rfc1104.txt

来自「RFC 的详细文档！」· 文本 代码 · 共 563 行 · 第 1/2 页


RFC 1104             Models of Policy Based Routing            June 1989


6. Policy based dynamic allocation of network resources (e.g.,
      bandwidth, buffers, etc.).

   Goals:

      Flexible and economical allocation of network resources based on
      current needs and certain policies.  Policies may be formulated at
      the network or Administrative Domain (AD) levels.  It is also
      possible to formulate policies which will regulate resource
      allocation for different types of traffic (e.g., Telnet, FTP,
      precedence indicators, network control traffic).

      Enforcement of policy based allocation of network resources might
      be implemented within the following parts of the network:

          routers for networks and Administrative Domain (AD) levels
          circuit switches for networks
          end systems establishing network connections

   Description:

      Policy based allocation of bandwidth could allow the modulation of
      the circuits of the networking infrastructure according to real
      time needs.  Assuming that available resources are limited towards
      an upper bound, the allocation of bandwidth would need to be
      controlled by policy.  One example might be a single end system
      that may or may not be allowed to, perhaps even automatically,
      take resources away from other end systems or users.  An example
      of dynamic bandwidth allocation is the currently implemented
      circuit switched IDNX component of the NSFNET, as well as the MCI
      Digital Reconfiguration Service (DRS) which is planned for the
      NSFNET later this year.

      Another model for resource allocation occurs at the packet level,
      where the allocation is controlled by multiple packet queues.
      This could allow for precedence queuing, with preferences based on
      some type of service and preferred forwarding of recognized
      critical data, such as network monitoring, control and routing.
      An example can be found in the NSFNET, where the NSFNET nodes
      prefer traffic affiliated with the NSFNET backbone network number
      over all other traffic, to allow for predictable passing of
      routing information as well as effective network monitoring and
      control.  At the other end of the spectrum, an implementation
      could also allow for queues of most deferrable traffic (such as
      large background file transfers).






Braun                                                           [Page 6]

RFC 1104             Models of Policy Based Routing            June 1989


   Benefits:

      Dynamic allocation of bandwidth could allow for a truly flexible
      environment where the networking infrastructure could create
      bandwidth on a per need basis.  This could result in significant
      cost reductions during times when little bandwidth is needed.
      This method could potentially accommodate real time transient high
      bandwidth requirements, potentially by reducing the bandwidth
      available to other parts of the infrastructure.  A positive aspect
      is that the bandwidth allocation could be protocol independent,
      with no impact on routing protocols or packet forwarding
      performance.

      Policy based allocation of bandwidth can provide a predictable
      dynamic environment.  The rules about allocation of bandwidth at
      the circuit level or at the packet level need to be determined by
      a consistent and predictable policy, so that other networks or
      Administrative Domains can tune their allocation of networking
      resources at the same time.

   Concerns:

      The policies involved in making dynamic bandwidth allocation in a
      largely packet switching environment possible are still in the
      development phase.  Even the technical implications of
      infrastructure reconfiguration in result of events happening on a
      higher level still requires additional research.

      A policy based allocation of bandwidth could tune the network to
      good performance, but could cause networks located in other
      Administrative Domains to pass traffic poorly.  It is important
      that network resource policy information for a network be
      discussed within the context of its Administrative Domain.
      Administrative Domains need to discuss their network resource
      allocation policies with other Administrative Domains.

      The technical problem of sharing network resource policy
      information could be solved by a making a "network resource policy
      information" database available to all administrators of networks
      and Administrative Domains.  However, the political problems
      involved in creating a network resource policy with impact on
      multiple Administrative Domains does still require additional
      study.

7. Discussion

   Both the first and the second model of policy based routing are
   similar in the sense that their goal is to enforce certain flows.



Braun                                                           [Page 7]

RFC 1104             Models of Policy Based Routing            June 1989


   This enforcement allows the control of access to scarce network
   resources (if the resource is not scarce, there is no performance
   reason to control access to it).  The major difference is the level
   of enforcement: macroscopic level versus microscopic level control.

   Associated with the enforcement for a certain network resource is the
   cost.  If this cost is higher than the cost required to make a
   particular resource less scarce, then the feasibility of enforcement
   may be questionable.

   If portions of the Internet find that microscopic enforcement of
   policy is necessary, then this will need to be implementable without
   significant performance degradation to the networking environment at
   large.  Local policies within specific Routing Domains or
   Administrative Domains should not affect global Internet traffic or
   routing.  Policies within Administrative Domains which act as traffic
   transit systems (such as the NSFNET) should not be affected by
   policies a single network imposes for its local benefit.

