rare enough that it makes sense that many otherwise competent developers draw a blank when asked to write one in the context of a high-stress interview.

They may look clever, but they're often not the ideal solution compared to a simpler to understand, and often easier to optimise iterative solution. I write plenty of DSP, GPGPU programming, and computer vision, and I can't honestly remember the last time I wrote a recursive function.

Re:We Replaced an SSD with Storage Class Memory by sukilot on Monday August 31, 2020 @12:55PM (#24333368)
Recursive functions come up quite often when investigating the root cause of outages.

Unclear if that means we need candidates who are better at recursion, or better at avoiding recursion.

Re:We Replaced an SSD with Storage Class Memory by dhosek on Tuesday September 01, 2020 @09:51AM (#24341697)
Recursion is pretty fundamental in functional programming—it's preferred over loops. A smart compiler will do tail-call optimization and eliminate the overhead of increasing the stack size. Even if you're working in a non-FP development environment, knowledge of FP techniques makes one a stronger programmer. And recursion is not some esoteric technique in any event.
Re:We Replaced an SSD with Storage Class Memory by throwaway_pdp09 on Monday August 31, 2020 @12:03PM (#24332751)
> I write plenty of DSP, GPGPU programming, and computer vision ... I can't honestly remember the last time I wrote a recursive function.

Not remotely surprising is it. Hopefully if you went for a suitable job they wouldn't ask unsuitable questions. Then again, writing a recursive factorial should pain no-one.

Re:We Replaced an SSD with Storage Class Memory by dhosek on Monday August 31, 2020 @12:44PM (#24333221)
Exactly. I wasn't asking people to do something difficult. I don't think that there's a simpler recursive function than this, if you can't do this, then you don't have basic knowledge of what a recursive function is. I learned how to do this in high school.
Re:We Replaced an SSD with Storage Class Memory by lmilcin on Monday August 31, 2020 @12:21PM (#24332957)
It is not about how often they come up but whether you can think in an abstract way on a convoluted problem.

Think about IQ test which shows some abstract puzzles that have no connection with the real world. They are part of the test because they provide measurement points on tasks that require different types of abstract thinking without prerequisite knowledge.

Again, it is not possible to verify you have all the necessary knowledge for the job, nor would it be useful (no candidate knows everything for the position and if he did he would likely be overqualified).

I of course ask questions that test candidate's knowledge in the most useful areas, but at least part of the interview I want to devote to finding out how smart he/she is on abstract tasks that don't require some special knowledge.

Re:We Replaced an SSD with Storage Class Memory by sukilot on Monday August 31, 2020 @12:56PM (#24333380)
I don't want to let abstract convoluted problem solvers anywhere near my production systems
Re:We Replaced an SSD with Storage Class Memory by mlyle on Monday August 31, 2020 @01:29PM (#24333769)
You probably want to avoid people with a propensity for viewing things in abstract and convoluted terms... not to avoid people with the ability to solve abstract and/or convoluted problems.
Re:We Replaced an SSD with Storage Class Memory by dekhn on Monday August 31, 2020 @11:40AM (#24332556)
They come up about once a year for me. That's about how often I have to write a tree traversal.
Re:We Replaced an SSD with Storage Class Memory by sukilot on Monday August 31, 2020 @12:57PM (#24333392)
Why traverse a tree recursively and risk blowing your stack?
Re:We Replaced an SSD with Storage Class Memory by mlyle on Monday August 31, 2020 @01:25PM (#24333723)
You'd sure need a pretty tiny stack and/or an unbelievable amount of context on stack to worry too much about this.

If our per-callstack frame is 10k bytes, and we can spend 1 megabyte of stack on recursion (both are pessimistic).. we can go 100 levels deep. Which means we can expect to run into problems when there are more than ~5E29 elements.

This assumes we don't write the recursion in a way that the compiler can elide it entirely with tail-call optimization.

I'm not a huge fan of recursion, but let's not resort to too much hyperbole in our arguments against it. :P

Re:We Replaced an SSD with Storage Class Memory by jlokier on Monday August 31, 2020 @04:32PM (#24335623)
> If our per-callstack frame is 10k bytes, and we can spend 1 megabyte of stack on recursion (both are pessimistic).. we can go 100 levels deep. Which means we can expect to run into problems when there are more than ~5E29 elements.

This is incorrect.

Your limit is only for fully balanced trees. On a fully unbalanced tree your solution will crash at about 100 elements.

