Directory Assistance Problem and High-Level Design Proposal

Abstract

Edge TRILL (Transparent Interconnection of Lots of Links) switches currently learn the mapping between MAC (Media Access Control) addresses and their egress TRILL switch by observing the data packets they ingress or egress or by the TRILL ESADI (End-Station Address Distribution Information) protocol. When an ingress TRILL switch receives a data frame for a destination address (MAC&Label) that the switch does not know, the data frame is flooded within the frame's Data Label across the TRILL campus.

This document describes the framework for using directory services to assist edge TRILL switches in reducing multi-destination frames, particularly unknown unicast frames flooding, and ARP/ND (Address Resolution Protocol / Neighbor Discovery), thus improving TRILL network scalability and security.

Status of This Memo

This document is not an Internet Standards Track specification; it is published for informational purposes.

This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Not all documents approved by the IESG are a candidate for any level of Internet Standard; see Section 2 of RFC 5741.

Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc7067.

Copyright Notice

Copyright © 2013 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.

Table of Contents

   1. Introduction ....................................................3
   2. Terminology .....................................................4
   3. Impact of Massive Number of End Stations ........................5
      3.1. Issues of Flooding-Based Learning in Data Centers ..........5
      3.2. Two Examples ...............................................6
   4. Benefits of Directory-Assisted TRILL Edge .......................7
   5. Generic Operation of Directory Assistance .......................8
      5.1. Information in Directory for Edge RBridges .................8
      5.2. Push Model and Requirements ................................9
      5.3. Pull Model and Requirements ...............................11
   6. Recommendation .................................................12
   7. Security Considerations ........................................12
   8. Acknowledgements ...............................................13
   9. Informative References .........................................14

1. Introduction

Edge TRILL (Transparent Interconnection of Lots of Links) switches (devices implementing [RFC6325], also known as RBridges) currently learn the mapping between destination MAC addresses and their egress TRILL switch by observing data packets or by the ESADI (End-Station Address Distribution Information) protocol. When an ingress RBridge (Routing Bridge) receives a data frame for a destination address (MAC&Label) that RBridge does not know, the data frame is flooded within that Data Label across the TRILL campus. (Data Labels are VLANs or FGLs (Fine-Grained Labels [FGL]).

This document describes a framework for using directory services in environments where such services are available, such as typical data centers, to assist edge TRILL switches. This assistance can reduce multi-destination frames, particularly ARP [RFC826], ND [RFC4861], and unknown unicast, thus improving TRILL network scalability. In addition, the information provided by a directory can be more secure than that learned from the data plane (see Section 7).

Data centers, especially Internet and/or multi-tenant data centers, tend to have a large number of end stations with a wide variety of applications. Their networks differ from enterprise campus networks in several ways that make them attractive for the use of directory assistance, in particular:

1. Data center topology is often based on racks and rows. Furthermore, a Server/VM (virtual machine) Management System orchestrates the assignment of guest operating systems to servers, racks, and rows; it is not done at random. So, the information necessary for a directory is normally available from that Management System.

2. Rapid workload shifting in data centers can accelerate the frequency of the physical servers being reloaded with different applications. Sometimes, applications loaded into one physical server at different times can belong to different subnets. When a VM is moved to a new location or when a server is loaded with a new application with different IP/MAC addresses, it is more likely that the destination address of data packets sent out from those VMs are unknown to their attached edge RBridges.

3. With server virtualization, there is an increasing trend to dynamically create or delete VMs when the demand for resources changes, to move VMs from overloaded servers to less loaded servers, or to aggregate VMs onto fewer servers when demand is light. This results in the more frequent occurrence of multiple subnets on the same port at the same time and a higher change rate for VMs than for physical servers.

Both items 2 and 3 above can lead to applications in one subnet being placed in different locations (racks or rows) or one rack having applications belonging to different subnets.

2. Terminology

The terms "VLAN" and "Data Label" are used interchangeably with "Subnet" in this document, because it is common to map one subnet to one VLAN or FGL.

   Bridge:  Device compliant with IEEE Std 802.1Q-2011 [802.1Q].
   
