In the 1970s and 1980s, IBM mainframes were so dominant that the comment "no one ever got fired for buying IBM" became a cliché. Frame Relay now appears to have a similar cachet — the service is low cost, almost ubiquitous in the United States and reliable. Also, contrary to general perception, Frame Relay is expandable well beyond T1 speeds and, in fact, has no specific bandwidth limit (e.g., Verizon offers speeds up to 44 Mbps). So any organization considering a WAN deployment should include Frame Relay as a priority option.
Why Frame Relay rather than traditional circuits (e.g., T1s or ISDN)? Frame Relay costs less for the same throughput because it more efficiently uses bandwidth. As the successor to the hoary X.25 standard, [1] Frame Relay allows multiple customers to share the bandwidth of a physical connection by taking advantage of the bursty nature of data transmissions (bandwidth on demand). It supports applications such as host-to-host/LAN-to-LAN links, telecommuting, multiple user Internet access, PBX-to-PBX communications, and passable voice/video communications.
The cost for Frame Relay service usually includes three elements:
-
PVC (private virtual circuit), which is usually related to the CIR (committed information rate)
-
Port charges
-
Access to the premises
It would seem that with only three major cost elements, comparing service offerings would be straightforward. Unfortunately, there are a number of factors that complicate the analysis. Following are key factors to consider.
Port Size, CIR, and Discard Eligible Flag
A rough rule of thumb that some network designers use is to set the CIR at half the port size (e.g., a PVC with a port size of 512 kb might have a CIR of 256 kbps). A better approach is to understand the bandwidth requirements of the organization's users and applications and set port size and CIR at optimum levels.
Assume, for example, a Portland field office is connected to the New York headquarters building. Portland has low bandwidth requirements but needs to be able to connect at any time (and not be subject to bottlenecks during busy times of the day). Portland might have a port speed of 128 kbps and a CIR of 64 kbps. In addition, there are six other field offices that transmit to headquarters, with the same specifications. The headquarters port speed is set at 256 kbps, with a CIR of 128 kbps. Clearly, headquarters is seriously oversubscribed. That is, if all sites transmit at once, headquarters will not be able to handle the volume. If the business environment is such that the network designer knows all six will not be transmitting at once, this can be a practical way to minimize costs.
If the network designer also knows that users in field offices can tolerate some transmission delay, further savings can be obtained by reducing the CIR, maybe even down to zero. At zero CIR, all packets are marked as "discard eligible" and are marked for a later transmission.
Asymmetric PVCs
Some carriers, such as AT&T, allow PVCs to be configured with CIRs (committed information rates) that are not equal in both directions. For example, assume a firm's corporate office is in Knoxville, Tennessee, and one of its field offices is in Houston, Texas. Data transmission from Houston to Knoxville may require a CIR of 64 kbps, whereas Knoxville to Houston may only require a 16-kbps CIR. If the carrier permits asymmetric PVCs, they should be considered because many times traffic is unequal between sites. Because the CIR is one factor driving Frame Relay charges, use of this technique can drive down costs with no decrease in service levels to the organization. Many WANs using Frame Relay have been implemented without fine-tuning for unequal traffic.
Multi-Carrier Networks
Many Frame Relay networks are single vendor from the IXC (interexchange carrier) POP to the destination. The local access link may be provided by the LEC, but the Frame Relay network itself is all one vendor. An alternative and more economical solution is to use a LEC Frame Relay network to concentrate traffic to a hub within an intraLATA area, and then transmit to major sites using IXC Frame Relay facilities. The critical factor is the access link. There are two disadvantages to this approach: (1) additional time is required to negotiate and manage separate vendors, and (2) some network management information is lost when Frame Relay packets cross vendor boundaries.
Exhibit 1 illustrates the multi-carrier approach. This solution only makes sense if the organization's topology fits the scenario — smaller locations in relatively close proximity to a hub location (within an intraLATA boundary). The alternative to this approach is to connect each site directly to the IXC POP.
PVC versus SVC
Initially, carriers set up Frame Relay circuits with dedicated, permanent virtual circuits (PVCs) that required an always-up access circuit to the POP. However, switched virtual circuits (SVCs) are now available for organizations that need (1) less frequent access to the network, or (2) more dynamic connection requirements. An SVC is started by the user, then the data is sent and the connection is torn down as in a traditional telephone call. SVCs are less expensive than PVCs up to a point (similar to traditional dial-up per-minute charges versus a dedicated circuit). Aside from lower transmission costs for limited duration sessions, SVCs offer other potential benefits:
-
Reduced equipment costs (FRADs [2] and router serial ports) relative to a complete PVC implementation, particularly as the network grows in a highly meshed configuration.
-
Inexpensive disaster recovery capability. Ongoing backup PVC costs are not incurred and regular database updates for backups can be scheduled as appropriate.
-
Temporary, any-to-any connections. These limited-duration links eliminate the need for PVCs between sites that only occasionally communicate with each other.
-
Simplified administration. Preconfiguring and managing PVC changes are time-consuming. For highly meshed networks, SVCs can reduce network configuration maintenance.
The above advantages are contingent on the availability of SVCs from the carrier and on user requirements. Also, at certain volumes of traffic, SVCs are no longer economical — sites should be periodically reviewed for appropriate technology. Unfortunately, many carriers do not offer SVCs.
