MMS in AFTN System Applications

The Multi-protocol Message System (MMS), offered by Digital Resources Inc. (DRI) for AFTN system applications, is a general purpose store-and-forward message switching system that maximizes the advantages of modern network techniques. The MMS is completely parameterized, and can implement networks ranging from the obsolete point-to-point configuration, up through the most up-to-date wide area network (WAN) applications. It can be implemented for various types of message switching applications, including the ICAO defined AFTN, AMHS, ATN, the ITU defined F.31 networks, WMO, etc. The MMS is designed to maximize the economic and hardware support advantages of a personal computer (PC) based WAN approach, over the point-to-point design. The MMS takes full advantage of continually decreasing server and PC costs along with newly emerging network technologies, while avoiding dependence on costly skilled technical support staff.

Except for the very smallest networks (less than 7 lines), a WAN implementation that uses remotely located distributed intelligence and operates on low cost Intel based servers, PCs, and routers, is certain to minimize both purchase costs and annual operating costs. The MMS can be used as either a single purpose AFTN-only WAN, or the base for a new organization-wide WAN that includes AFTN along with any other application. The MMS can also be implemented on any existing WAN. The MMS is far more than a simple gateway between an obsolete point-to-point network and an external WAN. Since the MMS AFTN system operates entirely on the WAN, it provides all of the advantages, such as high-speed transfers, maximum resiliency, and automatic error correction that cannot possibly be provided by a simple gateway to some external WAN.

Even if the MMS is initially installed as a simple point-to-point network, it can be upgraded at any time to the more advanced WAN technologies at a relatively low cost and impact, by simply adding WAN network routers. Since both the MMS software and the routers are multi-protocol, any combination of WAN technologies can be readily mixed on the same MMS system.

The MMS is fully compliant with all items in the latest version of ICAO Annex 10. Additionally, in applicable areas such as flight plan validity checking and AMHS, the MMS is also compliant with:

All components of the system operate under any version of the Microsoft Windows operating system from Windows 95 through Windows 2000, and Windows XP. The reason for choosing MS Windows over Unix or Linux alternatives is the result of the great importance that DRI places on the attribute of vendor independence. DRI believes this is also the reason that Windows has dominated all versions of Unix (including Linux) in recent years in the number of new installations. Unlike software, hardware is subject to wear and failures caused by age alone. Therefore, it is much more important to be independent from hardware vendors rather than from software vendors

All of the supported versions of Unix are proprietary to a specific hardware platform vendor. In the case of highly specialized applications, this means that any Unix based application must be 'ported' by its developer to run on more than one vendor's proprietary hardware platform. Because of the elimination of once dominant hardware vendors, such as Wang, Digital Equipment Corporation (DEC), Tandem, etc., software vendors are naturally reluctant to risk the expense of porting their Unix application to some other hardware platform. History has demonstrated that any Unix based proprietary hardware platform might very well disappear from the market in the near future. In contrast, there are literally hundreds of hardware vendors that offer MS Windows based systems.

Since the hardware platform vendors profit from the reduced competition that their hardware specific version of Unix provides, they are therefore able to sustain relatively high prices. These high prices are inevitable for the case where customers using specialized applications are permanently locked into the only vendor capable of maintaining or enhancing the application. Even worse, in cases where market forces eliminate the hardware vendor entirely (Wang, DEC, Tandem, etc.), the customers are left stranded with orphaned systems that are impossible to expand and eventually become impossible to even maintain.

For the above reasons, it should be apparent that the term "COTS" (commercial off-the-shelf) is often misleading. It is misused to provide reassurance to the buyer that is only an illusion. In many cases, it obscures the fact the product offered is highly proprietary and available only from a single vendor. The fact that 'Unix' is supposedly an 'industry standard' in no way mitigates the substantial risk that it locks the system buyer into a very expensive hardware system supplied and supported by only a single vendor. On any networked applications, where data can be readily interchanged between MS Windows platforms and Unix platforms via TCP/IP, there is no long term reason to become locked into Unix based proprietary hardware platforms.

