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。
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.Runtime.Serialization;
6 using System.Web.Script.Serialization;
7 using System.IO;
8 using System.Xml.Serialization;
9 using System.Xml;
10 using Common;
11
12 namespace LoadDBData
13 {
14 class Program
15 {
16 static void Main(string[] args)
17 {
18 //模拟从数据库加载索引到内存中,形成内存中的数据库
19 //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
20 SerializableDictionary<int, List<int>> dic = new SerializableDictionary<int, List<int>>();
21
22 List<int> shopIDList = new List<int>();
23
24 for (int shopID = 300000; shopID < 300050; shopID++)
25 shopIDList.Add(shopID);
26
27 int UserID = 15;
28
29 //假设这里已经维护好了UserID与ShopID的关系
30 dic.Add(UserID, shopIDList);
31
32 XmlSerializer xml = new XmlSerializer(dic.GetType());
33
34 var memoryStream = new MemoryStream();
35
36 xml.Serialize(memoryStream, dic);
37
38 memoryStream.Seek(0, SeekOrigin.Begin);
39
40 //将Dictionary持久化,相当于模拟保存在Mencache里面
41 File.AppendAllText("F://1.txt", Encoding.UTF8.GetString(memoryStream.ToArray()));
42
43 Console.WriteLine("数据加载成功!");
44
45 Console.Read();
46 }
47 }
48 }
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
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.Xml.Serialization;
6 using System.Xml;
7 using System.Xml.Schema;
8 using System.Runtime.Serialization;
9
10 namespace Common
11 {
12 ///<summary>
13 /// 标题:支持 XML 序列化的 Dictionary
14 ///</summary>
15 ///<typeparam name="TKey"></typeparam>
16 ///<typeparam name="TValue"></typeparam>
17 [XmlRoot("SerializableDictionary")]
18 public class SerializableDictionary<TKey, TValue> : Dictionary<TKey, TValue>, IXmlSerializable
19 {
20
21 public SerializableDictionary()
22 : base()
23 {
24 }
25 public SerializableDictionary(IDictionary<TKey, TValue> dictionary)
26 : base(dictionary)
27 {
28 }
29
30 public SerializableDictionary(IEqualityComparer<TKey> comparer)
31 : base(comparer)
32 {
33 }
34
35 public SerializableDictionary(int capacity)
36 : base(capacity)
37 {
38 }
39 public SerializableDictionary(int capacity, IEqualityComparer<TKey> comparer)
40 : base(capacity, comparer)
41 {
42 }
43 protected SerializableDictionary(SerializationInfo info, StreamingContext context)
44 : base(info, context)
45 {
46 }
47
48
49 public System.Xml.Schema.XmlSchema GetSchema()
50 {
51 return null;
52 }
53 ///<summary>
54 /// 从对象的 XML 表示形式生成该对象
55 ///</summary>
56 ///<param name="reader"></param>
57 public void ReadXml(System.Xml.XmlReader reader)
58 {
59 XmlSerializer keySerializer = new XmlSerializer(typeof(TKey));
60 XmlSerializer valueSerializer = new XmlSerializer(typeof(TValue));
61 bool wasEmpty = reader.IsEmptyElement;
62 reader.Read();
63 if (wasEmpty)
64 return;
65 while (reader.NodeType != System.Xml.XmlNodeType.EndElement)
66 {
67 reader.ReadStartElement("item");
68 reader.ReadStartElement("key");
69 TKey key = (TKey)keySerializer.Deserialize(reader);
70 reader.ReadEndElement();
71 reader.ReadStartElement("value");
72 TValue value = (TValue)valueSerializer.Deserialize(reader);
73 reader.ReadEndElement();
74 this.Add(key, value);
75 reader.ReadEndElement();
76 reader.MoveToContent();
77 }
78 reader.ReadEndElement();
79 }
80
81 /**/
82 ///<summary>
83 /// 将对象转换为其 XML 表示形式
84 ///</summary>
85 ///<param name="writer"></param>
86 public void WriteXml(System.Xml.XmlWriter writer)
87 {
88 XmlSerializer keySerializer = new XmlSerializer(typeof(TKey));
89 XmlSerializer valueSerializer = new XmlSerializer(typeof(TValue));
90 foreach (TKey key in this.Keys)
91 {
92 writer.WriteStartElement("item");
93 writer.WriteStartElement("key");
