How to improve distributed app in Scala?

[Sebastian M Cheung]

What are best practices in improving distributed Scala app performances?

[Rex Kerr]

This is a topic for a several-day conference, or a topic to cover with a handful of books or dozens of papers.

I'm not sure it's possible to say anything of use on such a broad topic on a mailing list, save: "When your program is distributed, communication between the parts takes time, so don't forget to pay attention to that when picking what to optimize."  (You always try to optimize the things that take the most time when you're improving performance, right?)

Well, that and, "If you're not prepared to really dig into this (as the topic is both deep and none-too-easy), use a framework like Spark that hides as many of the details from you as possible."

  --Rex

On Fri, Jul 24, 2015 at 7:35 AM, Sebastian M Cheung <minsc...@googlemail.com> wrote:

What are best practices in improving distributed Scala app performances?

--
You received this message because you are subscribed to the Google Groups "scala-language" group.
To unsubscribe from this group and stop receiving emails from it, send an email to scala-languag...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[Bardur Arantsson]

On 07/24/2015 04:35 PM, Sebastian M Cheung wrote:
> What are best practices in improving distributed Scala app performances?
>
>

To somewhat echo Rex's message: I think the only general advice that one
could give would be: Try to minimize your sensitivity to latency.

Btw, there is no such thing as "best practice". There may be "generally
recommended practice", but there's no such thing as "this is always the
right thing to do". It's always dependent on context. So, to encourage
healthy debate: What is *your* context? Describe your scenario in more
detail and perhaps people can contribute ideas and potential solutions...

Regards,

[Kostya Golikov]

Also, there are two kinds of things people usually think when they talk about performance -- latency and throughput and those two oftentimes contradict.