TE
科技回声
首页24小时热榜最新最佳问答展示工作
GitHubTwitter
首页

科技回声

基于 Next.js 构建的科技新闻平台,提供全球科技新闻和讨论内容。

GitHubTwitter

首页

首页最新最佳问答展示工作

资源链接

HackerNews API原版 HackerNewsNext.js

© 2025 科技回声. 版权所有。

FAANG promo committees are killing Kubernetes: A Short Thread

218 点作者 lolsoftware大约 3 年前

29 条评论

paxys大约 3 年前
While it is still a stretch, the title should be &quot;Google&#x27;s promo committee is killing Kubernetes&quot;, since no other FAANG-equivalent is using or contributing to Kubernetes in a meaningful way.<p>The core point:<p>&gt; It&#x27;s too indirect, fixing a bug in kube-apiserver might retain a GCP customer or avoid a costly Apple services outage, but can you put a dollar value on that? How much is CI stability worth? Or community happiness?<p>While correct, the same is true for <i>most</i> projects at these companies. Very little work that a single engineer does can be assigned a provable revenue number. How does someone working on an internal build tool get promoted? Or a new training model? A large-scale refactor of a legacy codebase? (All of these are examples of very senior promos at my own FAANG company).<p>Kubernetes is no different. An engineer has to show how the work they did first and foremost aligns with their team&#x27;s goals. If it doesn&#x27;t, then well they shouldn&#x27;t have been working on it in the first place. While working on open-source projects might be tolerated, even on company time, it makes sense that it won&#x27;t put you on a path to promotion unless there is direct benefit to the company from it.
评论 #30939889 未加载
评论 #30939259 未加载
评论 #30939649 未加载
评论 #30939618 未加载
评论 #30940839 未加载
评论 #30940283 未加载
评论 #30950423 未加载
评论 #30940434 未加载
评论 #30942411 未加载
评论 #30940776 未加载
评论 #30939294 未加载
axg11大约 3 年前
This thread is ridiculous. At FAANG, and everywhere really, promos are decided based on impact and influence. FAANG are just more likely to be working on open source projects than employees at other companies.<p>All impact is difficult to quantify, not just contributions to OSS. The only easily quantifiable achievements for an engineer are delivered projects that directly generate revenue. In the teams I’ve worked in, I’d say that covers perhaps a third of projects.
评论 #30939907 未加载
devy大约 3 年前
Brendan Burns, the Kubernetes co-founder who currently works at Microsoft Azure as a Corporate Vice President disagrees the &quot;killing Kubernetes&quot; part.[1][2]<p>But yes, Big Tech&#x27;s promo committee&#x27;s short-sighted interests do NOT align well with multi-vendor sponsored open source projects&#x27; the long term prosperity. And it&#x27;s no secrets that open source contributors are overworked. And it&#x27;s not only affect K8s, but it affects across the board.[3]<p>[1]: <a href="https:&#x2F;&#x2F;www.linkedin.com&#x2F;in&#x2F;brendan-burns-487aa590" rel="nofollow">https:&#x2F;&#x2F;www.linkedin.com&#x2F;in&#x2F;brendan-burns-487aa590</a><p>[2]: <a href="https:&#x2F;&#x2F;twitter.com&#x2F;brendandburns&#x2F;status&#x2F;1511841189686771716" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;brendandburns&#x2F;status&#x2F;1511841189686771716</a><p>[3]: <a href="https:&#x2F;&#x2F;twitter.com&#x2F;kantrn&#x2F;status&#x2F;1511791447091003395" rel="nofollow">https:&#x2F;&#x2F;twitter.com&#x2F;kantrn&#x2F;status&#x2F;1511791447091003395</a>
评论 #30939619 未加载
colinmhayes大约 3 年前
&gt; going L6-&gt;7 at Google is worth ~200k&#x2F;year, 7-&gt;8 is ~400. Similar patterns at other places. People have kids, mortgages, student loans<p>Honestly this is a bit ridiculous. If making 500k as an L6 isn&#x27;t good enough for you maybe you should be working on projects that provably bring your company revenue?
评论 #30939874 未加载
评论 #30939757 未加载
sydthrowaway大约 3 年前
As someone in Europe, I am so sick of the implied salary pissing contest in every thread.
评论 #30940028 未加载
评论 #30943721 未加载
datalopers大约 3 年前
Google created Kubernetes so potential competitors are dragged down with immense technical debt, of course they’re not going to award internal advancement of it.
评论 #30939777 未加载
hintymad大约 3 年前