Our partners:
BlackBerry Accessories - Blackberry Accessories store. Get the latest collection of Blackberry accessories including BlackBerry charger, BlackBerry cases, Blackberry Batteries, Bluetooth Headsets and more
Free Nokia Applications - Free Cell Phone Software Applications and Themes
Cheap International Mobiles Calls - Vopium allows you to save 40-90% when making international calls or sending SMS from your mobile phone in your home country. Get free 30 minutes and 100 SMS.
Calling Cards - Offers cheap calling cards, prepaid phone card, international calling card for Russia, India, China & other countries at affordable prices. Unlock Nokia - Unlock Your Nokia mobile was never this easy
Cell Phone Accessories - CellphoneAccents is the largest cell phone accessory store online, providing cell phone accessories for Nokia, Apple, Samsung, LG, Motorola & other major cellphone brands at the lowest prices.
Unlock Nokia - Unlock Your Nokia mobile to reach its full pontential!
International phone cards - on sale online 24-7, low rates, best prices on calling cards
iPhone unlocking - They provide a full unlock solution for all iPhone versions currently up to 1.1.4 (latest version) for use on any network.

Only 3 hardware components are used in the MMS: industry standard Intel-based servers and personal computers (PCs), wide area network (WAN) routers, and local area network (LAN) hubs. The server/PCs are used for message switching and routing at both the switching center and the remotely located front-end-processors (FEPs). The PCs also serve as end-user terminals for message origination and reception. The LAN hubs are used at the switching center and remote concentrator sites to connect together the switching server/PCs and the control, message correction, archiving, and monitoring terminals. The routers are also used at the switching center and at the remote concentrator sites.

At the switching center, the servers and PCs that are connected together on the LAN are either switching server PCs, or terminal PCs. The switching server PCs have no video monitor or keyboard and are connected to the routers. These server PCs perform the actual message switching and high level routing functions, through their connected routers. The routers perform the low-level routing across the WAN. The LAN based terminal PCs allow the user to monitor and control the network, and also to prepare new messages and correct rejected messages. Terminal PCs always include a keyboard, mouse, and video monitor.

All components are 'commercial-off-the shelf' (COTS) items, but more importantly, all components are non-proprietary. To insure long-term maintainability, competitive prices, and to avoid future conflicts with operating system upgrades, no vendor specific hardware or software components are used. This precludes the use of any add-in cards that require a specialized vendor-specific software driver. This still allows up to the 4 standard COM ports and 4 LPT ports on PCs, since they are directly supported by the operating system. It precludes any 'multi-port' COM port add-in cards or add-in router interface cards, that require add-in vendor-specific software drivers. These vendor-specific cards pose serious long-term support risks, due to future incompatibility problems between device drivers and operating system software and product cancellation issues.

At least 7 competing vendors offer each of the 3 components that comprise the MMS. The switching center is composed entirely of servers and PCs connected on a dual-homed LAN. The front-end-processors (FEPs) at the remote concentrator sites use PCs that employ neither a keyboard nor video monitor. Depending on the application, a FEP may not necessarily employ a hard disk unit. On very large installations, this stripped-down FEP configuration is intended to further reduce the PC cost and approach the reliability of the routers and hubs that contain no moving parts.

In addition to the standard vendor hardware test programs and third party diagnostics, DRI provides additional specialized test programs for the LAN based units, and specialized simultaneous 4 port loopback testing on the FEPs.

Our partners: Mobile Phone Deals, Bluetooth Advertising

Although the MMS can be implemented as a point-to-point AFTN switch, or any hybrid combination of WAN and point-to-point architecture, the major cost reduction and resilience benefits are obtained for those networks which are entirely WAN implementations. It is also important to note that the MMS WAN is separate from the ICAO proposed Aeronautical Telecommunications Network (ATN). The MMS WAN is, however, fully compatible with the ATN. The ICAO ATN defines links between states, but does not mandate the much more numerous links within the state. Thus, the MMS WAN operates on the national AFTN level, while linking to other states at either the AFTN, Cidin, or ATN level. In the case of ATN links to adjacent states, the connectivity is provided through the ICAO defined AMHS gateway unit. The discussion below begins with the MMS WAN at the national level and near the end of this section it deals with the AMHS gateway link to the ATN.