   Data Label:  VLAN or FGL
   
   EoR:     End-of-Row switches in a data center.  Also known as
            aggregation switches.
   
   End Station:  Guest OS running on a physical server or on a virtual
            machine.  An end station in this document has at least one
            IP address, at least one MAC address, and is connected to a
            network.
   
   FGL:     Fine-Grained Label [FGL]
   
   IS-IS:   Intermediate System to Intermediate System routing protocol.
            TRILL uses IS-IS [IS-IS] [RFC6326].

RBridge:

"Routing Bridge", an alternate name for a TRILL switch.

   ToR:     Top-of-Rack switches in a data center.  Also known as access
            switches in some data centers.
   
   TRILL:   Transparent Interconnection of Lots of Links [RFC6325]
   
   TRILL Switch:  A device implementing the TRILL protocol [RFC6325].
   
   VM:      Virtual Machine

3. Impact of Massive Number of End Stations

This section discusses the impact of a massive number of end stations in a TRILL campus using Data Centers as an example.

3.1. Issues of Flooding-Based Learning in Data Centers

It is common for Data Center networks to have multiple tiers of switches, for example, one or two Access Switches for each server rack (ToR), aggregation switches for some rows (or EoR switches), and some core switches to interconnect the aggregation switches. Many aggregation switches deployed in data centers have high port density. It is not uncommon to see aggregation switches interconnecting hundreds of ToR switches.

                       +-------+         +------+
                     +/------+ |       +/-----+ |
                     | Aggr11| + ----- |AggrN1| +    EoR switches
                     +---+---+/        +------+/
                      /     \            /      \
                     /       \          /        \
                  +---+    +---+      +---+     +---+
                  |T11|... |T1x|      |T21| ..  |T2y| ToR switches
                  +---+    +---+      +---+     +---+
                    |        |          |         |
                  +-|-+    +-|-+      +-|-+     +-|-+
                  |   |... |   |      |   | ..  |   |
                  +---+    +---+      +---+     +---+ Server racks
                  |   |... |   |      |   | ..  |   |
                  +---+    +---+      +---+     +---+
                  |   |... |   |      |   | ..  |   |
                  +---+    +---+      +---+     +---+

Figure 1: Typical Data Center Network Design

The following problems could occur when TRILL is deployed in a data center with a large number of end stations and when the end stations in one subnet/Label are placed under multiple edge RBridges:

• Unnecessary filling of slots in the MAC address learning table of edge RBridges, e.g., RBridge T11, due to T11 receiving broadcast/multicast traffic (e.g., ARP/ND, cluster multicast, etc.) from end stations under other edge RBridges that are not actually communicating with any end stations attached to T11.

• Packets being flooded across a TRILL campus when their destination MAC addresses are not in the ingress RBridge's MAC address to the egress RBridge cache.

3.2. Two Examples

Consider a data center with 1,600 server racks. Each server rack has at least one ToR switch. The ToR switches are further divided into 8 groups, with each group being connected by a set of aggregation switches. There could be 4 to 8 aggregation switches in each set to achieve load sharing for traffic to/from server racks. Let's consider the following two scenarios for the TRILL campus boundary if TRILL is deployed in this data center environment:

• Scenario #1: TRILL campus boundary starts at the ToR switches:

If each server rack has one ToR, there are 1,600 edge RBridges. If each rack has two ToR switches, then there will be 3,200 edge RBridges.

In this scenario, the TRILL campus will have more than 1,600 (or 3,200) + 8*4 (or 8*8) nodes, which is a large IS-IS area. Even though a mesh IS-IS area can scale up to thousands of nodes, it is challenging for aggregation switches to handle IS-IS link state advertisement among hundreds of parallel ports.

If each ToR has 40 downstream ports facing servers and each server has 10 VMs, there could be 40*10 = 400 end stations attached. If those end stations belong to 8 Labels, then the total number of MAC&Label entries learned by each edge RBridge in the worst case might be 400*8 = 3,200, which is not a large number.

• Scenario #2: TRILL campus boundary starts at the aggregation switches:

With the same assumptions as before, the number of nodes in the TRILL campus will be less than 100, and aggregation switches don't have to handle IS-IS link state advertisements among hundreds of parallel ports.