94 keySerializer.Serialize(writer, key);
95 writer.WriteEndElement();
96 writer.WriteStartElement("value");
97 TValue value = this[key];
98 valueSerializer.Serialize(writer, value);
99 writer.WriteEndElement();
100 writer.WriteEndElement();
101 }
102 }
103
104 }
105 }
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。
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Runtime.Serialization;
5 using System.ServiceModel;
6 using System.Text;
7
8 namespace HeartBeatService
9 {
10 //CallbackContract:这个就是Client实现此接口,方便服务器端通知客户端
11 [ServiceContract(CallbackContract = typeof(ILiveAddressCallback))]
12 public interface IAddress
13 {
14 ///<summary>
15 /// 此方法用于Search启动后,将Search地址插入到此处
16 ///</summary>
17 ///<param name="address"></param>
18 [OperationContract(IsOneWay = true)]
19 void AddSearch(string address);
20
21 ///<summary>
22 /// 此方法用于IIS端获取search地址
23 ///</summary>
24 ///<param name="address"></param>
25 [OperationContract(IsOneWay = true)]
26 void GetService(string address);
27 }
28 }
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Runtime.Serialization;
5 using System.ServiceModel;
6 using System.Text;
7 using System.Timers;
8 using System.IO;
9 using System.Collections.Concurrent;
10 using SearhService;
11 using ClientService;
12
13 namespace HeartBeatService
14 {
15 //InstanceContextMode:主要是管理上下文的实例,此处是single,也就是单体
16 //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
17 [ServiceBehavior(InstanceContextMode = InstanceContextMode.Single, ConcurrencyMode = ConcurrencyMode.Multiple)]
18 public class Address : IAddress
19 {
20 static List<string> search = new List<string>();
21
22 static object obj = new object();
23
24 ///<summary>
25 /// 此静态构造函数用来检测存活的Search个数
26 ///</summary>
27 static Address()
28 {
29 Timer timer = new Timer();
30 timer.Interval = 6000;
31 timer.Elapsed += (sender, e) =>
32 {
33
34 Console.WriteLine("\n***************************************************************************");
35 Console.WriteLine("当前存活的Search为:");
36
37 lock (obj)
38 {
39 //遍历当前存活的Search
40 foreach (var single in search)
41 {
42 ChannelFactory<IProduct> factory = null;
43
44 try
45 {
46 //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。
47 factory = new ChannelFactory<IProduct>(new NetTcpBinding(SecurityMode.None), new EndpointAddress(single));
48 factory.CreateChannel().TestSearch();
49 factory.Close();
50
51 Console.WriteLine(single);
52
53 }
54 catch (Exception ex)
55 {
56 Console.WriteLine(ex.Message);
57
58 //如果抛出异常,则说明此search已经挂掉
59 search.Remove(single);
60 factory.Abort();
61 Console.WriteLine("\n当前时间:" + DateTime.Now + " ,存活的Search有:" + search.Count() + "个");
62 }
63 }
64 }
65
66 //最后统计下存活的search有多少个
67 Console.WriteLine("\n当前时间:" + DateTime.Now + " ,存活的Search有:" + search.Count() + "个");
68 };
69 timer.Start();
70 }
71
72 public void AddSearch(string address)
73 {
74
75 lock (obj)
76 {
77 //是否包含相同的Search地址
78 if (!search.Contains(address))
79 {
80 search.Add(address);
81
82 //search添加成功后就要告诉来源处,此search已经被成功载入。
83 var client = OperationContext.Current.GetCallbackChannel<ILiveAddressCallback>();
84 client.LiveAddress(address);
85 }
86 }
87 }
88
89 public void GetService(string address)
90 {
91 Timer timer = new Timer();
92 timer.Interval = 1000;
93 timer.Elapsed += (obj, sender) =>
94 {
95 try
96 {
97 //这个是定时的检测IIS是否挂掉
98 var factory = new ChannelFactory<IServiceList>(new NetTcpBinding(SecurityMode.None),
99 new EndpointAddress(address));
100
101 factory.CreateChannel().AddSearchList(search);
102
103 factory.Close();
104
105 timer.Interval = 10000;
106 }
107 catch (Exception ex)
108 {
109 Console.WriteLine(ex.Message);
