1
$\begingroup$

My 1st question ever on stackexchange, please be nice to me if it's too naive :)

I read the map reduce paper and i discover distributed system is expensive.

As an example in the paper, roughly speaking, approx. 1700 machines with 30000 mb/s throughput compares to non distributed single machine with 500mb/s ssd, it's 60 times through put increased with 1700 times machines! i understand achieving perfect linear scalability is impractical, since there's inherited overheads on network latency (prominent bottle neck in map reduce's case), data shuffling between machines, and synchronization issues.

But at the awe of this numbers gives me the scale feeling, I am wondering:

  1. how do we decide for example : my business is doing well enough (ex: online gambling soccer game platform) to adopt this fancy distributed system,
  2. what are some experiences of gradually integrating distributed systems in a startup company, bring in distributed system step by step, for different period of developing the business, we should design different distributed system that cost effective to that business phase, where can i find this material to read if there's ever such material? or this knowledge is mostly just in experienced people's head ?
$\endgroup$
2
  • $\begingroup$ I think this question fits better to serverfault.com $\endgroup$ Commented Dec 21, 2023 at 10:51
  • $\begingroup$ @ErelSegal-Halevi thanks this is very helpful! $\endgroup$
    – Effy
    Commented Dec 22, 2023 at 11:46

1 Answer 1

1
$\begingroup$

The main reason to use a distributed system is because we have to. If you can meet the business need using a single server, typically that is more cost-effective. Distributed systems bring their own headaches and pain, so generally, we only enter into that if we have to. So the typical reason to use a distributed system is because we have no choice: that's the only way to build the service (or the only way to do so within a reasonable cost).


I have no experience with your second question. I'm not sure whether that question is the best fit for the Stack Exchange format. Generally, on this site we are looking for posts that contain a single, focused technical question. We ask folks to avoid asking multiple questions in one post (they can be asked separately), to avoid questions that call for opinions (see https://cs.stackexchange.com/help/dont-ask), and on this site, we focus on the technical aspects of computer science rather than the business aspects of software engineering.

$\endgroup$
2
  • $\begingroup$ Hi thanks very much for the answer! which devision/sub i should go to ask the 2nd question if you know? $\endgroup$
    – Effy
    Commented Dec 21, 2023 at 6:47
  • $\begingroup$ @Effy, I'm not sure whether it is on-topic anywhere, as it seems to call for opinions, and requests for reading recommendations are also often opinion-based. If you can identify a more focused question, it might be on-topic at Software Engineering (but read their help center first), but I'm not sure whether "I don't know anything about this yet, I don't know where to start reading" will be a good fit there. Sorry to not be able to be of more assistance. $\endgroup$
    – D.W.
    Commented Dec 21, 2023 at 7:43

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.