The wide area network (WAN) consists of T1 or E1 lines, operating between 9.6 KBPS to 2 MBPS. Any combination of X.25, TCP/IP, Frame Relay, DSL, ISDN, or VPN network protocols can be used for the MMS. Additionally, V.90 or V.92 analog modems can be used, as either back-up or primary route, for any or all remote user lines. On the initial installation, depending on the network configuration, any of 5 possible Cisco routers are used. However, any other vendor or router model can be used to expand the WAN. All of these routers are multi-protocol, and 2 of the 5 routers are modular and therefore easily expanded. The Simple Network Management Protocol (SNMP) is available, but not required, to monitor an X.25, TCP/IP, or Frame Relay implementation.

5.1 Partial-mesh Considerations

The number of telecommunication lines required for the WAN is based on the number of remote concentrator sites connected to the main switch. In order to take full advantage of resiliency designed into the MMS software, each concentrator site requires 2 lines. Since the 2 lines are typically connected to adjacent concentrator sites, they will normally be much shorter, and probably lower cost, than the alternative configuration of lines connected directly to the switching site in a star topology.

For example, a system consisting of a main switch and 3 remote concentrator sites requires only 3 lines if resiliency is not a requirement or if automated V.90 dial back-up is also implemented to provide resiliency. Assuming dial-backup is not implemented, then, for the elimination of service interruptions due to line problems, a total of 4 lines are required in order to provide the resilience of at least 2 paths to all sites. No matter how many concentrator sites are added, the cost-saving benefit of the resiliency provided by the single added line is more than justified. Adding yet another additional line increases the resilience and available bandwidth even further, although it is not essential. This method of adding lines, to provide 2 or more routes to any network point, is termed a 'partial-mesh' network.

The future ATN is an example of a partial-mesh network, at least at the inter-state level. Unfortunately, this ATN uses the costly to implement OSI protocols, making it impractical to extend to local AFTN users. Furthermore, the future ATN does not specify any partial-mesh requirements for the much more numerous intra-state links. Thus, at the local CAA level, the ATN permits the obsolete point-to-point links, which remain vulnerable to service interruptions. The overwhelming majority of AFTN system vendors can only offer a centralized switch with a gateway connection to adjacent states and point-to-point connections at the local level. Even though this limited gateway architecture satisfies the ATN requirements, it fails to offer the same resilience benefits that the future ATN offers on the inter-state links at the local level.

The primary reason why almost all AFTN system vendors cannot offer partial-mesh solutions at the local CAA level is due to the fact that the AFTN system software must be completely recreated to optimize the benefits of a partial-mesh WAN architecture. It is not possible to simply retrofit the old point-to-point AFTN software package to properly function in a totally new architecture. The software must necessarily be designed at the outset to interface with routers as part of a distributed network architecture. A secondary reason is that most AFTN system vendors have absolutely no experience in network design. The obsolete point-to-point installations never required any understanding of actual networks as such, and viewed the connections as nothing more than simple wires. As long as these vendors believe there is a reasonable chance of selling their obsolete point-to-point software, they will continue to avoid the very expensive and time-consuming development effort required to adapt to a partial-mesh WAN. Simply implementing an AMHS gateway to the ATN allows these vendors to create an illusion that their obsolete point-to-point software is actually a modern up-to-date network solution.