110 }
111 };
112 timer.Start();
113 }
114 }
115 }
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.ServiceModel;
6
7 namespace HeartBeatService
8 {
9 ///<summary>
10 /// 等客户端实现后,让客户端约束一下,只能是这个LiveAddress方法
11 ///</summary>
12 public interface ILiveAddressCallback
13 {
14 [OperationContract(IsOneWay = true)]
15 void LiveAddress(string address);
16 }
17 }
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。
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Runtime.Serialization;
5 using System.ServiceModel;
6 using System.Text;
7
8 namespace SearhService
9 {
10 // 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”。
11 [ServiceContract]
12 public interface IProduct
13 {
14 [OperationContract]
15 List<int> GetShopListByUserID(int userID);
16
17 [OperationContract]
18 void TestSearch();
19 }
20 }
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Runtime.Serialization;
5 using System.ServiceModel;
6 using System.Text;
7 using Common;
8 using System.Xml;
9 using System.IO;
10 using System.Xml.Serialization;
11
12 namespace SearhService
13 {
14 public class Product : IProduct
15 {
16 public List<int> GetShopListByUserID(int userID)
17 {
18 //模拟从MemCache中读取索引
19 SerializableDictionary<int, List<int>> dic = new SerializableDictionary<int, List<int>>();
20
21 byte[] bytes = Encoding.UTF8.GetBytes(File.ReadAllText("F://1.txt", Encoding.UTF8));
22
23 var memoryStream = new MemoryStream();
24
25 memoryStream.Write(bytes, 0, bytes.Count());
26
27 memoryStream.Seek(0, SeekOrigin.Begin);
28
29 XmlSerializer xml = new XmlSerializer(dic.GetType());
30
31 var obj = xml.Deserialize(memoryStream) as Dictionary<int, List<int>>;
32
33 return obj[userID];
34 }
35
36 public void TestSearch() { }
37 }
38 }
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.ServiceModel;
6 using System.Configuration;
7 using System.Timers;
8 using SearhService.HeartBeatService;
9
10 namespace SearhService
11 {
12 public class SearchHost : IAddressCallback
13 {
14 static DateTime startTime;
15
16 public static void Main()
17 {
18 ServiceHost host = new ServiceHost(typeof(Product));
19
20 host.Open();
21
22 AddSearch();
23
24 Console.Read();
25
26 }
27
28 static void AddSearch()
29 {
30 startTime = DateTime.Now;
31
32 Console.WriteLine("Search服务发送中.....\n\n*************************************************\n");
33
34 try
35 {
36 var heartClient = new AddressClient(new InstanceContext(new SearchHost()));
37
38 string search = ConfigurationManager.AppSettings["search"];
39
40 heartClient.AddSearch(search);
41 }
42 catch (Exception ex)
43 {
44 Console.WriteLine("Search服务发送失败:" + ex.Message);
45 }
46 }
47
48 public void LiveAddress(string address)
49 {
50 Console.WriteLine("恭喜你," + address + "已被心跳成功接收!\n");
51 Console.WriteLine("发送时间:" + startTime + "\n接收时间:" + DateTime.Now);
52 }
53 }
54 }
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。
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.ServiceModel;
6
7 namespace ClientService
8 {
9 [ServiceContract]
10 public interface IServiceList
11 {
12 [OperationContract]
13 void AddSearchList(List<string> search);
14 }
15 }
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.ServiceModel;
6 using System.Configuration;
7 using System.Timers;
8 using System.Threading;
9
10 namespace ClientService
11 {
12 public class ServiceList : IServiceList
13 {
14 public static List<string> searchList = new List<string>();
15
16 static object obj = new object();
17
18 public static string Search
19 {
20 get
21 {
22 lock (obj)
23 {
24 //如果心跳没及时返回地址,客户端就在等候
25 if (searchList.Count == 0)
26 Thread.Sleep(1000);