However, the number of MAC&Label <-> Egress RBridge mapping entries to be learned and managed by the RBridge edge node can be very large. In the example above, each edge RBridge has 200 edge ports facing the ToR switches. If each ToR has 40 downstream ports facing servers and each server has 10 VMs, there could be 200*40*10 = 80,000 end stations attached. If all those end stations belong to 1,600 Labels (50 per Data Label) and each Data Label has 200 end stations, then under the worst-case scenario, the total number of MAC&Label entries to be learned by each edge RBridge can be 1,600*200=320,000, which is very large.

4. Benefits of Directory-Assisted TRILL Edge

In some environments, particularly data centers, the assignment of applications to servers, including rack and row selection, is orchestrated by Server (or VM) Management System(s). That is, there is a database or multiple databases that have the knowledge of where each application is placed. If the application location information can be fed to RBridge edge nodes through some form of directory service, then there is much less chance of RBridge edge nodes receiving unknown MAC destination addresses, therefore less chance of flooding.

Avoiding unknown unicast address flooding to the TRILL campus is especially valuable in the data center environment, because there is a higher chance of an edge RBridge receiving packets with an unknown unicast destination address and broadcast/multicast messages due to VM migration and servers being loaded with different applications. When a VM is moved to a new location or a server is loaded with a new application with a different IP/MAC addresses, it is more likely that the destination address of data packets sent out from those VMs is unknown to their attached edge RBridges. In addition, gratuitous ARP (IPv4 [RFC826]) or Unsolicited Neighbor Advertisement (IPv6 [RFC4861]) sent out from those newly migrated or activated VMs have to be flooded to other edge RBridges that have VMs in the same subnets.

The benefits of using directory assistance include:

• Avoids flooding an unknown unicast destination address across the TRILL campus. The directory-enforced MAC&Label <-> Egress RBridge mapping table can determine if a data packet needs to be forwarded across the TRILL campus.

When multiple RBridge edge ports are connected to end stations (servers/VMs), possibly via bridged LANs, a directory-assisted edge RBridge won't need to flood unknown unicast destination data frames to all ports of the edge RBridges in the frame's Data Label when it ingresses a frame. It can depend on the directory to locate the destination. When the directory doesn't have the needed information, the frames can be dropped or flooded depending on the policy configured.

• Reduces flooding of decapsulated Ethernet frames with an unknown MAC destination address to a bridged LAN connected to RBridge edge ports.

When an RBridge receives a unicast TRILL data packet whose destination Nickname matches with its own, the normal procedure is for the RBridge to decapsulate it and forward the decapsulated Ethernet frame to the directly attached bridged LAN. If the destination MAC is unknown, the RBridge floods the decapsulated Ethernet frame out all ports in the frame's Data Label. With directory assistance, the egress RBridge can determine if the MAC destination address in a frame matches any end stations attached via the bridged LAN. Frames can be discarded if their destination addresses do not match.

• Reduces the amount of MAC&Label <-> Egress RBridge mapping maintained by edge RBridges. There is no need for an edge RBridge to keep MAC entries of remote end stations that don't communicate with the end stations locally attached.

• Eliminates ARP/ND being broadcast or multicast through the TRILL core.

• Provides some protection against spoofing of source addresses (see Section 7).

5. Generic Operation of Directory Assistance

There are two different models for directory assistance to edge RBridges: Push Model and Pull Model. The directory information is described in Section 5.1 below, while Section 5.2 discusses Push Model requirements, and Section 5.3 Pull Model requirements.

5.1. Information in Directory for Edge RBridges

To achieve the benefits of directory assistance for TRILL, the corresponding Directory Server entries will need, at a minimum, the following logical data structure:

   [IP, MAC, Data Label, {list of attached RBridge nicknames}, {list of
   interested RBridges}]

The {list of attached RBridges} are the edge RBridges to which the host (or VM) is attached as specified by the [IP, MAC, Data Label] in the entry. The {list of interested RBridges} are the remote RBridges that might have attached hosts that communicate with the host in this entry.

When a host has multiple IP addresses, there will be multiple entries.

