今天是wcf系列的第四天，也该出手压轴戏了。嗯，现在的大型架构，都是神马的，
nginx鸡群，iis鸡群，wcf鸡群，DB鸡群，由一个人作战变成了群殴.......
今天我就分享下wcf鸡群，高性能架构中一种常用的手法就是在内存中维护一个叫做“索引”的内存数据库，
在实战中利用“索引”这个概念做出"海量数据“的秒杀。
好，先上图：
这个图明白人都能看得懂吧。因为我的系列偏重于wcf，所以我重点说下”心跳检测“的实战手法。
第一步：上一下项目的结构，才能做到心中有数。
第二步：“LoadDBService”这个是控制台程序，目的就是从数据库抽出关系模型加载在内存数据库中，因为这些东西会涉及一些算法的知识，
在这里就不写算法了，就简单的模拟一下。
LoadDBServcie
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Runtime.Serialization;
using System.Web.Script.Serialization;
using System.IO;
using System.Xml.Serialization;
using System.Xml;
using Common;
namespace LoadDBData
{
class Program
{
static void Main(string[] args)
{
//模拟从数据库加载索引到内存中,形成内存中的数据库
//这里的 "Dictionary" 用来表达“一个用户注册过多少店铺“，即UserID与ShopID的一对多关系
SerializableDictionary<int, List<int>> dic = new SerializableDictionary<int, List<int>>();
List<int> shopIDList = new List<int>();
for (int shopID = 300000; shopID < 300050; shopID++)
shopIDList.Add(shopID);
int UserID = 15;
//假设这里已经维护好了UserID与ShopID的关系
dic.Add(UserID, shopIDList);
XmlSerializer xml = new XmlSerializer(dic.GetType());
var memoryStream = new MemoryStream();
xml.Serialize(memoryStream, dic);
memoryStream.Seek(0, SeekOrigin.Begin);
//将Dictionary持久化，相当于模拟保存在Mencache里面
File.AppendAllText("F://1.txt", Encoding.UTF8.GetString(memoryStream.ToArray()));
Console.WriteLine("数据加载成功！");
Console.Read();
}
}
}因为Dictionary不支持序列化，所以我从网上拷贝了一份代码让其执行序列化
SerializableDictionary
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.Xml.Serialization;
using System.Xml;
using System.Xml.Schema;
using System.Runtime.Serialization;
namespace Common
{
///<summary>
/// 标题：支持 XML 序列化的 Dictionary
///</summary>
///<typeparam name="TKey"></typeparam>
///<typeparam name="TValue"></typeparam>
[XmlRoot("SerializableDictionary")]
public class SerializableDictionary<TKey, TValue> : Dictionary<TKey, TValue>, IXmlSerializable
{
public SerializableDictionary()
: base()
{
}
public SerializableDictionary(IDictionary<TKey, TValue> dictionary)
: base(dictionary)
{
}
public SerializableDictionary(IEqualityComparer<TKey> comparer)
: base(comparer)
{
}
public SerializableDictionary(int capacity)
: base(capacity)
{
}
public SerializableDictionary(int capacity, IEqualityComparer<TKey> comparer)
: base(capacity, comparer)
{
}
protected SerializableDictionary(SerializationInfo info, StreamingContext context)
: base(info, context)
{
}
public System.Xml.Schema.XmlSchema GetSchema()
{
return null;
}
///<summary>
/// 从对象的 XML 表示形式生成该对象
///</summary>
///<param name="reader"></param>
public void ReadXml(System.Xml.XmlReader reader)
{
XmlSerializer keySerializer = new XmlSerializer(typeof(TKey));
XmlSerializer valueSerializer = new XmlSerializer(typeof(TValue));
bool wasEmpty = reader.IsEmptyElement;
reader.Read();
if (wasEmpty)
return;
while (reader.NodeType != System.Xml.XmlNodeType.EndElement)
{
reader.ReadStartElement("item");
reader.ReadStartElement("key");
TKey key = (TKey)keySerializer.Deserialize(reader);
reader.ReadEndElement();
reader.ReadStartElement("value");
TValue value = (TValue)valueSerializer.Deserialize(reader);
reader.ReadEndElement();
this.Add(key, value);
reader.ReadEndElement();
reader.MoveToContent();
}
reader.ReadEndElement();
}
/**/
///<summary>
/// 将对象转换为其 XML 表示形式
///</summary>
///<param name="writer"></param>
public void WriteXml(System.Xml.XmlWriter writer)
{
XmlSerializer keySerializer = new XmlSerializer(typeof(TKey));
XmlSerializer valueSerializer = new XmlSerializer(typeof(TValue));