5.2 SVC, PVC and Bandwidth Considerations

The AFTN, F.31 or WMO application on the WAN use only a small fraction of the available router bandwidth and port connections. In the case of an X.25 application, this is achieved by delivering AFTN messages via switched virtual circuits (SVCs) 'calls', rather than permanently allocating bandwidth and ports, such as is the case in permanent virtual circuits (PVCs) or static SVCs. Once the queued traffic for that path is delivered, the 'call' is disconnected and the port becomes available for new incoming calls and the bandwidth becomes available to other non-AFTN applications. Thus at the time of initial installation, at least 80% of the ports and bandwidth of the routers are available for future applications, such as radar data, graphic weather maps, etc. These same SVCs are also critically important in providing resiliency outside of the WAN, since any remote station can be reached from the switching center via multiple servers, each connected to multiple routers.

5.3 AMHS and ATN Considerations

An AFTN system based on a WAN for national users can readily coexist with the eventual implementation of the Aeronautical Telecommunications Network (ATN). The ICAO ATN provides connectivity between states and provides for a gateway unit, designated as an AMHS gateway, to link the national AFTN WAN to the inter-state ATN. Both networks can function on the same national WAN, regardless of any difference in protocols.

The AFTN/AMHS gateway provides for the conversion of the rarely used OSI protocols, mandated by ICAO, to and from the much more widespread industry standard protocols, such as TCP/IP, Frame Relay, VPN, etc. Thus, this gateway approach allows the national AFTN network to continually evolve as newer functions and services become available via the widespread industry standard protocols. The seldom used static OSI protocols can be isolated to only those relatively few lines between states, without imposing constraints on future enhancements to the national AFTN system. For reasons cited below, the optimum CAA strategy is to implement a TCP/IP and X.25 wide-area-network to replace any existing point-to-point AFTN system.

Recommended Strategy: There are numerous reasons to simply install a replacement AFTN system based on a standard TCP/IP and X.25 network, and then defer all aspects of the implementation of ATN and AMHS until the final version of the specifications have been issued and actual extended operational trials have been successfully completed. Up until now, only very limited and simplified point-to-point trials have been completed. The major high risk factors in ATN and AMHS involve the dynamic routing of large numbers of unproven OSI based routers interconnected in a partial-mesh between regions (inter-domain) and within regions (intra-domain). Even In the best-case optimistic schedule for the first regional large scale ATN pilot operational testing, it will be 2006 before this critical high-risk period will even begin. Unlike the ICAO mandated OSI protocols in the ATN router, the universal industry standard TCP/IP protocol based commercial routers have had the benefit of 20 years of operational testing in tens of thousands of units in private WANs, and millions of units in the public Internet.

Thousands of significant bugs and specification interoperability issues have been corrected during this 20 year period of live TCP/IP operation. Thus, if ICAO actually persists in using the OSI protocols in the ATN, then it will be at least 15 years before comparable stable operation is achieved in the ATN. However, at each new ATN Transition Task Force meeting, it is becoming more obvious that the OSI protocols must ultimately be abandoned in favor of the TCP/IP protocols. The three factors driving this major protocol revision are: safety, economics, and deployment time. All three of these factors apply to both the ground-ground router and the air-ground router.

The safety issue results from attempting to introduce a new highly specialized niche-market product (ATN router) across international organizations with no prior background or defined trouble-shooting procedures or OSI routing protocol experience. In fact, in the case of AFTN, all routing was manual, static, and strictly pre-defined and limited, and thus the introduction of automatic dynamic routing will be a first-time experience for all involved CAA support staff members. There will be no proven products or tested procedures, such as the case with TCP/IP based routers, to fall back on when routing problems and undetected software bugs cause a collapse of the international ATN.

The economics issue arises from the fact that there are only 2 vendors now offering ATN routers to a very small niche market, at a cost 15 to 30 times as great as comparable commercial TCP/IP based routers. The original ICAO expectations, of mandating universal ATN use to distribute and offset the ATN router development cost across all applications, has already been eroded by the acceptance of interim and regional adaptations such as 'AFTN over X.25' and 'AMHS over X.25'. It was even further eroded in April 2002 by the partial introduction of TCP/IP into the ATN world. This new position accepts AMHS over a TCP/IP-only based network that is not at all compliant with the ATN specification. It also accepts 'tunneling' ATN traffic over IP routers, thereby substantially reducing the market for the specialized costly ATN router.

