📣 黑五价提前开抢|VIP一年通行证5折,蓝莓、Offer多多同步特惠
查看: 4452| 回复: 1
跳转到指定楼层
上一主题 下一主题
收起左侧

[找工就业] 关于系统设计和behavioral questions的面试经验和资源

   
全局:

2022(4-6月)-CS硕士+5-10年 | 内推|大纽约地区 码农类General全职@

注册一亩三分地论坛,查看更多干货!

您需要 登录 才可以下载或查看附件。没有帐号?注册账号

x
本帖最后由 daisy__pupu 于 2022-5-13 01:37 编辑 .google  и

最近面senior level sde,读地里帖子受益不少,因此想写篇总结贴回馈地里。coding没啥可谈的,按各厂频率刷lc和刷地里面经就完事。这里谈一谈behavioral questions和system design。System design准备起来确实能被折磨得痛不欲生,因为问题完全看面试官经验,完全不可预测。这就导致SD需要做的准备特别多,甚至可能多于刷LC。

Behavioral Questions
. 1point3acres
最早楼主特别痛恨香蕉厂lp。但是跟香蕉厂的朋友做了几个mock以后,慢慢就理解了lp的意义所在。越senior的职位,越看沟通能力。站在面试官的角度来看,几个lp下来确实能看出情商和经验。楼主的朋友在面试别人的时候真的遇到过回答差到你不想和这个人工作的情况。而且准备完香蕉厂lp, 面任何bq都是吹灰之力。所以一切需要做bq而且会面香蕉厂的小伙伴,从香蕉厂lp开始准备就够了。
很多贴子都建议看lp相关的问题去准备几个回答。我觉得其实可以反过来。因为香蕉厂每轮都有lp,而且故事不能重复。同时一个故事可能又涵盖了几个lp,所以我觉得可以先想想自己工作里的例子,然后想想这些例子能用来讲哪些lp。最后面试的时候,准备故事清单,被问的时候挑一个讲就好了。
这里推荐一个视频。视频里不少建议都特别好,比如要注重回答能体现出你的level:
Don't Get Down-Leveled or How to Tell a Good Story (From a Principal at Amazon): https://www.youtube.com/watch?v=hU6BVxtGd5g&t=70s


System Design
. 1point 3 acres
System design必备功课肯定是DDIA这本书,以及Scott Shi的讲解视频和mock interview 视频(https://www.youtube.com/c/ScottShiCS/playlists),这里不多赘述。
.1point3acres
但是即使在做完这些基础功课以后,楼主依然毫无头绪,后来找朋友mock了两次,又看了地里两个总结贴,瞬间豁然开朗。这里一个帖子总结了sd的时候present的大致框架,另一个讨论了每块要考虑什么。.--
跳槽拿了7个senior offer,分享一下经验:https://www.1point3acres.com/bbs/thread-768263-1-1.html. ----
猴子也能懂的系统设计套路:https://www.1point3acres.com/bbs/thread-776466-1-1.html

在这个基础框架上,我觉得接下来最有用的,还是读各大厂的engineering blog, 看各家是怎么处理各种问题的。毕竟没有人比Raffi Krikorian更能回答how to design Twitter。这篇帖子的主要目的就是想分享这些文章。对于senior以上级别,grokking基本可以不看。很多问题其实已经有了非常好的解法,读完以后绝大多数的问题你都会有一个答案。楼主主要做backend data architecture,对前端非常不熟悉,不过通过这些engineering blog,很容易找到别的领域的文章。

最后,除了读文章,我觉得很有用的准备方法是,写design doc。把各种system design问题当作一个项目,按前面提到的框架去写design doc。只看文章,有的时候并不明白其设计的精妙之处。自己开始写,多问问single point of failure, scalability, exactly once等等问题,也可以请朋友看看并且提问。想不出来怎么解决这些问题时候再去看看这些文章,你就更能理解为什么这个东西是这样做的,那个设计是那样的。融汇贯通以后,新题也会是旧瓶装新酒。


Concepts. Χ

分布式事务中的一致性和隔离性你真的懂了吗: https://juejin.cn/post/6844903712368885768#heading-5
Zookeeper and Raft: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=36
This is the MIT professor's website in previous video: https://timilearning.com/