foreach (TKey key in this.Keys)
{
writer.WriteStartElement("item");
writer.WriteStartElement("key");
keySerializer.Serialize(writer, key);
writer.WriteEndElement();
writer.WriteStartElement("value");
TValue value = this[key];
valueSerializer.Serialize(writer, value);
writer.WriteEndElement();
writer.WriteEndElement();
}
}
}
}
第三步： "HeartBeatService"也做成了一个控制台程序，为了图方便，把Contract和Host都放在一个控制台程序中,
代码中加入了注释，看一下就会懂的。
IAddress.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Runtime.Serialization;
using System.ServiceModel;
using System.Text;
namespace HeartBeatService
{
//CallbackContract：这个就是Client实现此接口，方便服务器端通知客户端
[ServiceContract(CallbackContract = typeof(ILiveAddressCallback))]
public interface IAddress
{
///<summary>
/// 此方法用于Search启动后，将Search地址插入到此处
///</summary>
///<param name="address"></param>
[OperationContract(IsOneWay = true)]
void AddSearch(string address);
///<summary>
/// 此方法用于IIS端获取search地址
///</summary>
///<param name="address"></param>
[OperationContract(IsOneWay = true)]
void GetService(string address);
}
}Address.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Runtime.Serialization;
using System.ServiceModel;
using System.Text;
using System.Timers;
using System.IO;
using System.Collections.Concurrent;
using SearhService;
using ClientService;
namespace HeartBeatService
{
//InstanceContextMode：主要是管理上下文的实例，此处是single，也就是单体
//ConcurrencyMode： 主要是用来控制实例中的线程数，此处是Multiple，也就是多线程
[ServiceBehavior(InstanceContextMode = InstanceContextMode.Single, ConcurrencyMode = ConcurrencyMode.Multiple)]
public class Address : IAddress
{
static List<string> search = new List<string>();
static object obj = new object();
///<summary>
/// 此静态构造函数用来检测存活的Search个数
///</summary>
static Address()
{
Timer timer = new Timer();
timer.Interval = 6000;
timer.Elapsed += (sender, e) =>
{
Console.WriteLine("\n***************************************************************************");
Console.WriteLine("当前存活的Search为：");
lock (obj)
{
//遍历当前存活的Search
foreach (var single in search)
{
ChannelFactory<IProduct> factory = null;
try
{
//当Search存在的话，心跳服务就要定时检测Search是否死掉，也就是定时的连接Search来检测。
factory = new ChannelFactory<IProduct>(new NetTcpBinding(SecurityMode.None), new EndpointAddress(single));
factory.CreateChannel().TestSearch();
factory.Close();
Console.WriteLine(single);
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
//如果抛出异常，则说明此search已经挂掉
search.Remove(single);
factory.Abort();
Console.WriteLine("\n当前时间：" + DateTime.Now + " ,存活的Search有：" + search.Count() + "个");
}
}
}
//最后统计下存活的search有多少个
Console.WriteLine("\n当前时间：" + DateTime.Now + " ,存活的Search有：" + search.Count() + "个");
};
timer.Start();
}
public void AddSearch(string address)
{
lock (obj)
{
//是否包含相同的Search地址
if (!search.Contains(address))
{
search.Add(address);
//search添加成功后就要告诉来源处，此search已经被成功载入。
var client = OperationContext.Current.GetCallbackChannel<ILiveAddressCallback>();
client.LiveAddress(address);
}
}
}
public void GetService(string address)
{
Timer timer = new Timer();
timer.Interval = 1000;
timer.Elapsed += (obj, sender) =>
{
try
{
//这个是定时的检测IIS是否挂掉
var factory = new ChannelFactory<IServiceList>(new NetTcpBinding(SecurityMode.None),
new EndpointAddress(address));
factory.CreateChannel().AddSearchList(search);
factory.Close();
timer.Interval = 10000;
}
catch (Exception ex)
{
Console.WriteLine(ex.Message);
}
};
timer.Start();
}
}
}ILiveAddressCallback.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.ServiceModel;