The {list of interested RBridges} could get populated when an RBridge queries for information, or information is pushed from a Directory Server. The list is used to notify those RBridges when the host (specified by the [IP, MAC, Data Label]) in the entry changes its RBridge attachment. An explicit list in the directory is not needed as long as the interested RBridges can be determined.

5.2. Push Model and Requirements

Under this model, Directory Server(s) push the MAC&Label <-> Egress RBridge mapping for all the end stations that might communicate with end stations attached to an RBridge edge node. If the packet's destination address can't be found in the MAC&Label <-> Egress RBridge table, the Ingress RBridge could be configured to:

simply drop a data packet,

flood it to the TRILL campus, or

start the pull process to get information from the Pull Directory Server(s).

It may not be necessary for every edge RBridge to get the entire mapping table for all the end stations in a campus. There are many ways to narrow the full set down to a smaller set of remote end stations that communicate with end stations attached to an edge RBridge. A simple approach is to only push the mapping for the Data Labels that have active end stations under an edge RBridge. This approach can reduce the number of mapping entries being pushed.

However, the Push Model will usually push more entries of MAC&Label <-> Egress RBridge mapping to an edge RBridges than needed. Under the normal process of edge RBridge cache aging and unknown destination address flooding, rarely used mapping entries would have been removed. But it can be difficult for Directory Servers to predict the communication patterns among applications within one Data Label. Therefore, it is likely that the Directory Servers will push down all the MAC&Label entries if there are end stations in the Data Label attached to the edge RBridge. This is a disadvantage of the Push Model compared with the Pull Model described below.

In the Push Model, it is necessary to have a way for an RBridge node to request Directory Server(s) to push the mapping entries. This method should at least include the Data Labels enabled on the RBridge, so that the Directory Server doesn't need to push down the entire set of mapping entries for all the end stations in the campus. An RBridge must be able to get mapping entries when it is initialized or restarted.

The Push Model's detailed method and any handshake mechanism between an RBridge and Directory Server(s) is beyond the scope of this framework document.

When a Directory Server needs to push a large number of entries to edge RBridges, efficient data organization should be considered, for example, with one edge RBridge nickname being associated with all the attached end stations' MAC addresses and Data Labels. As shown in Table 1 below, to make the data more compact, a representation can be used where a nickname need only occur once for a set of Labels, each of which occurs only once and each of which is associated with a set of multiple IP and MAC address pairs. It would be much more bulky to have each IP and MAC address pair separately accompanied by its Label and by the nickname of the RBridge by which it is reachable.

         +------------+---------+--------------------------------+
         | Nickname1  |Label-1  | IP/MAC1, IP/MAC2, ,, IP/MACn   |
         |            |-------- +--------------------------------+
         |            |Label-2  | IP/MAC1, IP/MAC2, ,, IP/MACn   |
         |            |-------- +--------------------------------+
         |            |  ...... | IP/MAC1, IP/MAC2, ,, IP/MACn   |
         +------------+-------- +--------------------------------+
         | Nickname2  |Label-1  | IP/MAC1, IP/MAC2, ,, IP/MACn   |
         |            |-------- +--------------------------------+
         |            |Label-2  | IP/MAC1, IP/MAC2, ,,IP/MACn    |
         |            |-------- +--------------------------------+
         |            |         | IP/MAC1, IP/MAC2, ,, IP/MACn   |
         +------------+-------- +--------------------------------+
         | -------    |-------- +--------------------------------+
         |            |         | IP/MAC1, IP/MAC2, ,, IP/MACn   |
         +------------+-------- +--------------------------------+

Table 1: Summarized Table Pushed Down from Directory

Whenever there is any change in MAC&Label <-> Egress RBridge mapping that can be triggered by end stations being added, moved, or decommissioned, an incremental update can be sent to the edge RBridges that are impacted by the change. Therefore, something like a sequence number has to be maintained by Directory Servers and RBridges. Detailed mechanisms will be specified in a separate document.