   Some models of policy routing are trying to deal with cases where
   network resources require rather complex usage policies.  One of
   scenarios in [4] is one in which a specific agency may have some
   network resource (in the example it is a link) which is sometimes
   underutilized.  The goal is to sell this resource to other agencies
   during the underutilization period to recover expenses.  This
   situation is equivalent to the problem of finding optimum routes,
   with respect to a certain TOS, in the presence of network resources
   (e.g., links) with variable characteristics.  Any proposed solution
   to this problem should address such issues as network and route
   stability.  More feasibility study is necessary for the whole
   approach where links used for global communication are also subject
   to arbitrary local policies.  An alternative approach would be to
   reconfigure the network topology so that underutilized links will be
   dropped and possibly returned to the phone company.  This is
   comparable to what the NSFNET is planning on doing with the MCI
   Digital Reconfiguration Service (DRS).  A DRS model may appear
   cleaner and more easy to implement than a complicated model like the
   one outlined in [4].

   The models for policy based routing emphasize that careful
   engineering of the Internet needs to decided upon the profile of
   traffic during normal times, outage periods, and peak loads.  This
   type of engineering is not a new requirement.  However, there could
   potentially be a significant benefit in deciding these policies ahead
   of time and using policy based routing to implement specific routing
   policies.





Braun                                                           [Page 8]

RFC 1104             Models of Policy Based Routing            June 1989


8. Accounting vs. Policy Based Routing

   Quite often Accounting and Policy Based Routing are discussed
   together.  While the application of both Accounting and Policy Based
   Routing is to control access to scarce network resources, these are
   separate (but related) issues.

   The chief difference between Accounting and Policy Based Routing is
   that Accounting combines history information with policy information
   to track network usage for various purposes.  Accounting information
   may in turn drive policy mechanisms (for instance, one could imagine
   a policy limiting a certain organization to a fixed aggregate
   percentage of dynamically shared bandwidth).  Conversely, policy
   information may affect accounting issues.  Network accounting
   typically involves route information (at any level from AD to end
   system) and volume information (packet, octet counts).

   Accounting may be implemented in conjunction with any of the policy
   models mentioned above.  Similar to the microscopic versus
   macroscopic policies, accounting may be classified into different
   levels.  One may collect accounting data at the AD level, network
   level, host level, or even at the individual user level.  However,
   since accounting may be organized hierarchically, microscopic
   accounting may be supported at the network or host level, while
   macroscopic accounting may be supported at the network or AD level.
   An example might be the amount of traffic passed at the interface
   between the NSFNET and a mid-level network or between a mid-level
   network and a campus.  Furthermore, the NSFNET has facilities
   implemented to allow for accounting of traffic trends from individual
   network numbers as well as application-specific information.

   Full-blown accounting schemes suffer the same types of concerns
   previously discussed, with the added complication of potentially
   large amounts of additional data gathered that must be reliably
   retrieved.  As pointed out in [4], policy issues may impact the way
   accounting data is collected (one administration billing for packets
   that were then dropped in the network of another administration).
   Microscopic accounting may not scale well in a large internet.

   Furthermore, from the standpoint of billing, it is not clear that the
   services provided at the network layer map well to the sorts of
   services that network consumers are willing to pay for.  In the
   telephone network (as well as public data networks), users pay for
   end-to-end service and expect good quality service in terms of error
   rate and delay (and may be unwilling to pay for service that is
   viewed as unacceptable).  In an internetworking environment, the
   heterogeneous administrative environment combined with the lack of
   end-to-end control may make this approach infeasible.



Braun                                                           [Page 9]

RFC 1104             Models of Policy Based Routing            June 1989


   Lightweight approaches to accounting can be used (with less impact)
   when specific, limited goals are set.  One suggested approach
   involves monitoring traffic patterns.  If a pattern of abuse (e.g.,
   unauthorized use) develops, an accounting system could track this and
   allow corrective action to be taken, by changing routing policy or
   imposing access control (blocking hosts or nets).  Note that this is
   much less intrusive into the packet forwarding aspects of the
   routers, but requires distribution of a policy database that the
   accounting system can use to reduce the raw information.  Because
   this approach is statistical in nature, it may be slow to react.

9. References

   [1] Rekhter, Y., "EGP and Policy Based Routing in the New NSFNET
       Backbone", RFC 1092, IBM Research, February 1989.

   [2] Braun, H-W., "The NSFNET Routing Architecture", RFC 1093,
       Merit/NSFNET Project, February 1989.

   [3] Collins, M., and R. Nitzan, "ESNET Routing", DRAFT Version 1.0,
       LLNL, May 1989.

   [4] Clark, D., "Policy Routing in Internet Protocols", RFC 1102,
       M.I.T. Laboratory for Computer Science, May 1989.

Author's Address

   Hans-Werner Braun
   Merit Computer Network
   University of Michigan
   1075 Beal Avenue
   Ann Arbor, Michigan 48109

   Telephone:      [phone]
   Fax:            313 747-3745
   EMail:          [e-mail]















Braun                                                          [Page 10]