Register a SA Forums Account here!
JOINING THE SA FORUMS WILL REMOVE THIS BIG AD, THE ANNOYING UNDERLINED ADS, AND STUPID INTERSTITIAL ADS!!!

You can: log in, read the tech support FAQ, or request your lost password. This dumb message (and those ads) will appear on every screen until you register! Get rid of this crap by registering your own SA Forums Account and joining roughly 150,000 Goons, for the one-time price of $9.95! We charge money because it costs us money per month for bills, and since we don't believe in showing ads to our users, we try to make the money back through forum registrations.
 
  • Post
  • Reply
FamDav
Mar 29, 2008


Emacs Headroom posted:

What are they issueing in RSU?

I know in the US Amazon has a company-wide salary cap of like $175k, with RSUs and bonuses (in proportion and timing designed to maximize retention) making up an increasingly large amount of comp as you climb the ladder.

it's something like 165-170 in seattle, and around 180 in sf. less elsewhere in the country/world. an sde 1 is probably making around ~10-20% of their total income off RSUs, whereas an sde 2 is closer to 40% and sde 3 is 50+.

from what i saw on glassdoor for seattle last time i looked, these ranges seem to include either pretty out of date info or people who only put in their salary.

Adbot
ADBOT LOVES YOU

FamDav
Mar 29, 2008


i dunno, how much do they need to reduce their run rate by to make it to next month

FamDav
Mar 29, 2008


Trimson Grondag 3 posted:

just curious, do they cap like that for managers/directors etc or just for high paid individuals?

Amazon caps everyone, but you canít even hit the cap until youíre senior/principal because of the salary/rsu ratio.

looking to buy this 900k place in Capitol Hill and knowing I still donít have to touch like 60% of my income is p needs suiting though

EDIT: first against the wall, etc.

FamDav fucked around with this message at 03:11 on Jan 16, 2018

FamDav
Mar 29, 2008


Star War Sex Parrot posted:

Systems-y stuff: drivers, embedded, file systems, recently expanding that to DBMSs (working on an in-memory database project currently and finding it cool)

Also doing CUDA stuff right now which is neat. Iím all over the place as long as itís C or C++

dm if youre interested in filesystems in boston, dbs/system runtimes in seattle

FamDav
Mar 29, 2008


dividertabs posted:

I like most of the questions from that list but this one:

is really strange to me. I don't see how it has any correlation with how much I'll enjoy a job. Amazon talks a lot about Leadership Principles but (in the org I worked in when I was there) any decision could be spun as following one of the principles, even if it was harmful to customers or to other teams. 'Customer obsession' sounds great but putting it on a list doesn't make it true.

I enjoy my current not-at-Amazon job but don't remember what's on the list of company values (I know there's an official list); if I was asked by a candidate today I'd say 'uhhhhhhhhhh.' I don't think that should be a red flag to candidates. I still get to build useful stuff, work with people smarter tham myself, and make deece figgies without stress.

yeah for any company greater than 50-100 people the implementation of whatever values your company has are going to vary quite a bit team to team. I'd probably ask a list of questions like

"What are your team's primary values?"

"How much would you say those values are shared with other teams? leadership?"

"How often do you ignore those values?"

FamDav
Mar 29, 2008


dividertabs posted:

I wouldn't even ask about the team values; what they supposedly value doesn't (in my experience) correlate well with how decisions are made. I would drop the two questions:

  • what are the company's primary values? what characteristics are you looking for in a candidate in relation to those primary values?
  • what's the most impressive thing you've seen out of someone else you've interviewed recently
I'd replace them with:
  • When you think about your strongest teammates [or reports when interviewing with a manager], what qualities make them successful in your org?
A bad answer is, "resolves issues quickly and without bugs" as the only important quality. A good answer also considers how they impact other developers (e.g. spreading knowledge or improving tooling), especially cross-team impact, and how they contribute to planning + design tasks. They should also be able to speak to how those qualities are measured and rewarded; this helps you figure out what managers encourage, which is more important than what they value.

