Question

真的可以在这里使用一些帮助。在展示包含大量数据的仪表板时苦苦挣扎。

使用@ 2k记录平均@ 2秒。

MySql Console中的查询只需不到3.5秒即可返回150k行。 Ruby中的相同查询需要超过4 +分钟执行查询，直到所有对象都准备就绪。

目标：在添加缓存服务器之前进一步优化数据。使用Ruby 1.9.2，Rails 3.0和Mysql（Mysql2 gem）

问题：

使用Hashes会影响性能吗？
我应该先将所有内容放在一个主哈希中，然后再操作我需要的数据吗？
我还能做些什么来帮助提高表现吗？

DB中的行：

Gasstations和美国人口普查有<150,000条记录
人有@ 100,000条记录
汽车有@ 200,000条记录
FillUps有230万美元

仪表板必需（基于过去24小时，上周等的时间段查询）。所有数据都以JSON格式返回给JS。

加油站，有FillUps和美国人口普查数据（邮政编码，姓名，城市，人口）
填写最多的前20个城市
排名前10的汽车
根据他们装满水箱的次数分组的汽车

代码（6个月的样本。返回@ 100k +记录）：

# for simplicity, removed the select clause I had, but removing data I don't need like updated_at, gas_station.created_at, etc. instead of returning all the columns for each table.
@primary_data = FillUp.includes([:car, :gas_staton, :gas_station => {:uscensus}]).where('fill_ups.created_at >= ?', 6.months.ago) # This would take @ 4 + minutes

# then tried

@primary_data = FillUp.find_by_sql('some long sql query...') # took longer than before.
# Note for others, sql query did some pre processing for me which added attributes to the return.  Query in DB Console took < 4 seconds.  Because of these extra attributes, query took longer as if Ruby was checking each row for mapping attributes

# then tried

MY_MAP = Hash[ActiveRecord::Base.connection.select_all('SELECT thingone, thingtwo from table').map{|one| [one['thingone'], one['thingtwo']]}] as seen http://stackoverflow.com/questions/4456834/ruby-on-rails-storing-and-accessing-large-data-sets
# that took 23 seconds and gained mapping of additional data that was processing later, so much faster

# currently using below which takes @ 10 seconds
# All though this is faster, query still only takes 3.5 seconds, but parsing it to the hashes does add overhead.
cars = {}
gasstations = {}
cities = {}
filled = {}

client = Mysql2::Client.new(:host => "localhost", :username => "root")
client.query("SELECT sum(fill_ups_grouped_by_car_id) as filled, fillups.car_id, cars.make as make, gasstations.name as name,  ....", :stream => true, :as => :json).each do |row|
  # this returns fill ups gouged by car ,fill_ups.car_id, car make, gas station name, gas station zip, gas station city, city population 
  if cities[row['city']]
    cities[row['city']]['fill_ups']  = (cities[row['city']]['fill_ups']  + row['filled'])
  else
    cities[row['city']] = {'fill_ups' => row['filled'], 'population' => row['population']}
  end
  if gasstations[row['name']]
    gasstations[row['name']]['fill_ups'] = (gasstations[row['name']]['fill_ups'] + row['filled'])
  else
    gasstations[row['name']] = {'city' => row['city'],'zip' => row['city'], 'fill_ups' => row['filled']}
  end
  if cars[row['make']]
    cars[row['make']] = (cars[row['make']] + row['filled'])
  else
    cars[row['make']] = row['filled']
  end
  if row['filled']
    filled[row['filled']] = (filled[row['filled']] + 1)
  else
    filled[row['filled']] = 1
  end
end

拥有以下型号：

def Person
 has_many :cars 
end

def Car
  belongs_to :person
  belongs_to :uscensus, :foreign_key => :zipcode, :primary_key => :zipcode
  has_many :fill_ups
  has_many :gas_stations, :through => :fill_ups
end

def GasStation
  belongs_to :uscensus, :foreign_key => :zipcode, :primary_key => :zipcode
  has_many :fill_ups
  has_many :cars, :through => :fill_ups
end

def FillUp
  # log of every time a person fills up there gas
  belongs_to :car
  belongs_to :gas_station
end

def Uscensus
  # Basic data about area based on Zip code
end

Answer 1

我不使用RoR，但为仪表板返回100k行永远不会非常快。我强烈建议构建或维护摘要表并在数据库中运行GROUP BY以在演示之前汇总数据集。

使用大型数据集和ruby

1 个答案: