Airline Product Set Training Supplement

• Describe how the Airline Product Set (ALPS) feature operates
• Discuss the problems inherent to current airline networks and Cisco's solution
• Explain Cisco's three-phase plan for network migration

To support these objectives, this training supplement contains the following (major) sections:

Table of Contents

Feature Description
Target Audience
List of Terms
Current Airline Network
Customer Migration Requirements
SITA
Current Airline Topologies
Cisco's Three-Phase Delivery Plan
Mapping of Airline Traffic over IP (MATIP)
Platforms
Airline Control (ALC) Protocol
ALPS IOS Architecture
Configuration Example
ALC Router, or remote customer premises equipment (CPE)
Primary Central CPE (Data Center Router)
Backup Central CPE (Data Center Router)
Feature Review Questions

Feature Description: ALPS

Cisco's ALPS feature provides an end-to-end solution for airlines or central reservation systems. The ALPS feature is integrated in the Cisco IOS software and allows airlines to replace their existing hardware and software with Cisco routers. For customers who already use Cisco routers, this feature allows them to consolidate networking overhead and functionality.

Airline protocols are some of the oldest still being used in commercial networks today. The reservation systems used to book airplane flights and hotel rooms were among the first computer applications put out over telephone lines, predating both SNA and Binary Synchronous Communications (Bisync). Airlines and travel agencies now require the functionality, security, and reliability provided by modern networks, but are faced with the problem of retaining their existing applications. The problem is how to update the networks while maintaining the current applications.

This training supplement is an overview of Cisco's Airline Product Set (formerly called Airline Protocol Support). It outlines the strategy and product set currently being developed by Cisco in cooperation with the Societe Internationale de Telecommunications Aeronautiques (SITA).

Target Audience

These customers will have some or all of the following in their network:

• A mainframe running IBM's Transaction processing facility (TPF) or Unisys’ operating system
• A front-end processor (FEP) that connects to the mainframe via either AX.25 or EMTOX
• Remote terminal concentrators, often called agent set control units (ASCUs)
• Terminals and printers connected to those ASCUs via a remote access protocol, either P1024B in the IBM environment or P1024C in the Unisys environment

Current Airline Network Dilemma

Customers use terminals and printers to make airline reservations, print tickets and boarding passes, and perform other travel-related functions. Unfortunately, the network that currently connects these terminals is isolated from the airline's other networks. Typically, an airline has an SNA network and a TCP/IP network, in addition to its legacy airline protocol network. By connecting the legacy airline network into the router, Cisco makes it possible to consolidate those three networks into one standards-based network.

This diagram shows the current networking environments in place at most airlines. Most major airline networks evolved from the same starting point, so their current situations and migration requirements are similar. The TPF/Unisys graphic represents the oldest system, while the others are subsequent developments. In most airline networks, all three systems run simultaneously, yet distinctly.

Customer Migration Requirements

The ALC/SITA network could be either an airline’s own leased network or it could be connected to SITA’s network. The SNA and IP networks are typically private.

MVS Systems

Some smaller airlines use a program similar to TPF called Airline Control System (ALCS), which was developed by IBM and operates on Multiple Virtual Storage (MVS). The MVS System is a newer development and is used for the business applications; finance, personnel, payroll, etc. This is a standalone, standard SNA/SDLC network.

UNIX or NT Systems

The UNIX or NT system depicted is a combination of network media and computing platforms bound together by the use of TCP/IP. It includes PCs for word processing, spreadsheets, etc. UNIX or Windows NT are typically used for operational management; data warehousing, for example.

Other smaller airlines outsource their reservations system to another airline or service provider and do not maintain a separate reservation system. Virtually all large airlines, however, have at least one TPF system, and the largest airlines have many.

Customer Migration Requirements

Maintaining separate networks for TPF, SNA, and TCP/IP implies overhead in costs and complexity if users have to access applications using more than one protocol. P1024B protocol networking eliminates the need for a separate network for TPF by incorporating that traffic into the TCP/IP network. Furthermore, SNA traffic can be consolidated into a TCP/IP network using technologies such as data-link switching (DLSw). With a single TCP/IP network, cost savings can be realized in line costs, management, and a reduction in customized software and hardware. Customer migration requirements come directly from the dilemma of the three parallel networks described on the previous slide.