namespace HeartBeatService
{
///<summary>
/// 等客户端实现后，让客户端约束一下，只能是这个LiveAddress方法
///</summary>
public interface ILiveAddressCallback
{
[OperationContract(IsOneWay = true)]
void LiveAddress(string address);
}
}第四步： 我们开一下心跳，预览下效果：
是的，心跳现在正在检测是否有活着的Search。
第五步："SearhService" 这个Console程序就是WCF的search，主要用于从MemerCache里面读取索引。
记得要添加一下对“心跳服务”的服务引用。
IProduct.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Runtime.Serialization;
using System.ServiceModel;
using System.Text;
namespace SearhService
{
// 注意: 使用“重构”菜单上的“重命名”命令，可以同时更改代码和配置文件中的接口名“IService1”。
[ServiceContract]
public interface IProduct
{
[OperationContract]
List<int> GetShopListByUserID(int userID);
[OperationContract]
void TestSearch();
}
}Product.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Runtime.Serialization;
using System.ServiceModel;
using System.Text;
using Common;
using System.Xml;
using System.IO;
using System.Xml.Serialization;
namespace SearhService
{
public class Product : IProduct
{
public List<int> GetShopListByUserID(int userID)
{
//模拟从MemCache中读取索引
SerializableDictionary<int, List<int>> dic = new SerializableDictionary<int, List<int>>();
byte[] bytes = Encoding.UTF8.GetBytes(File.ReadAllText("F://1.txt", Encoding.UTF8));
var memoryStream = new MemoryStream();
memoryStream.Write(bytes, 0, bytes.Count());
memoryStream.Seek(0, SeekOrigin.Begin);
XmlSerializer xml = new XmlSerializer(dic.GetType());
var obj = xml.Deserialize(memoryStream) as Dictionary<int, List<int>>;
return obj[userID];
}
public void TestSearch() { }
}
}SearchHost.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.ServiceModel;
using System.Configuration;
using System.Timers;
using SearhService.HeartBeatService;
namespace SearhService
{
public class SearchHost : IAddressCallback
{
static DateTime startTime;
public static void Main()
{
ServiceHost host = new ServiceHost(typeof(Product));
host.Open();
AddSearch();
Console.Read();
}
static void AddSearch()
{
startTime = DateTime.Now;
Console.WriteLine("Search服务发送中.....\n\n*************************************************\n");
try
{
var heartClient = new AddressClient(new InstanceContext(new SearchHost()));
string search = ConfigurationManager.AppSettings["search"];
heartClient.AddSearch(search);
}
catch (Exception ex)
{
Console.WriteLine("Search服务发送失败：" + ex.Message);
}
}
public void LiveAddress(string address)
{
Console.WriteLine("恭喜你," + address + "已被心跳成功接收！\n");
Console.WriteLine("发送时间：" + startTime + "\n接收时间：" + DateTime.Now);
}
}
}第六步：此时Search服务已经建好，我们可以测试当Search开启获取关闭对心跳有什么影响：
Search开启时：
Search关闭时：
对的，当Search关闭时，心跳检测该Search已经死掉，然后只能从集群中剔除。
当然，我们可以将Search拷贝N份，部署在N台机器中，只要修改一下endpoint地址就OK了，这一点明白人都会。
第七步："ClientService" 这里也就指的是IIS，此时我们也要添加一下对心跳的服务引用。
IServiceList.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.ServiceModel;
namespace ClientService
{
[ServiceContract]
public interface IServiceList
{
[OperationContract]
void AddSearchList(List<string> search);
}
}ServiceList.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.ServiceModel;
using System.Configuration;
using System.Timers;
using System.Threading;
namespace ClientService
{
public class ServiceList : IServiceList
{
public static List<string> searchList = new List<string>();
static object obj = new object();
public static string Search
{
get
{
lock (obj)
{
//如果心跳没及时返回地址，客户端就在等候
if (searchList.Count == 0)
Thread.Sleep(1000);
return searchList[new Random().Next(0, searchList.Count)];