Exactly-once
Avoiding Double Payments in a Distributed Payments System: https://medium.com/airbnb-engine ... system-2981f6b070bb
kafka exactly once delivery: https://www.confluent.io/blog/ex ... ache-kafka-does-it/
kafka exactly once delivery: https://www.confluent.io/blog/en ... once-kafka-streams/
kafka design doc:https://docs.google.com/document ... ding=h.xq0ee1vnpz4o-baidu 1point3acres
transactions in kafka: https://www.confluent.io/blog/transactions-apache-kafka/
uber real-time exactly-once ad event processing: https://eng.uber.com/real-time-exactly-once-ad-event-processing/
flink exactly-once: https://nightlies.apache.org/fli ... and-fault-tolerance
. 1point3acres
. From 1point 3acres bbs
Database and CAP

Hbase consistency: https://docs.cloudera.com/HDPDoc ... %2C%20has%20had%20a,the%20most%20recently%20committed%20data.
cassandra cap guarantees: https://cassandra.apache.org/doc ... ure/guarantees.html
hbase architecture: https://data-flair.training/blogs/hbase-architecture/
Cassandra and linearizability: https://blog.yugabyte.com/apache ... unable-consistency/
uber payment strong consistency https://eng.uber.com/money-scale-strong-data/
uber Consistent Data Partitioning through Global Indexing for Large Apache Hadoop Tables https://eng.uber.com/data-partitioning-global-indexing/


Sharding

pinterst sharding: https://www.infoq.com/presentations/pinterest/


Job Queue / Job Scheduler.--

uber job queue: https://eng.uber.com/cherami-message-queue-system/. 1point3acres
dropbox job scheduler: https://dropbox.tech/infrastruct ... heduling-at-dropbox
uber kafka consumer proxy and dead letter queue: https://eng.uber.com/kafka-async-queuing-with-consumer-proxy/
uber data workflow: https://eng.uber.com/managing-data-workflows-at-scale/. 1point 3acres
uber cadence overview: https://eng.uber.com/open-source ... l-cadence-overview/ ..


Design Instagram / Twitter

twt timeline scalability: https://www.infoq.com/presentations/Twitter-Timeline-Scalability/ ..
twt timeline scalability: http://highscalability.com/blog/ ... m-active-users.html
twt real-time delivery: https://www.infoq.com/presentations/Real-Time-Delivery-Twitter/.--
Real-Time Delivery Architecture at Twitter: https://www.youtube.com/watch?v= ... _wJ&index=14&t=537s-baidu 1point3acres
Real-Time & Personalized Notifications @Twitter: https://www.youtube.com/watch?v= ... wJ&index=18&t=1407s.1point3acres
Keynote: Twitter's search architecture: https://www.youtube.com/watch?v= ... _wJ&index=28&t=997s
.1point3acresfacebook tao: https://engineering.fb.com/2013/ ... power-of-the-graph/
ins photo storage https://www.youtube.com/watch?v=34e_g-Ji_30
how to scale Instagram: https://www.youtube.com/watch?v=hnpzNAPiC0E&t=2281s
Scaling the Data Infrastructure at Instagram: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=58.google  и


Design Slack / Whatsapp. 1point 3 acres

design whatsapp: http://highscalability.com/blog/2022/1/3/designing-whatsapp.html
whatsapp Scaling to Millions of Simultaneous Connections: https://vimeo.com/44312354
fb messenger from hbase to myrocks: https://engineering.fb.com/2018/ ... timize-performance/
fb myrocks: rocksDB instaed of innoDB as storage engine: https://engineering.fb.com/2016/ ... zed-mysql-database/. Χ
how myrocks works: https://www.opensourceforu.com/2 ... cksDB%20and%20MySQL,database%20like%20RocksDB%2C%20using%20MySQL.&text=MyRocks%20requires%20C%2B%2B%20compilers%20and%20libraries%20for%20development.
fb MySQL for messaging: https://www.youtube.com/watch?v=eADBCKKf8PA
Slack architecture: https://www.infoq.com/podcasts/slack-keith-adams/
Slack Flanne edge cache: https://slack.engineering/flanne ... o-make-slack-scale/
how Slack works: https://www.youtube.com/watch?v= ... _wJ&index=10&t=307s
scaling Slack: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=12 ..
Streaming a Million Likes/Second: Real-Time Interactions on Live Video: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=13 (group message fanout)
Scaling Push Messaging for Millions of Devices @Netflix: https://www.youtube.com/watch?v= ... _wJ&index=30&t=874s (group message fanout)