The first requirement is to reduce cost.

Customers hope to achieve economies of scale gained by consolidating three separate data networks onto a common backbone. If a network vendor were to provide a consolidated backbone, this simplified network infrastructure could be configured, managed, and troubleshot using one set of applications, which reduces the network support costs.

Once the consolidated data network has been established, the customer could move to add voice support to the same network. For many airlines, about 50 percent of their communications budget is spent on voice.

Customers also need a migration strategy. The customers have a need to access information across platforms, especially from the mainframe hosts to the UNIX workstations. The network architecture would be optimized for applications such as data mining.

Airlines require security and QoS guarantees before they will consider opening their environments to outside users and customers. Many smaller airlines buy their mainframe time from a service or from their larger competitors. Security of this internal data is important to them. They require QoS measuring tools to be sure they are providing proper service to their customers and also to monitor the QoS received from their service provider.

Internet access via the World Wide Web (Web) opens additional business opportunities. When external customers (the travel agency or end-user customers) want to dial in, they want to do it via the Web and use standard browser interfaces. In addition, airlines can now include their legacy applications in their own intranet.

Cisco Airline Internetworking Initiatives

Cisco plans to meet the airlines’ requirements with a coordinated set of products that provide a roadmap for migrating to standards-based platforms. Cisco is working on four fronts:

• Remote legacy protocol support
• High-speed, scalable data center solutions
• Leveraging Cisco IOS QoS, security, and voice features for the network backbone
• A coordinated effort with SITA and the standards bodies for MATIP, the next generation of Airline-to-IP support, as well as participation in other air transport industry associations

For the future, Cisco is working with SITA on the definition of the next generation of ALC-to-host access in the form of a proposed standard called MATIP. This proposal is described later in this supplement.

SITA

As part of our execution strategy, Cisco is partnering with Societe Internationale de Telecommunications Aeronautiques (SITA) to develop and test airline features in SITA’s network. SITA is an international value-added backbone network provider whose backbone network is considered the benchmark for both the airline and the travel industry overall. In addition to providing the bandwidth and access mechanism, they provide configuration, troubleshooting, and network management features.

Cisco is jointly developing P1024B and P1024C protocol networking as well as the AX.25 and EMTOX components of ALPS with SITA. Cisco and SITA are jointly coordinating and supporting the beta sites. All beta sites are existing SITA customers. After the completion of beta testing, SITA will have a 90-day window of exclusivity with the product. It will sell these products to existing and potential customers as part of its many service offerings. After 90 days, the products can be sold by any Cisco-authorized direct or reseller channel. This 90-day window is a one-time event. Any future enhancements to the product set will be immediately available to all customers.

For more information about SITA, visit:

http://www.sita.int/index.html

Current Airline Topologies

• AX.25 standard for IBM networks
• EMTOX standard for IBM networks
• AX.25 standard for Unisys

AX.25 Standard for IBM Networks

This diagram shows the standard topology for an airline network running AX.25 to an IBM mainframe.

TPADs connect to ASCU. Typically, ASCUs are on a leased line connected to the TPAD in SITA’s point of presence (POP). The TPAD is typically a Westinghouse or Northrop Grumman model and has been around for many years. The TPAD is a concentration box with 12 to 16 ALC ports, and, depending on the size of the remote site, there could be up to 500 ALC lines; all serial. Upstream ,they use X.25 to talk to the host PAD (HPAD).

On the host side, the HPADs communicate with the mainframe using a protocol called AX.25. AX.25 is almost identical to X.25. It is only for permanent virtual circuits (PVCs), and the major difference between X.25 and AX.25 is that AX.25 uses a nonstandard frame size of 240 bytes. (X.25 usually uses 128, 256, 512, 1024, all powers of two.)