so that's why you need to follow up with a question that asks them about when they've had to compromise on their values. if they can't think of a time when they've had to compromise or explain well why they never have to compromise, then thats super suspicious to me.

i do like your question tho

FamDav
Mar 29, 2008


Shaman Linavi posted:

so i guess there are now literal tournaments to get job interviews? why is touching computers so bad

edit: also editing this to say i ran into another good one
a company that had a dropdown list for college major. Computer Science, which was clearly asked for in the job listing, was not actually on the list
they also had a mandatory "Desired Salary" field that would not take 0 (but it took 1)

if you aren't located in south america or eastern europe, i dont think youre the target audience

FamDav
Mar 29, 2008


Fiedler posted:

Ah, yes, that's right. I forgot some humans enjoy living in hives.

a thing a normal person says

FamDav
Mar 29, 2008


why do you always end up working in awful situations

FamDav
Mar 29, 2008


lol if you define your job prospects by a specific language

but if you are pick English and java

FamDav
Mar 29, 2008


Progressive JPEG posted:

whoops, minor point release upgrade just wiped out the cluster, oh well

kubernetes is javascript is rails. a lot of good stuff and superficially easy to start using, marred by the lie (by omission) that it can be hard to maintain and has quite a few sharp edges. much like anything successful.

cargo culting doesnít mean you made a bad decision, it just means youíre an idiot who doesnít know why you do anything (much like windows developers)

FamDav
Mar 29, 2008


freeasinbeer posted:

Donít disagree that this is major issue with k8s, the devs basically code to a pretty far departure of what traditional sysadmins are comfortable with. But to me fighting it to fit your legacy network design is a fools errand. Iíve also gotten in some pretty nasty fights with this in the grey forums, but I tend to agree with the vision that k8s has set forth for designs and think that maybe legacy design patterns need a shake up.

FWIW I brought it up to refute the FUD that docker isnít used in production(it has warts but there not that bad) and to shed light on the fact that operations folks are trying new things all the time as well, to try and make their lives easier. Docker is one of them.

people are running containers in production happily; they arenít running docker in production with the same kind of joy. docker has encouraged a culture of never upgrading past a version that has been stable for you because who knows what theyíve broken in the intervening time. the company just canít wrap its head around the fact that backwards compatability and stability are what the paying customers want. ipv6 is more or less broke with recent versions of docker, and apparently exec (and especially docker health checks since theyíre just repeated execs) will lock up docker and make it completely unresponsive.

docker will stick around as the dev tool of choice and weíll continue to see it in build farms for a long time. docker in production is heading out the door within the next few years.

FamDav
Mar 29, 2008


Sapozhnik posted:

is it better to start with k8s and deploy something that is literally used to manage google's millions of servers when you have 1 server and 0 users or do you migrate to it painfully later on

(borg is not k8s, btw)

containers are becoming a new deployment layer, so soon rather than later your 1 server will just be your 1 application container running on some cloud provider. eventually weíll even see VPSí running super oversubscribed container platforms for deploying Wordpress and poo poo.

FamDav
Mar 29, 2008


Sapozhnik posted:

quarter mil household income

thereís quite a few people in yospos who bring home 250-500k tho

the best part is Pollyanna has already been given most of the advice you need to get there.

FamDav
Mar 29, 2008


TimWinter posted:

Hey can someone recommend me any one of the many container solutions that exist as a viable alternative to docker? I don't want to derail the thread so I'm asking for literally one example here.

so others have mentioned rkt (though imo i see it petering out), or with k8s they have cri-o available in beta that can target runc and clear containers. the reality is that you should probably be running docker in production because nothing else has been as rigorously tested up to this point.

its not so much that docker has bugs that's the problem, because everything has bugs and people generally only complain about them because they're actively using something. the reason most people are tired of docker is because of their stewardship, and how the runtime itself has become such a commodity that nobody wants a lot the stuff that docker has had shoved into it.

