[erlang-questions] Why Distributed Is So Hard

Jayson Vantuyl <>
Mon Aug 30 12:53:30 CEST 2010


Well, I didn't mean to single out Mnesia specifically as being bad.  Rather, most people make assumptions when discussing things.  For example, when Mnesia has issues during a netsplit, it's often painful to even try to talk about the tradeoffs because people don't have a frame of reference.

In my post, I related that a coworker had asked me why distributed programming was any different than threaded programming.  His attitude was that it was just concurrency, right?  Well, we both know there's more, but I found that this question was a really good way to expose the differences between the two, and then blogged about it.

On Aug 30, 2010, at 3:27 AM, Evgeniy Khramtsov wrote:

> 30.08.2010 20:01, Jayson Vantuyl wrote
>> In particular, most distributed programming in Erlang is usually based on the assumption of a highly-available network layer (Mnesia, I'm looking at you).
> 
> Well, the main problem is network split. But I agree, distributed mnesia tables are not a good way to scale, especially if you have lots of nodes.
> 
>> This post talks about why that isn't good enough for distributed, in general, and gives a way to think about the problem specifically.
>>   
> 
> Everybody knows that this approach has several drawbacks, but, due to CAP-theorem, we don't have lots of choices - there is no silver bullet for this subject.
> 
> -- 
> Regards,
> Evgeniy Khramtsov, ProcessOne.
> xmpp:
> 
> 
> ________________________________________________________________
> erlang-questions (at) erlang.org mailing list.
> See http://www.erlang.org/faq.html
> To unsubscribe; mailto:
> 

-- 
Jayson Vantuyl
417-207-6962 (mobile)




More information about the erlang-questions mailing list