5.3. Pull Model and Requirements

Under this model, an RBridge pulls the MAC&Label <-> Egress RBridge mapping entry from the Directory Server when its cache doesn't have the entry. There are a couple of possibilities for triggering the pulling process:

• The RBridge edge node can send a pull request whenever it receives an unknown MAC destination, or

• The RBridge edge node can intercept all ARP/ND requests and forward them or appropriate requests to the Directory Server(s) that has the information on where the target end stations are located.

The Pull Directory response could indicate that the address being queried is unknown or that the requestor is administratively prohibited from getting an informative response.

By using a Pull Directory, a frame with an unknown MAC destination address doesn't have to be flooded across the TRILL campus and the ARP/ND requests don't have to be broadcast or multicast across the TRILL campus.

The ingress RBridge can cache the response pulled from the directory. The timer for such a cache should be short in an environment where VMs move frequently. The cache timer could be configured by the Management System or sent along with the Pulled reply by the Directory Server(s). It is important that the cached information be kept consistent with the actual placement of addresses in the campus; therefore, there needs to be some mechanism by which RBridges that have pulled information that has not expired can be informed when that information changes or becomes invalid for other reasons.

One advantage of the Pull Model is that edge RBridges can age out MAC&Label entries if they haven't been used for a certain configured period of time or a period of time provided by the directory. Therefore, each edge RBridge will only keep the entries that are frequently used, so its mapping table size will be smaller. Edge RBridges would query the Directory Server(s) for unknown MAC destination addresses in data frames or ARP/ND and cache the response. When end stations attached to remote edge RBridges rarely communicate with the locally attached end stations, the corresponding MAC&VLAN entries would be aged out from the RBridge's cache.

An RBridge waiting for a response from Directory Servers upon receiving a data frame with an unknown destination address is similar to an Layer-3/Layer-2 boundary router waiting for an ARP or ND response upon receiving an IP data packet whose destination IP is not in the router's IP/MAC cache table. Most deployed routers today do hold the packet and send ARP/ND requests to the target upon receiving a packet with a destination IP not in its IP-to-MAC cache. When ARP/ND replies are received, the router will send the data packet to the target. This practice minimizes flooding when targets don't exist in the subnet.

When the target doesn't exist in the subnet, routers generally resend an ARP/ND request a few more times before dropping the packets. So, if the target doesn't exist in the subnet, the router's holding time to wait for an ARP/ND response can be longer than the time taken by the Pull Model to get IP-to-MAC mapping from a Directory Server.

RBridges with mapping entries being pushed from a Directory Server can be configured to use the Pull Model for targets that don't exist in the mapping data being pushed.

A separate document will specify the detailed messages and mechanism for RBridges to pull information from Directory Server(s).

6. Recommendation

TRILL should provide a directory-assisted approach. This document describes a basic framework for directory assistance to RBridge edge nodes. More detailed mechanisms will be described in a separate document or documents.

7. Security Considerations

For general TRILL security considerations, see Section 6 of [RFC6325].

Accurate mapping of IP addresses into MAC addresses and of MAC addresses to the RBridges from which they are reachable is important to the correct delivery of information. The security of specific directory-assisted mechanisms for delivering such information will be discussed in the document or documents specifying those mechanisms.

A directory-assisted TRILL edge can be used to substantially improve the security of a TRILL campus over TRILL's default MAC address learning from the data plane. Assume S is an end station attached to RB1 trying to spoof a target end station T and that T is attached to RB2. Perhaps S wants to steal traffic intended for T or forge traffic as if it was from T.

With that default TRILL data-plane learning as described in [RFC6325], S can impersonate T or any other end station in the same Data Label (VLAN or FGL [FGL]) as S and possibly other Data Labels, depending on how tightly VLAN admission and Appointed Forwarders [RFC6439] are configured at the port by which S is connected to RB1. S can just send native frames with the forged source MAC addresses of T, perhaps broadcast frames for maximum effectiveness. With this technique, S will frequently receive traffic intended for T and S can easily forge traffic as being from T.