The new ICAO position also recognizes that airborne IP networks already exist, and will now be considered for air-ground applications. Commercial TCP/IP services are now available that allow up to 9 airborne passengers to simultaneously access the public Internet using a normal LAN, or even a wireless on-board LAN. In the face of these alternatives, and with fewer than 200 CAA potential customers, it is extremely unlikely that this ATN router is an economically viable long term product. This uncertain ATN router outlook raises the serious issue of long term vendor support questions.

The deployment time issue results from the fact that significant routing and interoperability problems and latent software bugs won't even become visible until after 2006. Even if it were possible to realistically simulate the hundreds of ATN routers from all of the different regions in a single room, it would require 5 years of shakedown testing to debug the software and routing configurations enough to establish even a minimal confidence level. In reality however, this shakedown testing must take place across many widely separated support organizations, including independent service providers, speaking many different languages. During this crucial period the support staff is undergoing its first crash course in troubleshooting wide-area-networks. Even if there are no language problems, the challenge of regional or world-wide cooperative troubleshooting on an unproven new product will present a significant challenge. Much of this extended shakedown period could be eliminated by simply starting out with the field proven commercial TCP/IP routers, with over 20 years of product development debugging periods behind them.

5.4 Protocol Translation

Within the national AFTN system, the industry standard commercial router software handles any required protocol translation between any non-OSI protocols, such as Frame Relay and TCP/IP and X.25, etc., while the MMS itself handles any addressing conversion required locally, such as Telnet or X.121 to and from the familiar 8 character ICAO addresses. The MMS remote user terminals can be either the X.400 AMHS units or the standard MMS intelligent terminals. Since the standard MMS terminals use the familiar ICAO message format and provide even more capabilities and resilience that the AMHS defined terminals, there is no benefit obtained in undertaking the steep learning curve required by the AMHS terminals.

5.5 Mobile User and Wireless Considerations

An example of the benefits of this strategy, of limiting the use of OSI protocols in favor of industry-standard protocols, is the MMS capability of connecting mobile AFTN users, and remote stations that cannot be directly cabled. In this case, PCs, notebooks, or hand-held PDAs can use wireless links to connect to the MMS WAN wireless access points at either the switching center or any remote concentrator. This link is handled via a combination of the wireless LAN protocols IEEE 802.11x and TCP/IP. For all wireless links security is provided by activating the built-in 128 bit industry-standard wired-equivalent-privacy (WEP) protocol.

The IEEE 802.11b protocol also provides a means to connect, to any remote concentrator or the switching center itself, any building complexes within a radius of 6 miles. This wireless connection is implemented by using very small directional antennas. Unlike the mobile wireless links above, this high-speed wireless link is between 2 fixed points. The one-time antenna installation and equipment cost is only a fraction of the annual cost of a leased line between the same two points.

5.6 VSAT Considerations

The MMS can provide for VSAT links between the switching center and the remote concentrators and/or between remote user stations and the switching center. As a result of the multi-protocol capability of the MMS software and routers, the MMS can even be used as a gateway or bridge between 2 different VSAT networks employing different protocols and satellite access methods. The combination of VSAT, with automated dial back-up, provides the absolute lowest cost method of providing highly resilient connectivity by minimizing or eliminating the much more costly leased land-lines.

The MMS also offers the possibility of eliminating any VSAT Frame Relay monthly service charge. Since the AFTN switch itself functions as a hub, the links can be implemented in a star topology as either TCP/IP PPP, Telnet, X.25 PVC or SVC, or simply V24. Wherever it is available, implementing the VSAT links in a star topology also makes it possible to avoid the more costly TDMA satellite access method in favor of the minimal cost MCPC/FDMA or SCPC/DAMA method. For those cases where a regional VSAT network is involved, the MMS makes it possible to use Frame Relay only on the regional VSAT link, while employing TCP/IP PPP on local AFTN non-regional VSAT links.