In between the host and remote sites, SITA is limited to using X.25 over the network. Many airlines would like to migrate from X.25 to services like Frame Relay, ATM, and leased lines.

Migration to a common IP backbone increases network flexibility and service potential. For example, in this diagram, if a SITA TPAD is replaced with a router, then SDLC could be run to the ASCUs and data-link switching (DLSw) over a common IP backbone. Alternately, attaching routers to standard Ethernet interfaces allows the airlines to run normal IP routing. Airlines are motivated primarily to move from X.25 routing to other WAN technologies due to the significant increase in services. Upgrading feasibly could accommodate multiprotocol routing on a single network.

EMTOX Standard for IBM Networks

SITA also uses the Extended Mixed Traffic over X.25 (EMTOX) feature. Currently, EMTOX is the only option other than AX.25 to access the host in an SITA network. The major difference between the two features is that EMTOX uses SVC while AX.25 uses PVC.

EMTOX uses standard X.25 to transport data from the TPADs to the mainframe, usually via a FEP running NCP Packet Switching Interface (NPSI). The virtual circuit extends from the TPAD directly to the host without the HPAD required to tunnel the protocol. Data is transported through an X.25 switch to the FEP with load-balancing serial links. The FEP then passes the information to the Transaction Processing Facility (TPF) operating system.

ALC runs off the TPADs since ALC is almost always run with an IBM mainframe. (ALC is actually an IBM protocol called 1006 that predates SNA.)

This TPF system offers high throughput of transactions, which explains why some airlines are reluctant to migrate away from their current network.

1. TPF and MVS operating systems are incompatible and cannot be run simultaneously.

AX.25 Standard for Unisys

About 75 percent of airline systems use the IBM mainframe networks previously described. However, the remaining 25 percent predominantly use Unisys host reservation systems.

The Unisys network topology is very similar to the examples presented earlier.

The HPAD connections are the same as the IBM standards, except that they hook to a Unisys mainframe rather than an IBM mainframe. More important is that there is a different serial protocol running between the ASCUs and the TPADs. Unisys does not use ALC. Instead, it uses a protocol called Unisys Terminal System (UTS), also known as P1024C. (In contrast, ALC is SITA’s of P1024B).

Otherwise, the WAN functions the same as the TPF/MVS examples.

The additional HPAD pictured in this diagram represents a former SITA requirement to have redundancy.

Cisco's Three-Phase Delivery Plan

Phase I: ALC to AX.25/EMTOX

Phase I involves the use of ALC, Cisco's implementation of IBM-based P1024B. This protocol networking provides support for tunneling ALC (P1024B) traffic on properly configured Cisco 2500, 3600, 4500, and 4700 series router serial interfaces. ALC support is added to the Cisco IOS software as part of the Plus feature package. ALC tunneling allows customers to transport the six-bit traffic from ASCUs at remote locations across a TCP/IP network to central-site routers, where the traffic is converted to AX.25 or EMTOX and passed to the TPF host.

The first physical step in Phase I will be to replace the TPAD with a router, typically a Cisco 2523 or 2524. This consolidates a TCP/IP backbone in the middle so the customer has the option to run standards like Frame Relay or X.25.

Next, HPADs will be replaced with routers. These routers will run either AX.25 or EMTOX so they can communicate with the IBM mainframe.

The resulting changes will support ALC to the ASCUs, provide a transport in the middle, and provide a transport via the TCP/IP backbone, and then forward the data over the X.25 links using either AX.25 or EMTOX to communicate via NPSI on the FEP.

Phase I became generally available to customers in the second quarter of 1998.

• The LANAS depicted in the slide represents SITA's network management, which is SNMP based.

Phase II: Unisys Support

Phase II addresses the 25 percent of the airline market that uses Unisys mainframes. Although UTS is a more complicated protocol, Cisco provides support for tunneling P1024C protocol traffic on properly configured Cisco 2500, 3600, 4500, and 4700 series router serial interfaces. P1024C tunneling allows enterprises to transport this traffic from ASCUs at remote locations across a TCP/IP network to CPE routers. There traffic is converted to AX.25 or EMTOX and passed to the Unisys host.

