orm_benchmark.rb

require 'benchmark'
require 'pp'
require 'active_record'
require 'sequel'
require 'data_mapper'

@start = Date.parse("2012-12-01")
@end   = Date.parse("2012-12-31")

ActiveRecord::Base.establish_connection(adapter: 'mysql2', database: 'escobar_development', cast: false, username: 'root', password: nil, pool: 5, timeout: 5000)
class MCwAR < ActiveRecord::Base
  self.table_name = 'marketing_costs'
end

Sequel.connect(adapter: 'mysql2', user: 'root', password: nil, host: 'localhost', database: 'escobar_development')
class MCwSq < Sequel::Model(:marketing_costs)
end

class MCwDM
  include DataMapper::Resource
  storage_names[:default] = 'marketing_costs'

  property :id,         Serial # An auto-increment integer key
  property :ad_dimension, String
  property :ad_group, String
  property :ad_position, String
  property :ad_type, String
  property :ad_url, String
  property :campaign, String
  property :content, String
  property :cost, Decimal
  property :cost_events, Integer
  property :cost_model, String
  property :created_at, DateTime
  property :creative_filename, String
  property :date, Date
  property :datetime, DateTime
  property :import_id, Integer
  property :impressions, Integer
  property :match_type, String
  property :medium, String
  property :placement, String
  property :source_id, Integer
  property :source_name, String
  property :source_type, String
  property :sub_id, String
  property :term, String
  property :time, Time
  property :timestamp, DateTime
  property :tv_market, String
  property :updated_at, DateTime
end

# Connect to the database
DataMapper.setup(:default, 'mysql://root@localhost/escobar_development')

# Finalize all the database models
# This will drastically improve the performance of your app
DataMapper.finalize

# By default throw an exception if a model fails validation
# Never leave home without this or your towel.  -HGTTG
DataMapper::Model.raise_on_save_failure = true



###
### Pre-Populate the table with some data
###
# DataMapper.auto_migrate! # ==> Drops & Creates the table
# DataMapper.auto_upgrade! # ==> Perform any migrations on the table
# 1.upto(31) do |day|
#   timestamp = Date.parse("2012-12-%02d" % day)
#   'A'.upto('Z') do |campaign|
#     0.upto(5) do |source_id|
#       0.upto(5) do |cost_events|
#         0.upto(5) do |cost|
#           MCwDM.create(:campaign => campaign, :source_id => source_id, :cost_events => cost_events, :cost => cost, :timestamp => timestamp).save
#         end
#       end
#     end
#   end
# end



GC.start; GC.disable

Benchmark.bm do |x|
  x.report("[%15s]" % "activerecord") do 
    100.times do
      objs = MCwAR.select(['source_id', 'campaign', 'sum(cost)', 'sum(cost_events)']).where(date: (@start..@end)).group(['source_id', 'campaign'])
      objs.each do |obj|
        row_id = obj.source_id.to_s + obj.campaign.to_s
      end
    end 
  end
  
  GC.enable; GC.start; GC.disable
  
  x.report("[%15s]" % "sequel") do
    100.times do
      objs = MCwSq.select_group(:source_id, :campaign).select_append{sum(cost)}.select_append{sum(cost_events)}.where(date: (@start..@end))
      objs.each do |obj|
        row_id = obj.source_id.to_s + obj.campaign.to_s
      end
    end
  end

  GC.enable; GC.start; GC.disable

  x.report("[%15s]" % "datamapper") do
    100.times do
      # Find all the rows
      objs = MCwDM.aggregate(:source_id, :campaign, :cost.sum, :cost_events.count, :timestamp.gte => @start, :timestamp.lte => @end)
      # Map the rows to hashes
      objs.collect! { |o| { :source_id => o.shift, :campaign => o.shift, :cost_sum => o.shift, :cost_events_count => o.shift } }
      # Do the down and dirty
      objs.each do |obj|
        row_id = obj[:source_id].to_s + obj[:campaign].to_s
      end
    end
  end
end

                   user       system     total    real
[   activerecord]  0.050000   0.010000   0.060000 (  0.063170)
[         sequel]  0.050000   0.000000   0.050000 (  0.053465)
[     datamapper]  0.280000   0.020000   0.300000 (  0.312492)

A neat trick here is use DataMapper in your backend/scraper scripts to populate the tables,
and use Sequel or ActiveRecord in your frontend to render the data as fast as possible.

This makes the development easy. reduces migrations, and simplifies your code; at the cost of not using AR-migrations.