Do we see a lot of unbalanced trees? Yes, most of the time in my experience. If there's a balanced tree, there's probably a data structure and the function is already supplied. Writing a tree traversal function come up when working with unbalanced things like program ASTs, JSON/HTML/XML parsed data, Lisp-style lists, filesystem traversal, etc.

> This assumes we don't write the recursion in a way that the compiler can elide it entirely with tail-call optimization.

The compiler cannot elide away tree traversal with tail-call optimisation. Only one branch.

A really smart compiler could transform it into a loop with an explicit node-stack using an array, or avoid the stack and use in-place pointer-reversal if concurrency rules permit and there's a spare bit. But I've never seem a really smart compiler do either of those things.

Re:We Replaced an SSD with Storage Class Memory by mlyle on Monday August 31, 2020 @07:22PM (#24336902)
> Your limit is only for fully balanced trees. On a fully unbalanced tree your solution will crash at about 100 elements.

5E29 assumes not-quite-balanced. Yes, I'm assuming some kind of balanced binary tree data structure, not other things that are tree-like. There are some things that are much better than 5E29, like btrees. :P

If you have really deep unbalanced trees, you may want to have a smaller stack frame and/or pay the pain of doing things iteratively with your own stack. (Or have upward links in your tree and do it purely iteratively but slow).

> The compiler cannot elide away tree traversal with tail-call optimisation. Only one branch.

Yah, sorry. Just for search, not for full traversal.

Re:We Replaced an SSD with Storage Class Memory by joefourier on Monday August 31, 2020 @04:26PM (#24335568)
I have 20KB total RAM on the MCU I'm currently programming for. It's not too hyperbolic to imagine real-world cases where some recursive functions would be unsuitable.

They're handy for some parsing and scripting scenarios however, I'll grant you that.

Re:We Replaced an SSD with Storage Class Memory by dekhn on Monday August 31, 2020 @01:29PM (#24333770)
On my 1995 DOS-based computer, the C compiler I used at the time could handle roughly 6 levels deep.
Re:We Replaced an SSD with Storage Class Memory by mlyle on Monday August 31, 2020 @01:31PM (#24333799)
Probably you were limited to a single 64k segment for stack... still would have required a really big amount of context per call to blow stack so quickly.

Back in the mid 80's on DOS I had no problem recursing 30+ levels deep.

Re:We Replaced an SSD with Storage Class Memory by dekhn on Monday August 31, 2020 @01:09PM (#24333531)
Hmm, let's see. The last time I blew the computer's stack due to tree recursion was in 1995 (it was six levels deep). Since then, I've never seen a stack overflow due to tree recursion (presumably because none of the trees I've operated on were deeper than the hardware stack, and I switched from C to Python where the stack is far, far deeper).

If I were writing a server that needed better memory requirements, I could certainly transform my code if desired.

Re:We Replaced an SSD with Storage Class Memory by asdfasgasdgasdg on Monday August 31, 2020 @08:20AM (#24330680)
Even if you can do a particular problem on a single machine, sometimes that isn't the right call. In a work scheduled cluster environment, a task that wants the entire machine may have trouble getting a slot unless it has the priority to preempt everything else going on on that machine. We call such VMs "picky" and they don't get scheduling guarantees.
Re:We Replaced an SSD with Storage Class Memory by lrem on Monday August 31, 2020 @09:20AM (#24331153)
Heh, sure. But Borg was not part of the stated question, just of the expected answer. It wasn't even close to being needed to meet the stated parameters (IIRC the whole data set would fit into a single SSD card, so one could even have room for growth without scaling out).
Re:We Replaced an SSD with Storage Class Memory by michaelt on Monday August 31, 2020 @08:51AM (#24330921)
I'm having trouble thinking of how you'd end up with a "cluster" that couldn't provide the power of a single machine?
Re:We Replaced an SSD with Storage Class Memory by asdfasgasdgasdg on Monday August 31, 2020 @10:29AM (#24331796)
To make it concrete: consider a cluster of ten machines with 256G of ram. Team A and team B both schedule jobs with a requirement of 129G of RAM and five replicas each. The replicas get scheduled, one on each machine. Team C wants to schedule a task that takes an entire 256GB of ram. If they don't have sufficient priority to preempt the other jobs, then they will not schedule.

In real heterogenous-workload production clusters, every available machine likely has several VMs scheduled on it if the cluster isn't idle. There is never a full machine that's free unless some special effort has been taken to make it so.