Our partners: Network Monitoring - Dotcom-Monitor's remote network monitoring services keep you informed about your network's status, availability and performance of network applications - that means your entire network functions at top performance and is more profitable.

Ultimately, all CAAs will implement organization-wide administrative WANs for data interchange. Equally certain is that all AFTN systems will be eventually implemented on a WAN. For obvious cost reasons, these 2 applications should share the same physical network. Once the AFTN network is implemented on a partial-mesh WAN, the infrastructure is already in place to extend its functionality to a general purpose administrative network for the entire CAA organization. The same routers and links used for AFTN can readily be adapted to TCP/IP based servers, LAN switches, and personal computers throughout the entire organization. Since TCP/IP is essentially the universal protocol, any Windows or non-Windows server, work station, or personal computer can be accommodated by the infrastructure.

Thus, the AFTN network becomes part of an organization wide intranet, that provides e-mail, document collaboration, file transfer, etc., services between staff members in different locations. As an intranet, it also can be used to provide local web-site hosting that all staff members can access with commercial browser programs, such as MS Internet Explorer or Netscape. Wherever it is necessary, the AFTN network can be kept functionally separate from the general purpose network by the installation of virtual-private-network (VPN) units that incorporate built-in firewalls. For those cases where the CAA has an existing WAN in operation, then the MMS AFTN system can be installed as a component of the existing WAN, with VPN/firewall units isolating the AFTN system from other non-AFTN network users.

The MMS can be expanded at any point in the future by simply adding industry standard servers and PCs and implementing the DRI software packages for the MMS component added. In addition to the servers and PCs at the switching center and FEPs, this also includes adding dial-up and/or directly connected message preparation and editing terminals running the DRI MMSTERM software package. Providing the use of the MMSTERM program license to the end-users makes it possible to implement an error-correcting Ack/Nak protocol, with WAN connectivity and/or automated dial-up access. This user terminal license applies throughout the entire network at no additional costs, regardless of how many users are added in the future.

Our partners:
NextPage® offers Risk Management Software that lets you manage your enterprise documents that are subject to corporate compliance and allows you and your employees to comply with your current document retention regulations.
Offshore Software Testing Services

Each of the server PCs comprising the switching center can handle 150,000 messages per day without introducing any significant message delay. Each single switching center can be expanded to 30 switching servers and PCs, for a maximum capacity of 4,500,000 messages per day. A typical switching center of 7 servers/PCs, routers, and hubs can be installed in a single rack. The expanded system of 30 server/PCs, and the associated hubs and routers, can be easily installed in 4 standard equipment racks. If the traffic load exceeds 4,500,000 messages per day, then additional switching centers can be implemented and linked to the initial switching center by either WAN links or LAN switches.

The switching center and remote concentrators automatically distribute the traffic load across all the server PCs. For reliability, at least 4 server PCs must be installed. Although an ordinary PC can readily function as a switching PC, the MMS normally uses Intel based servers from IBM, Compaq, Dell, Hewlett-Packard, etc., as the switching units. These servers include RAID level hot-swappable hard disks and error-checking-and-correction (ECC) RAM. If a server PC fails, the remaining PCs automatically pick up the load of the failed server PC without any operator intervention. If somehow 3 out of 4 server PCs failed simultaneously, the remaining single server PC automatically carries the entire traffic load. However, even a single remaining server PC is capable of handling at least 200,000 messages per day by gradually accumulating message queues (backlog). Although a single server PC handling message loads above 150,000 messages per day might eventually introduce message delays, the system continues to function in a somewhat degraded fashion. Normally, at least one of the 3 failed server PCs would be rebooted and returned to service long before any queue at all developed.