The connections from the routers to the mainframe will be almost identical to the previous AX.25/EMTOX example.

Phase II is scheduled to be committed to Cisco IOS Release 12.0(1)T.

Phase: III

The goal of Phase III is to eliminate the need for a FEP. This is accomplished by using a channel-attached router with CIP.

The network stays the same from end-users to the remote CPE. You can have either ALC (P1024B) or UTS (P1024C) supported to the ASCUs.

However, there is a TCP endpoint where there used to be a TPAD, allowing the TCP endpoint to run all the way to the mainframe via a channel-attached router and CIP; no FEP is required.

Currently a TPF system can communicate TCP/IP only two ways:. with an IBM 3172 (not a popular choice) or with a CIP running CIP TCP offload. Because there is no native TCP stack in TPF, Cisco adds a TCP/IP stack on the CIP and uses the IP offload feature. The CIP also adds extremely high throughput to this environment.

This goal has been incorporated into a proposal for standardization called MATIP. MATIP is the strategic migration that the airlines want to move toward.

Cisco is beginning to investigate Phase III.

Mapping of Airline Traffic over IP

From the airlines or CRS perspective, those already implementing IP backbones are typically interested in MATIP, because it moves ALC over native IP. This means that their network is standards-based all the way from the end client to the host application.

TYPE A/B traffic refers to Layer 3 of ALC. Type A is interactive (terminal-to-host); Type B is host-to-host. Most remote applications will be Type A, because they’re carrying interactive terminal traffic. Some MATIP participants are applications developers (e.g., at IBM), so they need to address both Type A and B.

With MATIP, the key point is that mapping occurs at the remote site and at the host application only. Even the central site router(s) does not perceive MATIP. From the router’s perspective, it is just feeding standard IP packets into the host channel or Unisys D channel packet-based service (DCP).

When looking at the remote CPE side of the network, imagine a ticket agent standing in front of the green screen, entering data on one of the terminals connected to the controller (the ASCU). The ASCU uses ALC over a serial connection to the Cisco router. If we’re not using MATIP, the Cisco router terminates the ALC and creates a tunnel across the cloud to the Cisco data center router.

With the MATIP alternative, we still terminate ALC in the router and provide local acknowledgment as we do with the tunneling method. However, rather than using a Cisco-specific tunneling format, the router encapsulates the data in an IP packet using the MATIP header format, which defines the type of traffic, addresses, etc. The MATIP header is essentially a multiplexing scheme.

Starting with the remote CPE, the MATIP header will be read by a new application interface on the host side. (MATIP does require a small amount of application modification on the host.) P1024 B/C work the same way for the purposes of this description. The arrow on the right side of the slide indicates an ASCU session between the ASCU and the router. If the user resends too soon, then Cisco's sequencing number mechanism will filter out the redundant request.

Through the WAN, the Cisco router will use a single Frame Relay or ATM or X.25 PVC to multiplex the various TCP/IP sessions, whether they are ALC or any other protocol like DLSw, IPX, etc. Frames look like an IP data packet from the router’s point of view.

At the central site, the router receives and forwards IP packets directly to the host. There will be a MATIP layer in the host that will recognize the MATIP header. The central site router strips off the TCP header before the host sees it. If there’s a TCP/IP layer on the host, the stripping can occur on the host. If there’s no TCP/IP support on the host, the customer can use Cisco's TCP offload feature; the stripping then occurs on the router, and the Common Link Access for Workstations (CLAW) protocols are used to communicate with the host.

Platforms

The main issue is that ALC is a 6-bit protocol and requires special firmware.

