From: Marcin Raczkowski <mailing.mr@...>
Date: 2007-10-03T02:26:06+09:00
Subject: Re: Ruby vs. PHP

Richard Conroy wrote:
> On 10/1/07, Marcin Raczkowski <mailing.mr@gmail.com> wrote:
>>> But you have committed to an ActiveRecord rewrite, and your reasoning
>>> is not clear. It sounds like you have read something that 'ActiveRecord
>>> doesn't scale' without investigating thoroughly.
>> I didn't "hear" anything, i performed tests and benchmarks, with few
>> ORMs (AR, Og, sequel) and pure postgres library and obviously pure hand
>> made SQL is best - of course i can use AR's find_by_sql, and sometimes i
>> do but it's slower then pure call.
> 
> Yeah, but slower for which end, the Database or the Rails pizza box?
> Pursuing as-optimal-as-possible can introduce issues elsewhere. In general
> Rails ActiveRecord just needs to be 'fast enough' - i.e. not be the bottleneck.
> 

in my expirience when in my test (but based on real case) database  DB 
is taking 85% of request time i think it can be bottleneck don't you :]

> 
>>> Rails gets a lot of its productivity from ActiveRecord. By getting rid of it,
>>> you lose that, and you also close the door on the community development
>>> of ActiveRecord and its supporting plugins.
>>>
>>
>> I'm planning to use ORM, and use pure sql for most frequent calls (like
>> checking for new messages, or new stories) that are made by ajax
> 
> Thats a bad design. The most frequent calls should be hitting your memcache.
> If you have already rendered the new messages resource once, then it
> should be in your memcache. Subsequent calls to the resource should
> return the cached result, no Rails rendering or DB requests involved.
> 
> You cannot scale big without memcaching. It should be factored into your
> design already.

Why are you treating me like an idiot ? seriously? im using memcached, 
but in some cases i have to generate query based on parameters provided 
by user, and each user have diffirent excluding conditions based on his 
permision level, how you suggest i use memcached for that?

what if each request is unique and comes in random 2-20s intervals, how 
you suggest i use memcached?

I'm not saying memcached is bad, i even advertised it as big part of my 
project but what you said is like saying (using your pizza analogy) you 
need tomato sauce on each pizza, you can't make good pizza without 
tomato souce, use tomato souce becouse it's good, even if client have 
alergy give him tomato souce becouse you can't make pizza without it