Rate Limiting

aws token bucket: https://docs.aws.amazon.com/apig ... est-throttling.html
cloudflare limiter: https://blog.cloudflare.com/coun ... f-different-things/
github shareded rate limiter using redis: https://github.blog/2021-04-05-h ... rate-limiter-redis/
google rate limiting strategies: https://cloud.google.com/archite ... rategies-techniques-baidu 1point3acres
basic rate limter algos: https://medium.com/geekculture/s ... imiter-351c09a57d14.
rate limiter sticky session: impl https://docs.aws.amazon.com/elas ... ticky-sessions.html
rate limiter pros cons: https://towardsdatascience.com/d ... imiter-6351bd8762c6. 1point 3acres
flickr global incrementing id using smart sql functionality: https://code.flickr.net/2010/02/ ... -keys-on-the-cheap/
stripe rate limiter: https://stripe.com/blog/rate-limiters


Log Management

log mgmt: https://www.influxdata.com/time-series-platform/telegraf/

.
Caching

nflx app data caching: https://netflixtechblog.com/appl ... g-ssds-5bf25df851ef
fb scale memcache :https://www.usenix.org/system/fi ... final170_update.pdf
Caching at Netflix: The Hidden Microservice: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=34


Others

lyft envoy circuit breaking: https://www.youtube.com/watch?v=CQvmSXlnyeQ
uber time series query engine: https://eng.uber.com/billion-data-point-challenge/. 1point 3acres
uber big data platform: https://eng.uber.com/uber-big-data-platform/
Building a Real Time Metrics Database at Datadog: https://www.youtube.com/watch?v= ... F_wJ&index=6&t=313s
Actors or Not: Async Event Architectures: https://www.youtube.com/watch?v= ... wJ&index=31&t=2455s
Uber Marketplace: Location Serving & Storage in the Uber Marketplace: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=44
RocksDB: A High Performance Embedded Key-Value Store for Flash Storage - Data@Scale: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=47
scaling Pinterest: https://www.youtube.com/watch?v= ... _wJ&index=62&t=103s
How Uber scaled its Real Time Infrastructure to Trillion events per day: https://www.youtube.com/watch?v= ... 4l9i1OF_wJ&index=66




补充内容 (2022-05-14 00:14 +8:00):
这个帖子更多是作为地里总结贴的补充。楼主面试下来的感觉是,虽然准备的时候按框架走,每一步的可能问题都想清楚,但是面试的时候真的是questions can go anywhere。有的面试官不care back-of-envelop, 有的连network bandwidth都需要估计。比如像design youtube这么大的题,有可能每个功能都简单过一遍,也可能在video upload上追问下去没时间讨论其它功能。包括tiny-url这种经典题,怎么保证database consistency, 怎么搞transaction, 都能问很多。但是话说回来,distributed system难点都有共性,所以这篇是按概率+话题的方式整理,方便大家按需阅读。

评分

参与人数 12大米 +13 收起 理由
jhliu99 + 1 赞一个
wsc_go + 1 给你点个赞!
好好学习12138 + 1 给你点个赞!
freakpao + 1 这么有用的帖子,分给足了!
littlebenz + 2 给你点个赞!

查看全部评分


上一篇:微软口头offer后一直没有written offer
下一篇:关于温哥华巨硬offer的一个问题
全局:
可以说楼主罗列的很细致了,学习了
回复

使用道具 举报

🔗
freakpao 2022-6-22 10:47:23 | 只看该作者
全局:
这么有用的帖子,分给足了!
回复

使用道具 举报

您需要登录后才可以回帖 登录 | 注册账号
隐私提醒:
  • ☑ 禁止发布广告,拉群,贴个人联系方式:找人请去🔗同学同事飞友,拉群请去🔗拉群结伴,广告请去🔗跳蚤市场,和 🔗租房广告|找室友
  • ☑ 论坛内容在发帖 30 分钟内可以编辑,过后则不能删帖。为防止被骚扰甚至人肉,不要公开留微信等联系方式,如有需求请以论坛私信方式发送。
  • ☑ 干货版块可免费使用 🔗超级匿名:面经(美国面经、中国面经、数科面经、PM面经),抖包袱(美国、中国)和录取汇报、定位选校版
  • ☑ 查阅全站 🔗各种匿名方法

本版积分规则

>
快速回复 返回顶部 返回列表