diff --git a/README.md b/README.md
index 24713d6..7826239 100644
--- a/README.md
+++ b/README.md
@@ -198,6 +198,10 @@
 - [22、说说生产环境下，你们是怎么实现网关对服务的动态路由的？](/docs/distributed-system/dynamic-route.md)[代码下载点击这里哦!](https://github.com/shishan100/Java-Interview-Advanced/raw/master/docs/distributed-system/code/code2.zip)
 - [23、如果网关需要抗每秒10万的高并发访问，你应该怎么对网关进行生产优化？](/docs/distributed-system/gateway-high-concurrency.md)
 - [24、作业：你们公司的网关是怎么技术选型的，假设有高并发场景怎么优化？](/docs/distributed-system/gateway-technical.md)
+- [25、如果需要部署上万服务实例，现有的服务注册中心能否抗住？如何优化？](/docs/distributed-system/registration-center-optimize.md)
+- [26、你们是如何基于网关实现灰度发布的？说说你们的灰度发布方案？](/docs/distributed-system/gray-environment.md)[代码下载点击这里哦!](https://github.com/shishan100/Java-Interview-Advanced/raw/master/docs/distributed-system/code/code3.zip)
+- [27、说说你们一个服务从开发到上线，服务注册、网关路由、服务调用的流程？](/docs/distributed-system/service-register-gateway-router.md)
+- [28、作业：看看你们公司的服务注册中心能否支撑上万服务实例的大规模场景？](/docs/distributed-system/work-register.md)
 
 
 ### 第二季-高并发
diff --git a/docs/distributed-system/code/code3.zip b/docs/distributed-system/code/code3.zip
new file mode 100755
index 0000000..9060d04
Binary files /dev/null and b/docs/distributed-system/code/code3.zip differ
diff --git a/docs/distributed-system/gray-environment.md b/docs/distributed-system/gray-environment.md
new file mode 100644
index 0000000..5f047ec
--- /dev/null
+++ b/docs/distributed-system/gray-environment.md
@@ -0,0 +1,76 @@
+
+#### 准备一个数据库和一个表（也可以用Apollo配置中心、Redis、ZooKeeper，其实都可以），放一个灰度发布启用表
+
+```
+id service_id path enable_gray_release
+
+CREATE TABLE `gray_release_config` (
+   `id` int(11) NOT NULL AUTO_INCREMENT,
+   `service_id` varchar(255) DEFAULT NULL,
+   `path` varchar(255) DEFAULT NULL,
+   `enable_gray_release` int(11) DEFAULT NULL,
+   PRIMARY KEY (`id`)
+ ) ENGINE=InnoDB DEFAULT CHARSET=utf8
+ 
+
+
+在zuul里面加入下面的filter，可以在zuul的filter里定制ribbon的负载均衡策略
+
+<dependency>
+			<groupId>io.jmnarloch</groupId>
+			<artifactId>ribbon-discovery-filter-spring-cloud-starter</artifactId>
+			<version>2.1.0</version>
+		</dependency>
+
+写一个zuul的filter，对每个请求，zuul都会调用这个filter
+
+@Configuration
+public class GrayReleaseFilter extends ZuulFilter {
+
+@Autowired
+private JdbcTemplate jdbcTemplate;
+
+    @Override
+    public int filterOrder() {
+        return PRE_DECORATION_FILTER_ORDER - 1;
+    }
+ 
+    @Override
+    public String filterType() {
+        return PRE_TYPE;
+    }
+ 
+    @Override
+    public boolean shouldFilter() {
+    	
+    }
+ 
+    @Override
+    public Object run() {
+        RequestContext ctx = RequestContext.getCurrentContext();
+        HttpServletRequest request = ctx.getRequest();
+
+Random random = new Random();
+int seed = random.getInt() * 100;
+
+        if (seed = 50) {
+            // put the serviceId in `RequestContext`
+            RibbonFilterContextHolder.getCurrentContext()
+                    .add("version", "new");
+        }  else {
+            RibbonFilterContextHolder.getCurrentContext()
+                    .add("version", "old");
+        }
+        
+        return null;
+    }
+}
+ ```
+
+eureka:
+instance:
+metadata-map:
+version: new
+
+
+
diff --git a/docs/distributed-system/images/registration-center-optimize.png b/docs/distributed-system/images/registration-center-optimize.png
new file mode 100755
index 0000000..3fec6a5
Binary files /dev/null and b/docs/distributed-system/images/registration-center-optimize.png differ
diff --git a/docs/distributed-system/registration-center-optimize.md b/docs/distributed-system/registration-center-optimize.md
new file mode 100644
index 0000000..ffe41c5
--- /dev/null
+++ b/docs/distributed-system/registration-center-optimize.md
@@ -0,0 +1,6 @@
+
+![分布式注册中心](/docs/distributed-system/images/registration-center-optimize.png)
+#### eureka：peer-to-peer，每台机器都是高并发请求，有瓶颈
+#### zookeeper：服务上下线，全量通知其他服务，网络带宽被打满，有瓶颈
+
+#### 分布式服务注册中心，分片存储服务注册表，横向扩容，每台机器均摊高并发请求，各个服务主动拉取，避免反向通知网卡被打满
diff --git a/docs/distributed-system/service-register-gateway-router.md b/docs/distributed-system/service-register-gateway-router.md
new file mode 100644
index 0000000..a65167b
--- /dev/null
+++ b/docs/distributed-system/service-register-gateway-router.md
@@ -0,0 +1,12 @@
+
+#### 生产环境，微服务生产实践
+
+开发了一个新的服务，线上部署，配合网关动态路由的功能，在网关里配置一下路径和新服务的映射关系，此时请求过来直接就可以走到新的服务里去
+
+对已有服务进行迭代和开发，新版本，灰度发布，新版本部署少数几台机器，通过一个界面，开启这个服务的灰度发布，**此时zuul filter启用，按照你的规则，把少量的流量打入到新版本部署的机器上去**
+
+观察一下少量流量在新版本的机器上运行是否正常
+
+版本改成current，全量机器部署，关闭灰度发布功能，网关就会把流量均匀分发给那个服务了
+
+
diff --git a/docs/distributed-system/work-register.md b/docs/distributed-system/work-register.md
new file mode 100644
index 0000000..5afd505
--- /dev/null
+++ b/docs/distributed-system/work-register.md
@@ -0,0 +1,13 @@
+
+#### Dubbo框架原理
+#### Spring Cloud框架原理
+#### 服务框架的技术选型
+
+#### 服务注册中心技术选型和核心原理
+#### 生产优化
+#### 架构优化
+
+#### 网关系统技术选型和核心原理
+#### 生产优化（灰度发布、动态路由）
+
+#### 生产级的分布式系统里，新服务开发如何做，老服务迭代如何做