Remote CPE (ALC-attached Router)

These require Cirrus Serial Communications Controller

• Cisco 2500 with Cirrus interfaces (Cobra)
• Cisco 3600 with Cirrus WIC card
• Cisco 4500 with Cirrus interfaces
• Cisco 2600 with Cirrus WIC (brand new platform, not currently tested)

1. Cisco 3600 has more processing power, which is important with ALC. Processing power is important when considering ALC because the firmware support will receive the 6-bit frame, and it will not do the frame building. The processor will do the packet building, but it eats up processing power.
2. ALC is being put on only certain Cisco remote access routers. P1024B is a 6-bit protocol that requires specific hardware to support its unique timing requirements for local acknowledgment. The Cisco 2520-23 and appropriately configured versions of the Cisco 3600 and Cisco 45xx/47xx are currently the only Cisco platforms that support it.

Central CPE (AX.25/EMTOX Router)

1. This is any IOS router with X.25 support.
• Cisco 2500
• Cisco 3600
• Cisco 4500/4700 (most popular choice)
• Cisco 7200
• Cisco 7500
2. Cisco 2500 does not have the horsepower to support a large number of remote sites.

Notice that the requirements are less stringent for the central CPE. There is no ALC local acknowledgment driver required at the central CPE, so any Cisco router can be used. Cisco intends to add ALC support to other Cisco platforms, which will be announced as they become available.

Images:

Basically this is the IBM feature set. (In 11.3T, it is everything named with an "s".)

• c2500-is-1, c2500-ds-1, c2500-js-1
• c4500-is-mz, c4500-ds-mz, c4500-js-mz

1. c2500-is-l is IP plus IBM with no desktop support. c2500-ds-l supports IP plus IBM with desktop support like IPX. This can be important to some airline customers.
2. When running Cisco IOS Release 11.3, if you run out of Flash memory, the image does not fit in 8 {Use either "MB" or "megabytes" }of Flash memory. This is a potential problem. To accommodate this shortage, SITA combines two 8 MB of Flash to form one 16-MB bank. This is a functional solution but does not allow backup image.

Supported Releases

Cisco IOS Release 11.3T and Release 12.0

Airline Control Protocol

ALC support is added to the Cisco IOS software as part of the Plus feature package. ALC is commonly called P1024B. ALC tunneling allows customers to transport the 6-bit traffic from ASCUs at remote locations across a TCP/IP network to central-site routers, where the traffic is converted to AX.25 or EMTOX and passed to the TPF host. Currently, only the master mode is supported, and Cisco is always primary.

Characteristics of ALC

• ALC is a 6-bit protocol and has no parity bits. It does have Cyclical Check Character (CCC) which is one byte for error detection.
• ALC is a character-oriented protocol. There are no 0x7E flags to easily demarcate the frames, as there are with most SDLC or HDLC frames. Instead, SYN1 or SYN2 precede each frame to designate the beginning of an ALC frame.
• ALC uses IPARS to represent data instead of ASCII or EBCDIC.
• ALC is a polled protocol with multidrop (like SDLC or Bisync).
• Unreliable data-link control. With ALC, data-link control frame drops are considered acceptable and, in some instances, desirable. This process prevents frames from getting stuck in a router for more than 3 to 4 seconds. With Type A interactive traffic, retransmission occurs when the user presses Enter again. This prevents the same frame from being sent to the mainframe if someone presses Enter twice. There are no sequence numbers for retransmission.
• Service messages are messages sent from the remote router to the ASCUs and to inform what is happening on the network. These service messages include network down, congestion, please retry (or reenter).

This diagram shows examples of ALC frames. You can read the format using the following key:

S1: SYN1 character (0x00).

S2: SYN2 character (0x20).

GA: Go ahead. Character code (0x30) that denotes the type of control message.

IA: Interchange address. Address of the ASCU to which the output control message is sent.

TA: Terminal address. Address of the device (terminal or printer) to which the output message is sent.

CMD: Command. Character denoting the type of write command.

LA: Line address. Control character indicating where the message must be displayed. The control character may be no-op, a new line character, or a specific line address.

EOM(c): End of message (complete).

NIA: Next interchange address. Character code (0x04) indicating that the next ASCU should be polled.

CCC: Cyclic cycle check.