Re:We Replaced an SSD with Storage Class Memory by fragmede on Monday August 31, 2020 @01:40PM (#24333912)
The knapsack problem is known to be NP complete, and your algorithm that fixates on getting jobs the size of a single machine to run, will fail to run multi-machine jobs as successfully as an algorithm that does not. It's a far more interesting algorithm to think about than sorting lists of numbers. Job priorities are easy enough to add in, but the far more practical issue is with noisy neighbors. Even limiting things to single jobs on a single machines, network and storage bandwidth has bottlenecks the cluster scheduler has to optimize for.

To make things more complicated, a long-lived cluster is going to be made up of different classes of machines, from different CPU micro-architecture, so 'single machine' is overly constraining. Eg it's not interesting that a job with a 4 MiB requirement can always run if your job needs 32 GiB.

Re:We Replaced an SSD with Storage Class Memory by throwaway_pdp09 on Monday August 31, 2020 @12:06PM (#24332776)
http://www.frankmcsherry.org/graph/scalability/cost/2015/01/...

"Rather than making your computation go faster, the systems introduce substantial overheads which can require large compute clusters just to bring under control.

In many cases, you’d be better off running the same computation on your laptop."

My limited experience fits this in that a bit of smarts on a single box beats a bunch of boxes.

(the link is a very good read BTW)

Re:We Replaced an SSD with Storage Class Memory by yencabulator on Tuesday September 01, 2020 @04:38PM (#24346778)
The biggest problem with that write-up is ignoring availability. "Fast all the way up to the crash" can be much worse than slow and steady.

Of course, for a batch job with a runtime under 11 minutes, that probably doesn't really matter too much. Just don't generalize that too much.

Re:We Replaced an SSD with Storage Class Memory by sukilot on Monday August 31, 2020 @12:53PM (#24333331)
Your interviewer may have lacked skill in guiding the process, but you could have taken the question in the spirit it was given and answered "now if the usage grows 10x as much as hardware advances" or "if we need to be resilient to hardware failures" or "if we need to roll out a logic upgrade slowly" and continued to solve the problem.

> I'm still wondering if that interview costed me a level.

Unlikely. Candidate level is decided before the interview.

Re:We Replaced an SSD with Storage Class Memory by twic on Monday August 31, 2020 @12:12PM (#24332848)
An interviewer once gave me a problem about scraping some information out of a log file. I wrote a short shell pipeline. The interviewer asked for a more complicated analysis. I wrote a longer pipeline. The interviewer added more requirements, with aggregation, state, etc in order to push me to write an actual program. I wrote an even longer pipeline.

By this point we had both realised that this was a battle of wits: could he come up with a problem that i couldn't solve with a pipeline?

At the end of the interview, i had a pipeline that took up most of a piece of A4 paper to write out. I had won the battle, and was offered the job.

Of course, i would not advise you to actually write a pipeline like that in production, but it's a fun exercise.

Anyway, the moral of this story is that if the interviewer wants you to solve a problem a certain way, and you can solve it in a simpler way, then a good interviewer will mark you up, not down. Perhaps at Google they didn't; they don't really seem like a company that has it together.

Re:We Replaced an SSD with Storage Class Memory by sukilot on Monday August 31, 2020 @12:59PM (#24333405)
Your thesis is that a company that has it together would only write software that is either a shell pipeline or can be solved in 45minues?

> would not advise you to actually write a pipeline like that in production,

So when you are interviewing for a production job, why would you fight for non-production quality solutions?

Re:We Replaced an SSD with Storage Class Memory by ur-whale on Monday August 31, 2020 @07:04AM (#24330174)
If the interviewer was gauging your political skills/ adaptability, it doesn't look like you fared too well indeed.
Re:We Replaced an SSD with Storage Class Memory by Frost1x on Monday August 31, 2020 @07:35AM (#24330380)
That seems like an excuse for poor behavior moreso than a deeper level of testing (testing when you know you're right but someone with more authority tells you otherwise). While possible, I'd have to think it's unlikely.
Re:We Replaced an SSD with Storage Class Memory by lrem on Monday August 31, 2020 @09:15AM (#24331113)
Do you test new grads for their political skills? I don't.
Re:We Replaced an SSD with Storage Class Memory by dominotw on Monday August 31, 2020 @12:02PM (#24332738)
> solution to his numbers.

Why did he have concrete numbers. Couldn't he simply have increased those numbers when you proposed single machine solution.