This is not really just the problem in Google. We might as well claim that promotion processes in big companies promote promotion-oriented work (yes, I intentionally use the word promotion this many times). The result is title inflation, lots of good employees leaving, and superfluous yet mediocre projects. I wouldn&#x27;t say only unqualified people get promoted, though. Big companies have very different dynamics than small ones. Some people are just good at navigating big companies and are capable of aligning multiple organizations, for good or for bad. Big companies need such talent as well. As for individuals, the reward of getting promoted is not worth the effort, at least statistically. A much better alternative, is focusing on solving truly interesting and meaning problems in a blow-out company. The financial return and title bumps will follow naturally.
mdaniel大约 3 年前
<a href="https:&#x2F;&#x2F;threadreaderapp.com&#x2F;thread&#x2F;1511791378497384454.html" rel="nofollow">https:&#x2F;&#x2F;threadreaderapp.com&#x2F;thread&#x2F;1511791378497384454.html</a>
wink大约 3 年前
That&#x27;s the most ridiculous bubble valley thing I read in a while, coming from a non-k8s using non-FAANG-employed European.<p>Which isn&#x27;t meant dismissive of the problem at hand, I can totally see how it&#x27;s happening and how it might be bad for the people involved in the project.<p>To phrase it differently - if the project is dying because it hinges on people working on a promotion that gives them a &#x2F;bigger paycheck increase than my yearly income&#x2F; - well, good riddance. I&#x27;m staying on my side of the fence and continue to not buy a new Tesla every year.
评论 #30944117 未加载
zozbot234大约 3 年前
Kubernetes&#x27; biggest problem by far is its mess of accumulated technical debt. FAANG promo committees very likely undervalue FLOSS contributions, but this hits all of FLOSS pretty much equally; it hardly explains why k8s specifically would be in so much trouble.
评论 #30939782 未加载
likpok大约 3 年前
Kubernetes doesn&#x27;t get patches from hyperscale companies because kubernetes is not really hyperscale software. If I recall correctly (probably don&#x27;t), clusters can handle tens of thousands of container hosts. That&#x27;s not manageable at large scale unless you&#x27;re selling people individual kubernetes instances directly (which, to be fair, google and amazon are).<p>In my experience the thing that kills open source at big companies is the divergence of needs: The company needs massive scale software, but the community wants ease-of-use and features. Most open source models are a big drag on productivity (you can&#x27;t just refactor all the callsites to do a migration!), and it&#x27;s hard to change directions when the company&#x27;s need does (since the other contributors will likely want the original software).<p>The tailscale people have been arguing this as well (with more data and better reasoning), but fundamentally you need different software when you&#x27;re a scrappy startup or a medium size company or a giant behemoth.
评论 #30940002 未加载
评论 #30940197 未加载
reilly3000大约 3 年前
This is just the law of inertia at play. Big projects slow down. K8s isn’t just the K8s core, it’s a massive ecosystem, and when taken as a whole it’s unbelievably vast and growing at an amazing pace. As it does, the core of it moves slower, there are more stakeholders, more arguments. Small changes make for big problems and not everyone can be happy. Patches will keep coming, major changes will become eventually impossible. The death of Kubernetes makes for great clickbait, but most of us won’t see it in our lifetimes. Not when our banks run on 60 year old code. Kubernetes is too big to fail at this point.<p>As far as FAANG promo committees go, let them value what they value. Kubernetes is a direct revenue driver for many companies; it’s health is tied to billions of dollars in investments. Just because one cohort of contributors age out, cash out, or fall in love with someone doesn’t mean there is no one to take their place. The new people won’t do it the same way and that is okay, even great. I’m grateful for the vision and effort that have make K8s the platform it is, and if I can translate that into a contribution in the future I will.<p>Finally, I’ll say that people who really love working on the project may not get support to work on it full time, but then may find themselves able to retire sooner than many and have ample opportunity to contribute.