27 return searchList[new Random().Next(0, searchList.Count)];
28 }
29 }
30 set
31 {
32
33 }
34 }
35
36 public void AddSearchList(List<string> search)
37 {
38 lock (obj)
39 {
40 searchList = search;
41
42 Console.WriteLine("************************************");
43 Console.WriteLine("当前存活的Search为:");
44
45 foreach (var single in searchList)
46 {
47 Console.WriteLine(single);
48 }
49 }
50 }
51 }
52 }
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.ServiceModel;
6 using System.Configuration;
7 using System.Threading;
8 using ClientService.HeartBeatService;
9 using SearhService;
10 using BaseClass;
11 using System.Data;
12 using System.Diagnostics;
13
14 namespace ClientService
15 {
16 class Program : IAddressCallback
17 {
18 static void Main(string[] args)
19 {
20
21 ServiceHost host = new ServiceHost(typeof(ServiceList));
22
23 host.Open();
24
25 var client = new AddressClient(new InstanceContext(new Program()));
26
27 //配置文件中获取iis的地址
28 var iis = ConfigurationManager.AppSettings["iis"];
29
30 //将iis的地址告诉心跳
31 client.GetService(iis);
32
33 //从集群中获取search地址来对Search服务进行调用
34 var factory = new ChannelFactory<IProduct>(new NetTcpBinding(SecurityMode.None), new EndpointAddress(ServiceList.Search));
35
36 //根据userid获取了shopID的集合
37 var shopIDList = factory.CreateChannel().GetShopListByUserID(15);
38
39 //.......................... 后续就是我们将shopIDList做连接数据库查询(做到秒杀)
40
41 Console.Read();
42 }
43
44 public void LiveAddress(string address)
45 {
46
47 }
48 }
49 }
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:
1 using System;
2 using System.Collections.Generic;
3 using System.Linq;
4 using System.Text;
5 using System.ServiceModel;
6 using System.Timers;
7 using System.Diagnostics;
8 using BaseClass;
9 using ClientService;
10 using ClientService.HeartBeatService;
11 using System.Configuration;
12 using SearhService;
13
14 namespace ClientService
15 {
16 class Program : IAddressCallback
17 {
18 static void Main(string[] args)
19 {
20
21 ServiceHost host = new ServiceHost(typeof(ServiceList));
22
23 host.Open();
24
25 var client = new AddressClient(new InstanceContext(new Program()));
26
27 //配置文件中获取iis的地址
28 var iis = ConfigurationManager.AppSettings["iis"];
29
30 //将iis的地址告诉心跳
31 client.GetService(iis);
32
33 //从集群中获取search地址来对Search服务进行调用
34 var factory = new ChannelFactory<IProduct>(new NetTcpBinding(SecurityMode.None), new EndpointAddress(ServiceList.Search));
35
36 //根据userid获取了shopID的集合
37 //比如说这里的ShopIDList是通过索引交并集获取的分页的一些shopID
38 var shopIDList = factory.CreateChannel().GetShopListByUserID(15);
39
40 var strSql = string.Join(",", shopIDList);
41
42 Stopwatch watch = new Stopwatch();
43
44 watch.Start();
45 SqlHelper.Query("select s.ShopID,u.UserName,s.ShopName from [User] as u ,Shop as s where s.ShopID in(" + strSql + ")");
46 watch.Stop();
47
48 Console.WriteLine("通过wcf索引获取的ID >>>花费时间:" + watch.ElapsedMilliseconds);
49
50 //普通的sql查询花费的时间
51 StringBuilder builder = new StringBuilder();
52
53 builder.Append("select * from ");
54 builder.Append("(select ROW_NUMBER() over(order by s.ShopID) as NumberID, ");
55 builder.Append(" s.ShopID, u.UserName, s.ShopName ");
56 builder.Append("from Shop s left join [User] as u on u.UserID=s.UserID ");
57 builder.Append("where s.UserID=15) as array ");
58 builder.Append("where NumberID>300000 and NumberID<300050");
59
60 watch.Start();
61 SqlHelper.Query(builder.ToString());
62 watch.Stop();
63
64 Console.WriteLine("普通的sql分页 >>>花费时间:" + watch.ElapsedMilliseconds);
65
66 Console.Read();
67 }
68
69 public void LiveAddress(string address)
70 {
71
72 }
73 }
74 }
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。