FamDav
Mar 29, 2008


also all your containers are going to start running in microvm's sooner or later so get loving in

FamDav
Mar 29, 2008


TerminalRaptor posted:

It's amazing how stressed I was able to get myself in undergrad. In retrospect it was silly, but at the time it felt like the world was ending. I wished I had goofed off more during college.

depending on your college, workloads have also increased over the past decade. some of it is the classes, and some is the escalating need for differentiating activities/credentials in this hosed up capitalist hellscape.

my college (after I left) had to tell professors to chill the gently caress out on their workload after a string of suicides/attempts in one semester :(.

FamDav
Mar 29, 2008


things they're probably dinging you for, in order of how likely it was the thing

* you're managing the list of open streams as a list and ordering them after every iteration instead of using a heap. you're doing O(k log k) work instead of O(log k) on every iteration.
* lineToRead is superfluous or misnamed. you can easily determine it from StreamReader.EndOfStream.
* The tuple stuff is hard to read, especially with a random primitive int and string. if you dropped lineToRead and made a "PeekableLineReader" I think the code would be easier to read.

FamDav
Mar 29, 2008


Space Whale posted:

I've been told that you should expect back and forth about it and iterate, not just do it optimally from the start.

I would be concerned about taking this approach with an assignment vs something like an hour interview session. i would kind of expect for a take-home that you iterated a few times already and produced something good/great.

quote:

There's also the fact that reading from disc is way slower than sorting 5000 items in memory. You are right but, wth?

yeah this is the lovely and annoying part of most interview questions, which is that for whatever synthetic problem they give you the scaling considerations are just as synthetic. Whatever zip file they gave you will probably see minimal performance improvement from switching from sorting to using a heap, but it's not technically the "optimal" solution.

quote:

Also, would a sortedlist or sortedictionary be best for "use a heap" but not reinvent the wheel?

no idea, I don't know c#.

FamDav
Mar 29, 2008


Shaman Linavi posted:

i think amazon did at one time, and i remember the online coding thing i did for them had a bunch of poo poo i had to agree to
i know Epic still did as of last year, and it wasnt even a coding specific platform either

you should have seen the internal e-mail thread that blew up when people found out recruiting had started that spyware coding app thing. that initiative was shut down very quickly

i get the appeal from a recruiting perspective (a way to effectively screen candidates without trying to constantly wrangle SDEs to do phone screens!) but you just explode trust by forcing spyware on people. almost better to let the cheaters through anyways and have them get filtered at a phone screen

Steve Jorbs posted:

Lol @ having a family and working at Amazon. Pick one.

I know teams/projects where it is pretty darn hard to balance work/family, but in my org (fairly large AWS service) there's quite a few people who have families of various sizes and shapes and they seem pretty content. ymmv

FamDav
Mar 29, 2008


qhat posted:

the fact that cheating is an effective method of getting an interview at amazon shows how utterly worthless 1hour online tests or whatever are at screening out candidates. spend half an hour talking to them on the phone and you'll screen out 90% of the idiots, as opposed to like 20% from an online exam.

i agree, and they still had a phone screen anyways. really it was just filtering out the bottom NN% of candidates at the expense of pissing off all of the candidates.

FamDav
Mar 29, 2008


Gazpacho posted:

BEFORE
- review the candidate's resume AND the job requisition. mark required skills on the resume and note any significant skill gaps.
- prepare and write down your questions. allocate 3-5 minutes for short answer, 10-15 for long answer/design, 20 for coding (i.e. you probably should leave the coding exercise to someone else)
  • describe some ubiquitous features of required tech (e.g. how do you define a subclass in java, what's a foreign key in SQL, where are system configuration files in Linux). these should be features that someone experienced in the tech would necessarily know from memory. I call these the "spell cat" questions.
  • design a solution to a simple problem using tech that they claim experience in that is relevant to the job (e.g. schematize a simple data model, create a "select/deselect all" UI)
  • describe a difficult work situation in their experience and how they dealt with it
  • if there are any significant skill gaps in the resume, ask whether they have experience with those.

