System Design Interview budget constraints

Mar 5, 2021 7 Comments

I see sample system design interviews and I'm appalled by the lack of regards to resource planning that the supposed "solutions" give.

"Just shard these jobs across 500 machines then replicate them and put them behind a reverse proxy and several layers of caches to provide reliability and lower latency to users on multiple geographical regions bla bla bla". These kinds of answers are very detached from reality where managers actually need to fight for resources annually/quarterly just to get CPU, storage, and even headcount to develop all of these. Not to mention the number of SRE/production engineers you have to pay just to give these two 9's of reliability in all time zones.

I even hear some bullshit about storing data in Spanner / Bigtable or some other paper they read in VLDB. Yeah good luck with that because that's another set of engineers and SRE's that you have to pay just to make your system work.

So I guess my question is: Is there a limit on how candidates can get crazy with their system design?

#systemdesign #interview

TC: 170k
YoE: 8
Loc: Hong Kong

comments

Want to comment? LOG IN or SIGN UP
TOP 7 Comments
  • New / Eng
    NotFermat

    New Eng

    NotFermat
    Asking the real questions. Most companies don’t have $200B sitting in their account in order to hire head count and server resources to implement these architectures. Heh, most of the time I’m just told work with the resources you have and make the application distributed and fault-tolerant. I would love to see system designs where you have to use a relational database (Oracle, MySQL) and only have a RESTful API and front-end to work with.
    Mar 5, 2021 2
    • Google
      lx500

      Go to company page Google

      lx500
      Just use spanner
      Mar 5, 2021
    • New / Eng
      NotFermat

      New Eng

      NotFermat
      Thanks for the option. However, most large companies I have worked for need to redesign their entire application portfolio to work in the cloud. Even if I moved one database to Spanner, we have lots of database links to other databases (Oracle) so how does that work with Spanner? Can Spanner easily call out an on-premise database server, assuming firewall rules are in-place?

      The main point is the adoption of cloud technologies requires a 5-10 year roadmap as the F500 companies barely hire enough developers to handle normal operations. I would love to work somewhere that keeps up with current trends (e.g. doesn’t have 15 year-old legacy apps that nobody is maintaining).
      Mar 5, 2021
  • Amazon
    dojabja

    Go to company page Amazon

    dojabja
    For a system design round, expected DAU / scale / volume has to be stated prior to actual design.

    This can be served as your limit. Do not leave your candidate clueless as your interviewer should be more experienced than the candidate.
    Mar 5, 2021 0
  • Google
    tsVr63

    Go to company page Google

    tsVr63
    In practice you will end up using GCP for spanner anyway.
    Mar 5, 2021 0
  • Google
    lx500

    Go to company page Google

    lx500
    170k TC in HK? That's nice. What's your level?
    Mar 5, 2021 1
    • I'm a VP which is equivalent to Senior SWE at Google. Life can comfortable for finance people in HK (many crypto exchanges, investment banks, hedge funds, prop trading). Tech people on the other hand ....
      Mar 5, 2021