ALPS IOS Architecture

An ALC frame enters the remote CPE router from an ASCU via the Cirrus firmware and the Cirrus driver interface. The frame is next passed to either ALC (P1024B) or UTS (P1024C) protocol handling and through to the circuit manger code. It is the peer manager will make the TCP connection, thus creating a virtual circuit which is multiplexed across a TCP peer circuit. (This virtual circuit functions similarly to a DLSw connection.) The result binds TCP together with the ALC code.

Once the frame leaves the peer manager, it is drawn through TCP and then across the WAN. The frame transverses the WAN to the central CPE. This is the X.25 attached router that will lead to the host.

At the central CPE, the frame begins a migration path similar to the remote CPE, from the Wan to TCP/IP, then to the TCP driver and through to the peer manger code. It is at peer manager code that interfaces X.25 and TCP. This circuit manager has a function similar to that between ALC and TCP, the interface is slightly different. The frame forwards it out through X.25 to go to the FEP.

Encapsulation services control frame headers using ALPS Tunneling Protocol (ATP). ATP works much like DLSw’s Switch-to-Switch protocol. In TCP, everything is byte-oriented, so there has to be a way to demarcate what frames are, headers to designate control frames to tell type and for passing information between sides.

Configuration Example

The purpose of this section is to provide an example of how ALPS can be configured to run both AX.25 and EMTOX on a single network.

ALC Router (or Remote CPE)

hostname alps_ccpe
!
alps local-peer 200.100.40.2 promiscuous
alps enable-alarms circuit
!
interface Loopback0
 ip address 200.100.40.2 255.255.255.0
!
interface Ethernet0
 ip address 172.18.3.27 255.255.255.0
!
interface Serial0
ip address 200.100.50.3 255.255.255.0
encapsulation frame-relay IETF
clockrate 56000
frame-relay map ip 200.100.50.2 20
frame-relay intf-type dce
!
interface Serial2
no ip address
encapsulation x25 dce ax25
x25 ltc 500
alps host-hld 4040 host-link 2 ax25
clockrate 64000
!
interface Serial3
no ip address
encapsulation x25 dce
x25 accept-reverse
alps host-hld 5050 host-link 6 emtox 2222
alps translate 161* 200.100.25.2
clockrate 64000
!
ip route 200.100.25.0 255.255.255.0 200.100.50.2
snmp-server community public RO

hostname alps_bk_ccpe
!
alps local-peer 200.100.70.2 promiscuous
alps enable-alarms peer
!
interface Loopback0
 ip address 200.100.40.2 255.255.255.0
!
interface Ethernet0
 ip address 172.18.3.27 255.255.255.0
 media-type 10BaseT
!
interface Serial0
ip address 200.100.50.3 255.255.255.0
encapsulation frame-relay IETF
clockrate 56000
frame-relay map ip 200.100.50.2 20
frame-relay intf-type dce
!
interface Serial2
no ip address
encapsulation x25 dce ax25
x25 ltc 500
alps host-hld 4040 host-link 2 ax25
clockrate 64000
!
ip route 200.100.25.0 255.255.255.0 200.100.50.2
snmp-server community public RO
!

Questions

1. What are P1024B and P1024C and how do they differ?
2. What is the primary difference between AX.25 and EMTOX?
3. What is another name for Cisco's Airline Control protocol? Why does it require special firmware?
4. True or False? In MATIP, mapping occurs at the primary CPE and remote CPE routers.
5. Describe the potential benefits of Cisco's ALPS plan.

Answers

1. Both are airline protocols run between ASCUs and TPADs (or remote CPEs). However, P1024B is used in IBM mainframe networks, and P1024C is used exclusively in Unisys mainframe networks.
2. EMTOX uses standard X.25 to transport data over switched virtual circuits (SVCs). Although similar to EMTOX, AX.25 uses only permanent virtual circuits (PVCs).
3. ALC is Cisco's implementation of P1024B. It requires Cirrus interfaces because it is an irregular 6-bit protocol.
4. False. In MATIP, mapping occurs at the remote site and the host application only. The routers do not distinguish these from standard IP packets.