DURING
- shake the candidate's hand, introduce yourself, and make small talk about something on their resume
- if the candidate blanks at a question, ask whether they understood. if they continue to blank, move on.
- don't ask questions touching on the candidate's race, sex, age, national origin, religion, marital status, pregnancy, citizenship status, veteran status, disability, sexual orientation or gender identity/expression (e.g. "where are you from?", "when did you graduate high school?")

AFTER
- write up feedback promptly, with a hire/don't hire decision and brief reasons

I would add to during "don't be a dick". Even if the candidate is bombing at minute 5 in a 60 minute interview, don't mentally check out.

best case, they recover and maybe you change your mind. Worst case they get rejected but feel like things went fairly and they were treated with respect.

Fiedler posted:

This is a half-step removed from asking "what's your greatest weakness."

a weakness is a self-reflective ongoing thing, whereas technical/social conflicts and challenges are events.

FamDav
Mar 29, 2008


qhat posted:

Your advice is highly specific to generalized programming roles where the team isn't fixed, like if you're hiring new grads. If you're hiring experienced people, what you want is extremely specific. I never hire someone experienced thinking "well they suck wrt the job description, but hey Paul's team might need him in 12 months".

you still shouldn't be a dick. even if they're not a fit for your specific position, they probably know someone who is and they're going to ask how your interviews go.

FamDav
Mar 29, 2008


Notorious b.s.d. posted:

you would be surprised how many sysadmins can't code

like, at all

this past phone interview was exactly that :(.

jony neuemonic posted:

"devops" as a position is a massive signal that the company doesn't actually understand what it is.

unfortunately, that's apparently every company.

on the other side of it, interviewing people who put devops on their resume has gone fairly poorly for us over the past year.

maybe the answer is most companies/candidates arenít so great

FamDav
Mar 29, 2008


Notorious b.s.d. posted:

process and formal communication are not free; they're not a tickbox on "we're a big boy company now"

they're expensive, stultifying choices that can make your business into a kafka-esque hellscape where it takes six weeks to communicate the new documents to overseas parties for committee sign-off

its actually Apache Kafka-esque

FamDav
Mar 29, 2008


Notorious b.s.d. posted:

yep that happens sometimes

sucks, don't it?

i live in SOA hell, and the robustness/responsiviness of a given service is directly proportional to how cordial your relations are with its authors/maintainers

lots of effort is put into tools and documentation and poo poo and none of it counts for a hill of beans compared to getting coffee with the right developer or manager

(let' face it humans are just monkeys who make cellphone and software)


the only way this is ever really the case is when you are working with a third party api

the real hellscape is when you are a satellite office of a larger company, and the "third party API" is some half-documented bullshit from HQ and the original authors give zero fucks about you

viewed through a certain lens, conway's law is a software-specific way of restating the theory of the firm: if you need tighter communication than is economic via formal means, you form a firm that enables it. otherwise, things happen between firms.

just because your company is dysfunctional in this particular way doesnt mean all other companies are or have to be

FamDav
Mar 29, 2008


Notorious b.s.d. posted:

i fuckin hate it when this comes up in techie forums

the job skills that you are being paid for are associated with extroversion. you are who you are, but that doesn't change the market circumstances

if people wanted to pay a weeaboo to crank out code with minimal outside interaction, they would outsource it. period.

the only thing that supports your stateside salary and carrying costs is superior interaction with stakeholders. if you're not up to the task, you will be left behind

now this is something i completely agree with. after your entry level hiring and maybe your first promotion, your ability to do introverted coding tasks becomes a smaller and smaller component of your overall value as a software developer. its replaced by your ability to lead increasingly larger groups of people on projects of increasingly complexity, while growing people's skillset at the levels below you. if all you want to do is the introverted tasks then there is probably a place for you most anywhere, but you're going to hit a job level and pay cap pretty quickly.

this is also why i'd say its best to work at the headquarters of your company unless they are really all-in on distributed/remote working. I'd be surprised if there is a real study on this, but my anecdata suggests that its 2-3x harder to get a promotion to a job level in a company location if nobody at that location has held that job title before, and its because you're missing out on the facetime and mentorship from the level above you.

