Network Working Group M. Tuexen INTERNET DRAFT Siemens AG expires April 20, 2001 October 20, 2000 Requirements for Reliable Server Pooling Status of this Memo This document is an Internet-Draft and is in full conformance with all provisions of Section 10 of [RFC2026]. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet-Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. Abstract The goal is to develop an architecture and protocols for the management and operation of server pools supporting highly reliable applications, and for client access mechanisms to a server pool. This document defines requirements and architecture for management and access to server pools, including requirements from a variety of applications, building blocks and interfaces, different styles of pooling, security requirements and performance requirements such as failover times and coping with heterogeneous latencies. Important requirements of this architecture are - network fault tolerance, - host fault tolerance, Tuexen [Page 1] Internet Draft Requirements for Reliable Server Pooling November 2000 - resistance against malicious attacks, - and scalability. 1. Introduction 1.1. Overview This document defines the requirements of the reliable server pooling architecture and the protocol used. Reliable server pools can be used for providing high available services by using a set of servers in a pool. Therefore fault tolerance is very important. Real time applications must also be supported which leads to requirements on the processing time needed. Scalability is another important requirement. Given that the server pool can be attacked by hackers, if one or more of the servers are hijacked then the server pool is compromised. Therefore, the security requirement is to catalog the threats to the reliable server pool and identify appropriate responses to those threats. 1.2. Terminology Pool: A collection of clients or servers providing the same service. Pool Element: A client or server which belongs to a pool. 1.3. Abbreviations SCTP: Stream Control Transmission Protocol TCP: Transmission Control Protocol 2. General Requirements The general architecture should be based on a peer to peer basis. However, the binding should be based on a client server model. It should be possible to use the protocol stack in small devices, like cellular phones. Therefore it should be possible to have lightweight implementations that are able to operate with the servers. Furthermore, it is expected that there is a transition phase with some systems are supporting the rserpool architecture and some are not. To Tuexen [Page 2] Internet Draft Requirements for Reliable Server Pooling November 2000 make this transition as seamless as possible it should be possible for hosts not supporting this architecture to use also the new pooling based services via some mechanism. There is also another transition process: it should be possible to adopt existing services to the new architecture to make use of the better services. The impact on existing protocols should be limited. Candidates for these services may be the adaptation layers developed in the SIGTRAN group and described in [RFC2719]. Another important requirement is that servers should be able to enter or leave the pool transparently without an interruption in service. The protocols used for the pool handling should not cause network congestion. This means that it should not generate heavy traffic, even in case of failures, and has to use flow control and congestion avoidance algorithms which are interoperable with currently deployed techniques, especially the flow control of TCP [RFC793] and SCTP [RFC2960]. Using this protocols it should be possible to implement an IP-based service without a single point of failure. This includes network failures and host failures. Especially, an IP-address is considered as a single point of failure. Therefore namespaces have to be used. 3. Namespaces and Pools Services are provided to the clients through a pool of servers. Clients will access these servers by name. Selection of a server in the pool will be performed on behalf of the client. The name resolution results in access to one specific server out of a pool of servers. The selection of the server is transparent to the client and is governed by server pool policy. Servers are registered by name in a namespace to join a server pool. There will be no globally unique namespace available, so multiple independent namespaces must be supported. Since it is necessary to support multiple namespaces, it should also be possible for a host to refer to entities in namespaces the host does not belong to. Editors note: Really? What about the proxies? It must also be possible for a host to be registered in more than one namespace or using multiple names in one namespace. A namespace can consist of a great number of pools. This is important since the system will be used for real time applications. So handling of name resolution has to be fast. Another consequence of the real time Tuexen [Page 3] Internet Draft Requirements for Reliable Server Pooling November 2000 requirement is the supervision on the pool entities. The name resolution should not result in a pool element which is not able to provide the required service. The registration and deregistration process is a dynamic one. It must be possible for hosts to register in a pool without affecting the other elements of a pool. This will be used for example, if a pool is under high load and more servers are installed to provide the service of the pool. It must also be possible to remove host from a pool without affecting the rest. Editors Note: Make clear if we can use DNS or not! 4. Server selection Services are provided by a pool of servers. If a client wants to connect to a server one of the servers of the pool has to be chosen. This functionality is called server selection. This selection is based on the server pooling policy. Server selection is driven by server pool policy. Some examples of selection policies are load balancing and round robin. The set of supported policies should be extensible in the sense that new policies can be added as required. The server selection should not be based on internal features of the underlying transport protocol. This means, for example, in the case of SCTP that only the state of associations will be taken into account and not the state of the paths of the associations. For specific applications it can be important that a client use a server of a pool, which was used before. Therefore the clients must be able to get handles of a server and to use this handle later in the server selection to connect to the same server again, if it is still possible. 5. Reliability aspects Host failures are handled by the pool concept. If one pool element fails and there are other pool elements which are able to provide the service than the other pool elements will be used. Transaction failover is not provided by reliable server pooling. If a host fails during processing of a transaction this transaction may be lost. Some services may provide a way to handle the failure, but this is not guaranteed. Network failures have to be handled by the underlying transport protocol. This means that the transport layer protocol must provide at least a acknowledged error-free non-duplicated transfer data delivery Tuexen [Page 4] Internet Draft Requirements for Reliable Server Pooling November 2000 service. Furthermore for fulfilling realtime requirements it may be necessary that the transport layer also provides network fault tolerance like the usage of multihoming in SCTP. 6. Security aspects Security is a major point of this architecture. There are several aspects which have to be considered: - The transport layer protocol used should support concepts for dealing with denial of service attacks. - It should not be possible for a host to register unauthorized in a pool. - It should not be possible for a host to deregister another host from a pool. - It should not be possible to attack the name resolution system. - The client-server relation should be based on an authentication. - It should be possible to encrypt the communication between the client and the host. Editors note: should the encryption be done at the application layer or at the transport layer. I think it should not be done at the rserpool layer. 7. Acknowledgements The author would like to thank the members of the rserpool discussion list, especially Maureen Stillman, for their valuable comments and suggestions. 8. References [RFC793] J. B. Postel, "Transmission Control Protocol", RFC 793, September 1981. [RFC2026] S. Bradner, "The Internet Standards Process -- Revision 3", RFC 2026, October 1996. [RFC2719] L. Ong et al., "Framework Architecture for Signaling Transport", RFC 2719, October 1999. [RFC2960] R. R. Stewart et al., "Stream Control Transmission Protocol", RFC 2960, November 2000. Tuexen [Page 5] Internet Draft Requirements for Reliable Server Pooling November 2000 9. Authors' Addresses Michael Tuexen Tel.: +49 89 722 47210 Siemens AG e-mail: Michael.Tuexen@icn.siemens.de ICN WN CS SE 51 D-81359 Munich Germany This Internet Draft expires April 20, 2001. Tuexen [Page 6]