GauntletWizard大约 3 年前
I basically can&#x27;t stand the comment section of Kubernetes threads. Lots of big teams are getting big things done with k8s, and you would not know it from here.
cletus大约 3 年前
Not sure why this says FAANG because it&#x27;s talking about Google.<p>Calibration, promo committees, feedback and the like are intended to create equivalent expectations on impact across different orgs. The dirty little secret however is that at best it has limited success.<p>The best advice I can give anyone in such an organization is to be liked by your manager and their manager. If that&#x27;s true, good things will tend to happen. If it&#x27;s not, good things will be a lot less frequent.<p>Put another way: you can take the exact same set of objective facts and use them to say a person did a good job or a bad job. There&#x27;s a popular meme about feedback at Google that goes something like:<p>&gt; This project would&#x27;ve failed without this person. It failed anyway but it definitely would&#x27;ve failed without them.<p>The difference ultimately boils down to whether or not they like you.<p>Here are a few Google-specific tidbits worth knowing:<p>1. Ratings are fit to a curve across a sufficiently large pool, typically at the director level and usually over 100-150+ people. This means there will be a percentage range of people who get Meets All, Exceeds Expectations, Greatly Exceeds, etc. This is intended to stop ratings inflation;<p>2. A consequenc eof (1) is that ultimately you are competing with people in your org for those better ratings. This can create some perverse incentives and a toxic environment;<p>3. It is almost always better to let something blow up and come and fix it rather than preventing that from ever happening. The first will get you a lot of recognition. The latter will get you almost none;<p>4. Promos at Google are stack-ranked. Each committee gets 10-15 packets that will be for a particular level. The committee will rank those packets. After that the promotion target will come into play. This is set by management and was allegedly cut as a cost-saving measure when Ruth Porat came on board. If it&#x27;s 20% then the top 20% from that ranking process be promoted.<p>You will find people who serve on those committees who say this isn&#x27;t how it works and they&#x27;ll argue they&#x27;re evaluating if someone is operating at the next level or not. This is partially true. Thes packets will be divided between promote, don&#x27;t promote and on-the-bubble. The on-the-bubble group will be sufficiently large to allow for the promotion target to be met;<p>5. For SWEs. L5 is the &quot;terminal&quot; level, meaning there is an expectation of growth to that level. L3-&gt;L4 and L4-&gt;L5 once went through promo committee but now don&#x27;t. Management within orgs decide this. These too have target percentages and there have been cases where the promot rate has been too &quot;high&quot; and orgs have been told to cut back on promotions to meet the targets;<p>6. There is a massive backup at L5-&gt;L6. Because of the low target percentages the impact required keeps going up and really you need your management to really push for this to happen. There are limited slots so you may be waiting eyars and again this is why them liking you matters so much. Google is full of L6s who got promoted 5-10+ years ago that would never make the grade by today&#x27;s standard. For really old cases you can find archives of why there were promoted and you&#x27;ll find cases like &quot;promoted unit testing&quot;.<p>I say all this because the author of this thread seems to fundamentally misunderstand how this process works.
评论 #30943490 未加载
cbushko大约 3 年前
There are certain classes of work that are completely ignored by most companies such as build, automation, developer tooling and infrastructure cost cutting. These are not flashy features that you can show off to customers but they can pay in dividends year over year.<p>(Personally, I have saved millions a year in infrastructure costs and my promotions&#x2F;compensation were not close to what a sales person would get if they brought in millions of recurring revenue)