FamDav
Mar 29, 2008


Progressive JPEG posted:

kafka is extremely good op

but people who use kafka are frequently p bad

FamDav
Mar 29, 2008


to give nbsd the benefit of the doubt, the version of remote work where offshoring at lowest cost isn't the usual desired outcome is a recent thing.

and every company i've seen does a cost of living adjustment for your salary, where have you seen otherwise? though that can be very different from pegging salary to the local market rate for software developers, which is another difference between hiring remote and offshoring.

FamDav
Mar 29, 2008


jit bull transpile posted:

Apple does it. It's p common because this is the United States of America.

everywhere does it. mononcqc gave an example of how to get around it because most companies won't cut your salary on a move (though its more likely if you switch countries).

the difference is the good places pay you a % of the rate in SF/Seattle/NYC, and not the rate you'd normally get in wherever the heck you are.

FamDav
Mar 29, 2008


here's the buffer calculator if you wanted to find out how much of a haircut you'd have to take to live in the same place

https://buffer.com/salary/director-of-engineering-systems/high

i appreciate the honesty in telling me I'd get paid 70% less. that takes courage

FamDav
Mar 29, 2008


ADINSX posted:

Those salaries seem pretty low across the board...

both of them are targeting median salary in SF for the position.

FamDav
Mar 29, 2008


ADINSX posted:

I was looking at the software/senior software developer salaries. Idk what directors usually make

the two red flags are:

targeting across all companies as opposed to a set of peer companies that you want to remain competitive with. thereís a lot of people with your job title equivalent that are getting paid awful wages. also, median salary and what good devs are paid do not (and often are not) the same number.

targeting median. the only people who are enticed by your salary are the ones making worse than average salary. if you want to compete for good people, you need to offer better than median.

FamDav
Mar 29, 2008


Fiedler posted:

Translation: Do not ever work for this company or use their products.

gitlab is riding high on the k8s and ďletís run everything ourselves what do those github guys know?Ē bandwagon

surprisingly they are even worse at operations and infrastructure than their customers

FamDav
Mar 29, 2008


qhat posted:

Welp just spent 3 hours on leetcode trying to come up with a good algorithm for validating a binary tree, managed it, then checked other people's solutions and realised I could've just done in order traversal and checked the current value against the previous. I guess I'll be putting off that Amazon assessment for another day.

three hours does seem long

my favorite way for validating a bst is you start at the root with a clopen range [-inf,inf), verify that the root is in the range, then split the range by the root for the left and right nodes respectively and recurse

FamDav
Mar 29, 2008


TimWinter posted:

Although, CI is continuous integration which is easy to do several times a day even with large codebases. CD is the continuous deployment part, which does suggest several daily deploys.

Which is why I've seen CI thrown around as a great option for all cases, and CD be less universal.

I still haven't heard a good answer to how you do CD when you have literally eight hours of automated tests or, good forbid, flakey tests peppered throughout your eight hour automated test suite.

flaky tests are an issue, but if you have an automated full CD deployment pipeline why does it matter if the turn around time is 8 minutes or 8 hours so long as your test coverage and deployment velocity is acceptable for what you need

FamDav
Mar 29, 2008


this aviation place isnt in oregon is it

FamDav
Mar 29, 2008



ok good because I had a job that was in exactly that domain 5 years ago and it was awful

FamDav
Mar 29, 2008


Space Whale posted:

I believe I'm at their rear end kicking competitor. Did your job name start with a J?

no this was flight optimization for the Air Force

Adbot
ADBOT LOVES YOU

FamDav
Mar 29, 2008


Notorious b.s.d. posted:

amazon is famous for violating OSHA rules with respect to toilets per worker, especially for their white collar programming hives

buzzing with the sound of men doing the pee-pee dance

this was specifically in Blackfoot, which is an awful rear end leased building that I cannot wait for us to leave. all of the amazon built buildings are much saner about everything.

the spheres are very 👌 btw, ask your lunch buddy during the interview to take you there.

  • 1
  • 2
  • 3
  • 4
  • 5
  • Post
  • Reply