Thus, as long as at least one of the server PCs is still operating, the system will not crash. By connecting each of the PCs and both LAN hubs to separate small dedicated UPS units, each fed from separate AC power sources, a system crash is virtually impossible. This yields a system availability of 99.999 % ('five nines'). If somehow both LAN hubs failed simultaneously, the message traffic continues without interruption or delay, since each of the switch server PCs are autonomous and independent of any other PC. Each server PC is connected to at least 2 routers through serial ports in addition to its LAN connection.

Each node in the system is capable of holding up to 2,000 messages on its individual queue. For example, if there were 6 switching center server PCs and 12 FEPs handling 6 remote concentrators, then the combined distributed queues, between the switch PCs and the FEP PCs, can indefinitely hold at least 36,000 messages in the event of a prolonged major nationwide communication line outage. If an operator is present to put some of the blocked destination traffic on hold, then over a million messages could be retained indefinitely until the communication lines were restored. If automated dial-backup to the concentrators is implemented, then not even a total failure of the network itself can cause a service outage.

Although LAN hubs and routers crash only about once in 12 years, servers and PCs crash more frequently. This higher rate of PC failure is the result of intermittent faults due to moving parts and operating system flaws. However, in almost all cases, the server or PC can be quickly restored to service by simply rebooting it. Thus, a server PC needs to be replaced only for a hard recurring fault, which typically occurs about once every 7 years for a brand-name server employing hot-swappable RAID drives and ECC RAM. For an ordinary PC, a recurring hard fault may occur as often as once every 3 years. Therefore, in the typical case of 6 switching units sharing the traffic load, the approximate mean-time-between-failure (MTBF) is 46,000 years. If ordinary PCs are used as the 6 switching elements, this MTBF drops to 216 years.

Thus, in either MTBF case, a total system failure is virtually impossible, barring a catastrophe such as a flood, fire, or explosion. Even in the event of a physical catastrophe, it is still possible to maintain service without interruption if the option for a remotely located contingency switch is implemented. This option is described in a later section of this document.

Within the MMS network, there is no 'single-point-of-failure', such as is common on the obsolete dual-mode hot-standby architectures. This MMS redundancy applies not only to the communication paths between the routers, but also between FEPs and their associated end-user terminals or printers. All possible message flow paths for each route are continually used in a load-balanced manner. This is done in order to avoid the problem of an idle fall-back line that is out of service for weeks without detection, until it is actually needed.If any of the paths are blocked or develop high message queues, the system automatically shifts all the traffic to the remaining good paths without any need for operator intervention.

If all paths to a particular user station or FEP are blocked, then the system recirculates the traffic within the system, until at least one path is restored, or the operator establishes an alternate route. When the faulty path is restored to normal service, the system then automatically resumes distributing the traffic across both paths. Thus, at a level above the WAN, the MMS system automatically 'routes around' problems outside of the WAN, in the same manner as the routers themselves do for faults within the WAN.

	Line costs are reduced, with 4 to 7 digital WAN lines replacing 30 - 500 lower speed analog communication lines.

	Line costs are reduced for any station transferring less than 250 messages per day by implementing automated on-demand V.90 dial-up calls.

	Line costs are reduced by converting, on an ongoing basis, all or some lines to newer emerging low cost network technologies, such as Frame Relay, VSAT, ISDN, DSL, VPN, or V.90 dial-up.

	Elimination of 24/7 hardware maintenance staffing costs by added resiliency, and the total elimination of local component level repairs. All hardware maintenance is reduced to highest level unit replacement only. (See section 13 below).

	Reduction in operator staffing costs due to the elimination of communication line problems, which otherwise would require message correction or reentry by operators.

	Reduction in operator staffing costs due to automated alternate routing on hardware problems, and fully automated service-message handling and retrievals.

	Reduction in cost of on-site spare parts resulting from very low cost components from multiple competitive sources, and duplication of on-line units.

	Reduction in line costs and staffing by distributing costs over a combined AFTN and administration wide IT network that shares the same network infrastructure. (See section 6 above).