ChrisArchitect大约 3 年前
Threadrip to make &#x27;this could have been a blog post&#x27; more readable:<p><a href="https:&#x2F;&#x2F;the.rip&#x2F;@kantrn-1511791378497384454-121542889" rel="nofollow">https:&#x2F;&#x2F;the.rip&#x2F;@kantrn-1511791378497384454-121542889</a><p><a href="https:&#x2F;&#x2F;unrollthread.com&#x2F;t&#x2F;1511791378497384454&#x2F;" rel="nofollow">https:&#x2F;&#x2F;unrollthread.com&#x2F;t&#x2F;1511791378497384454&#x2F;</a>
thockingoog大约 3 年前
There&#x27;s a perception that this is true, and like all perceptions it is based in reality, but it is not the reality itself.<p>We have plenty of data that shows that people DO get promoted at all levels based on OSS work.
throwaway787544大约 3 年前
K8s is a great example of what murders open source: monolithic platforms that exist unto themselves, but have some seemingly benign model allowing arbitrary &quot;plugins&quot; to pretend it&#x27;s not really a monolith in sheep&#x27;s clothing. It looks like open source and has the right license, but it ends up just mirroring corporate interests rather than those of the Commons that the open source movement was created to support.
评论 #30942535 未加载
lumost大约 3 年前
This is the case with most major products from most major tech companies. There is a reason you don&#x27;t (often) see steady incremental progress on products, but rather big flashy releases that never quite get their bugs cleaned up unless they are wildly* successful.<p>* Where wildly is defined as a success level which would leave any rational founder very wealthy, and very content.
pjmlp大约 3 年前
Great, the sooner we get rid of that yaml spaghetti, and container everywhere, the better.
henning大约 3 年前
Hopefully Kubernetes does die. Most companies write shit slow software and then think they&#x27;re big smart boys because they have to buy $20,000&#x2F;month of servers to make it not be unusable. In many shops it could and should be replaced by a handful number of efficient servers that do not suck shit.
dbish大约 3 年前
I love open source but I’m happy if we take the credit for killing kubernetes :)
otabdeveloper4大约 3 年前
&gt; killing Kubernetes<p>Really? Finally some good news in these troubled times.
wallfacer120大约 3 年前
Oh man, another rant that boils down to &quot;big evil corporations are KILLING open source software by not supporting it in the specific ways and in the specific amounts that I want.&quot;
avs733大约 3 年前
It sounds like FAANG has reinvented tenure and promotion processes from academia…and the resulting inherent pressures.<p>A moment to reflect on Han’s critique of publish or perish academia?
MuffinFlavored大约 3 年前
What&#x27;s the alternative to k8s then?
faddypaddy34大约 3 年前
I mean you should really be demoted for continuing to push the k8s microservices crap.
eigen-vector大约 3 年前
This seems like a strictly Google only problem. And by extension this is the problem with promotion committees that are removed from day to day of the candidate getting promoted. People are ridiculously bad at measuring growth based on some obscure company wide criteria - which is abstract at best and useless at worst at a company large enough as Google.<p>I&#x27;m at AWS and the promotion process here is largely contained within the candidate&#x27;s org which is at best two levels higher but in the same business. So if your team&#x27;s business relies on contributions to FOSS, its impact is measurable and leaders and the &#x27;committee&#x27; can easily tell you if your contributions are at the next level.
评论 #30939821 未加载
评论 #30939862 未加载
评论 #30939818 未加载
hapless大约 3 年前
Go figure, contributions to a multi-vendor consortium that just barely makes a marginal profit for an already marginal business unit (GCP) are not as highly valued as things that can be directly connected to revenues. This is not surprising in the least.<p>Perhaps obviously, Mr. Kantrowitz hasn&#x27;t quit, or been poached by a competitor, so I would say Google&#x27;s (cynical) comp and promo policy is working out just great for them. Whatever personal interest keeps Mr. Kantrowitz going is working out for them!
评论 #30940001 未加载