}
}
set
{
}
}
public void AddSearchList(List<string> search)
{
lock (obj)
{
searchList = search;
Console.WriteLine("************************************");
Console.WriteLine("当前存活的Search为:");
foreach (var single in searchList)
{
Console.WriteLine(single);
}
}
}
}
}Program.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.ServiceModel;
using System.Configuration;
using System.Threading;
using ClientService.HeartBeatService;
using SearhService;
using BaseClass;
using System.Data;
using System.Diagnostics;
namespace ClientService
{
class Program : IAddressCallback
{
static void Main(string[] args)
{
ServiceHost host = new ServiceHost(typeof(ServiceList));
host.Open();
var client = new AddressClient(new InstanceContext(new Program()));
//配置文件中获取iis的地址
var iis = ConfigurationManager.AppSettings["iis"];
//将iis的地址告诉心跳
client.GetService(iis);
//从集群中获取search地址来对Search服务进行调用
var factory = new ChannelFactory<IProduct>(new NetTcpBinding(SecurityMode.None), new EndpointAddress(ServiceList.Search));
//根据userid获取了shopID的集合
var shopIDList = factory.CreateChannel().GetShopListByUserID(15);
//.......................... 后续就是我们将shopIDList做连接数据库查询（做到秒杀）
Console.Read();
}
public void LiveAddress(string address)
{
}
}
}
然后我们开启Client,看看效果咋样：
当然，search集群后，client得到search的地址是随机的，也就分担了search的负担，实现有福同享，有难同当的效果了。
最后： 我们做下性能检测，看下“秒杀”和“毫秒杀”的效果。
首先在数据库的User表和Shop插入了180万和20万的数据用于关联。
ClientService改造后的代码：
Program.cs
using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.ServiceModel;
using System.Timers;
using System.Diagnostics;
using BaseClass;
using ClientService;
using ClientService.HeartBeatService;
using System.Configuration;
using SearhService;
namespace ClientService
{
class Program : IAddressCallback
{
static void Main(string[] args)
{
ServiceHost host = new ServiceHost(typeof(ServiceList));
host.Open();
var client = new AddressClient(new InstanceContext(new Program()));
//配置文件中获取iis的地址
var iis = ConfigurationManager.AppSettings["iis"];
//将iis的地址告诉心跳
client.GetService(iis);
//从集群中获取search地址来对Search服务进行调用
var factory = new ChannelFactory<IProduct>(new NetTcpBinding(SecurityMode.None), new EndpointAddress(ServiceList.Search));
//根据userid获取了shopID的集合
//比如说这里的ShopIDList是通过索引交并集获取的分页的一些shopID
var shopIDList = factory.CreateChannel().GetShopListByUserID(15);
var strSql = string.Join(",", shopIDList);
Stopwatch watch = new Stopwatch();
watch.Start();
SqlHelper.Query("select s.ShopID,u.UserName,s.ShopName from [User] as u ,Shop as s where s.ShopID in(" + strSql + ")");
watch.Stop();
Console.WriteLine("通过wcf索引获取的ID >>>花费时间:" + watch.ElapsedMilliseconds);
//普通的sql查询花费的时间
StringBuilder builder = new StringBuilder();
builder.Append("select * from ");
builder.Append("(select ROW_NUMBER() over(order by s.ShopID) as NumberID, ");
builder.Append(" s.ShopID, u.UserName, s.ShopName ");
builder.Append("from Shop s left join [User] as u on u.UserID=s.UserID ");
builder.Append("where s.UserID=15) as array ");
builder.Append("where NumberID>300000 and NumberID<300050");
watch.Start();
SqlHelper.Query(builder.ToString());
watch.Stop();
Console.WriteLine("普通的sql分页 >>>花费时间:" + watch.ElapsedMilliseconds);
Console.Read();
}
public void LiveAddress(string address)
{
}
}
}性能图：
对的，一个秒杀，一个是毫秒杀，所以越复杂越能展示出“内存索引”的强大之处。
![[PolarDB实操课] 01.PolarDB分布式版架构介绍](https://ucc.alicdn.com/pic/developer-ecology/zdtg5ua724qza_70a3f73ea9cf4243ab9e48d813773fb7.png?x-oss-process=image/resize,h_160,m_lfit)