Replication简介:
"Replication is the process of sharing information so as to ensure consistency between redundant resources, such as software or hardware components, to improve reliability, fault-tolerance, or accessibility." (wikipedia)
replication技术的目的主要是用于提高分布式系统(distributed object, database, file system等)的可依赖性(dependability), 依赖于冗余资源, 需要维护这些资源间的一致性.
按照复制的对象, 可分为data replication与computation replication两种.
按照实现方式, 可分为active replication与passive replication两种.
详见 http://en.wikipedia.org/wiki/Replication_(computer_science)
论文(一)
OSRAEL J, FROIHOFER L, GOESCHKA K M. What service replication middleware can learn from object replication middleware [C]// Proceedings of the 1st Workshop on Middleware for Service Oriented Computing. Melbourne, Australia: Association for Computing Machinery, 2006: 18-23
(1) 本文认为SO系统中replication尚处于起步阶段. 从架构(architectural)层次, 将现有的一些service replication middleware, 与object replication middleware进行了比较.
得到的结果是: SO系统中与分布式对象系统中的replication middleware有很多共同点, 存在的些许区别是由replicated entity的粒度和不同的transaction模型引起的.
(2) 现有的SO replication middleware
a)primary-backup replication middleware: "Fault tolerant web service"
b)active replication middleware: "WS-Replication: a framework for highly available web services", "A middleware for replicated Web services"
(3) 对于stateless service的replication比较简单, 难点主要在于针对stateful service的replication.
(4) 提取了6种architectural units (适用于object and service replication middleware):
Multicast service
monitoring service
replication manager
replication protocol unit
invocation service
transaction service(optional)
(5) 与WSC无关; 只关注dependability, 与QoS无关
论文(二)
LIANG D, FANG C L, CHEN C et.al. Fault tolerant Web service [C]// Proceedings of the Tenth Asia-Pacific Software Engineering Conference Software Engineering Conference (APSEC). Washington, DC: IEEE Computer Society, 2003: 310-319
本文gs: 32
(1) 提出了一种fault tolerant web service (FT-SOAP), 在消息层(SOAP)截获用户请求; 如果请求的WS fail了, 就自动重定向到backup WS. 本文的版面有点乱.
(2) 不涉及stateful WS; 不涉及WSC, QoS; 针对的是单个服务的fault tolerant; 适用replication manager生成备份服务, 不支持服务的动态发现等
论文(三)
YE X, SHEN Y. A Middleware for Replicated Web Services [C]// IEEE International Conference on Web Services (ICWS). Los Alamitos: IEEE Computer Society, 2005: 631-638
(1)提出了一个基于active replication, 支持reliable web service的middleware. 这个middleware负责维护replicas的状态.
使用这个middleware, 现有的单线程WS无需更改实现即可被复制到不同的站点.
每个replica包括两个实体: proxy service site(PWSS)和普通的web service site(WSS). 用户通过PWSS发送请求, PWSS负责维护对应的WSS的一致性和出错处理.
(2) 不涉及WSC QoS
论文(四)
SALAS J, PEREZ-SORROSAL F, PATINO-MARTINEZ M, et.al. WS-replication: a framework for highly available web services [C]// Proceedings of the 15th international conference on World Wide Web (WWW). Edinburgh, Scotland: Association for Computing Machinery, 2006: 357 - 366
(1) 本文提出了WS-Replication架构(infrastructure), 用于WAN环境中WS的replication.
WS-Replication的目的是使WS具有高度的availability, 从而能够满足mission critical system的需求.
这个架构基于group communication WS, WS-Multicast. WS-Multicast的传输是基于SOAP, 依赖WS进行跨组织的交互.
WS-Replication中采用了active replication的方式.
这四篇论文看的都比较粗略, 只是了解了一下论文的大概工作, 没有通读全文. 读了之后, 对WS replication有以下印象:
1. 作为replica的web services一般都是同一份实现, 需要维护保证代码和状态的一致性
2. WS不是动态发现的
3. 不是在WSC的场景下讨论, 主要针对WS的可靠性, 不涉及其他QoS
4. 都是通过middleware来支持web service replication
5. 需要考虑replica的location