Such spoofing can be prevented to the extent that the network RBridges (1) use trusted directory services as described above in this document, (2) discard native frames received from a local end station when the directory says that end stations should be remote, and, (3) when appropriate, intercept ARP and ND messages and respond locally. Under these circumstances, S would be limited to spoofing targets on the same RBridge as the ingress RBridge for S (that is, RB1 = RB2). RB1 would still need to learn which local end stations were attached to which port, and S could confuse RB1 by sending frames with the forged source MAC address of other end stations on RB1. Although it would also still be restricted to frames in a VLAN that would both be admitted by S's port of attachment and for which that port is an Appointed Forwarder.

Security against spoofing could be even further strengthened by adding port of attachment information to the directory and discarding native frames that are received on the wrong port. This would limit S to spoofing targets that were on the same link as S and in a VLAN admitted by the port of that link's attachment to RB1 and for which that port is an Appointed Forwarder (or, if the link is multiply connected, in the same way at all of the ports by which the link is attached to an RBridge).

Even without directory services, secure ND [RFC3971] or use of secure ESADI (as described in [ESADI]) may also be helpful to security.

8. Acknowledgements

Thanks for comments and review from the following:

Sam Aldrin, David Black, Charlie Kaufman, Yizhou Li, and Erik Nordmark

9. Informative References

   [802.1Q]   IEEE Std 802.1Q-2011, "IEEE Standard for Local and
              metropolitan area networks - Virtual Bridged Local Area
              Networks", May 2011.
   
   [IS-IS]    ISO/IEC, "Intermediate System to Intermediate System
              intra-domain routeing information exchange protocol for
              use in conjunction with the protocol for providing the
              connectionless-mode network service (ISO 8473)", ISO/IEC
              10589:2002.
   
   [RFC826]   Plummer, D., "Ethernet Address Resolution Protocol: Or
              Converting Network Protocol Addresses to 48.bit Ethernet
              Address for Transmission on Ethernet Hardware", STD 37,
              RFC 826, November 1982.
   
   [RFC3971]  Arkko, J., Ed., Kempf, J., Zill, B., and P. Nikander,
              "SEcure Neighbor Discovery (SEND)", RFC 3971, March 2005.
   
   [RFC4861]  Narten, T., Nordmark, E., Simpson, W., and H. Soliman,
              "Neighbor Discovery for IP version 6 (IPv6)", RFC 4861,
              September 2007.
   
   [RFC6325]  Perlman, R., Eastlake 3rd, D., Dutt, D., Gai, S., and A.
              Ghanwani, "Routing Bridges (RBridges): Base Protocol
              Specification", RFC 6325, July 2011.
   
   [RFC6326]  Eastlake, D., Banerjee, A., Dutt, D., Perlman, R., and A.
              Ghanwani, "Transparent Interconnection of Lots of Links
              (TRILL) Use of IS-IS", RFC 6326, July 2011.
   
   [RFC6439]  Perlman, R., Eastlake, D., Li, Y., Banerjee, A., and F.
              Hu, "Routing Bridges (RBridges): Appointed Forwarders",
              RFC 6439, November 2011.
   
   [ESADI]    Zhai, H., Hu, F., Perlman, R., Eastlake 3rd, D., and O.
              Stokes, "TRILL (Transparent Interconnection of Lots of
              Links): ESADI (End Station Address Distribution
              Information) Protocol", Work in Progress, July 2013.
   
   [FGL]      Eastlake 3rd, D., Zhang, M., Agarwal, P., Perlman, R., and
              D. Dutt, "TRILL (Transparent Interconnection of Lots of
              Links): Fine-Grained Labeling", Work in Progress, May
              2013.

Authors' Addresses

Linda Dunbar
Huawei Technologies
5430 Legacy Drive, Suite #175
Plano, TX 75024, USA

   Phone: +1-469-277-5840
   EMail: ldunbar@huawei.com

Donald Eastlake
Huawei Technologies
155 Beaver Street
Milford, MA 01757 USA

   Phone: +1-508-333-2270
   EMail: d3e3e3@gmail.com

Radia Perlman
Intel Labs
2200 Mission College Blvd.
Santa Clara, CA 95054-1549 USA

   Phone: +1-408-765-8080
   EMail: Radia@alum.mit.edu

Igor Gashinsky
Yahoo
45 West 18th Street 6th floor
New York, NY 10011 USA

   EMail: igor@yahoo-inc.com