Frame Relay over DSL
Increasingly, CLECs are offering Frame Relay via a DSL link (FRoDSL). Combined with the increased ability of providers to monitor commercial DSL and provide service-level guarantees, this option can provide significantly lower access costs.
Voice Communications Networking
Voicemail
Voicemail, which became widespread in the 1980s, was originally considered a substitute for a live person at the other end of the line. More recently, however, a shift in usage toward intentional messaging has occurred. Where there is no need for dialogue, voice messages can be recorded and sent quickly to an individual extension or distribution list.
Most major voicemail vendors have long provided the ability to transfer voicemail messages from one location to another over dedicated lines or the PSTN (public switched telephone network). For example, Avaya's Audix system can forward messages to another Audix server or to a different vendor's voicemail system using the AMIS (Audio Messaging Interchange Specification).
More recently, a new standard called VPIM (Voice Profile for Internet Messaging) has been developed, which allows voice messages to be packetized and sent over IP networks (or the public Internet). Most major voicemail vendors, including Avaya, Nortel, Siemens, and others, are implementing this standard into their voice messaging products. VPIM provides both economic and functional benefits:
-
Conserves bandwidth. The message is packetized and compressed to one half its original size.
-
Simplifies distribution. As more voicemail systems become VPIM compatible, distribution to multiple locations is easier.
-
Improves efficiency of message broadcast. The older AMIS system sent messages one at a time, even if many users at a distant location were receiving the same message. VPIM sends a single message, which is then addressed to multiple recipients, resulting in both a quicker and more efficient (i.e., less bandwidth) transmission.
-
Integrates easily with unified messaging. Sending and receiving voicemail messages in VPIM format is more straightforward, because the transmission is treated as a special, multimedia e-mail.
Exhibit 2 illustrates the use of VPIM for voicemail message transmission.
Virtual Private Network (VPN)
The term "virtual private network" has become closely linked with substitution of an IP-based public network (usually the Internet) for dedicated or leased facilities. Instead of leasing a T1 or Frame Relay circuit to link office A to a distant office B, an encrypted "tunnel" can be established across the Internet to securely transport data packets. Originally, carriers such as AT&T used the concept of VPN (called SDN by AT&T) to describe a logical private network for each customer using the service. The term "virtual" was used because the actual hardware, software, and circuits are shared among all the carrier's customers, but the end customer perceives the service as a dedicated facility.
VPNs reduce long-distance communications costs — particularly for international sites — by eliminating much of the IXC expense. However, there are start-up and maintenance charges that can make a VPN implementation uneconomical for certain volumes of traffic. Also, VPNs that use the public Internet are subject to the vagaries of events on the Net — congestion, irregular quality of service, etc.
Exhibit 3 shows a typical VPN configuration. The example shown is for data communications only. Although voice over the public Internet may yet have its day, currently the quality of service (QoS) on the Internet is not adequate for most enterprises. Voice-over-IP, using private transmission facilities with guaranteed QoS, is discussed in another section of this chapter.
Generally, most medium to large organizations that have multiple, dispersed sites can use VPN technology to supplement (rarely to completely eliminate) their existing wide area networks. The likelihood of a good fit increases dramatically if the organization incurs a large dialup (800 number) bill, typically associated with a RAS (remote access service) implementation. Indeed, organizations such as PricewaterhouseCoopers, having thousands of professionals on the road, have saved hundreds of thousands of dollars annually by sharply reducing long-distance dialup minutes.
When considering implementation of a VPN, there are a number of financial, business, and security issues to consider:
-
Advantages:
-
Replace some dedicated lines, such as T1s, with transmission over the Internet (e.g., backup T1s could be eliminated). The organization must be aware of the caveats, such as the potential for Internet congestion and poor quality of service.
-
Eliminate some or most RAS dial-up charges. While ISPs may charge a per-hour charge for users tunneling through a VPN, those charges are significantly less than IXC per-minute charges. For example, a large organization might negotiate a $1-per-hour ISP connect time charge, whereas the same charge for an hour of toll-free dial-up could be $5.00.
-
Enable quick bandwidth increases by adding additional ports (compared to lead-times of two to eight weeks for new T1/T3 services).
-
Facilitate extranets for customers, suppliers, and partners, and provide additional E-commerce functions.
-
Make secure intranets available to field offices around the world (at a reasonable cost).
-
Provide high-speed services to telecommuters who have broadband access in the home/small office. For example, VPNs can operate over cable modem lines or DSL. With this capability, some jobs can be accomplished off site that might otherwise require office space/equipment.
-
Reduce management costs of a WAN by using a fully integrated, secure VPN solution, in contrast to the traditional plethora of network access gear.
-
Reduce the number of access lines for some field offices. If the office has a separate line for Internet access and data communications (e.g., for Frame Relay), VPN can eliminate one access line.
-
EDI (electronic data interchange) communications costs can be reduced by establishing an extranet using a VPN and eliminating use of a value-added network (VAN).
-
-
Disadvantages/concerns:
-
VPNs are more complex to manage. Some organizations outsource the management of the VPN network.
-
VPN is not always the answer. For example, a small network with low bandwidth requirements may be better served via a Frame Relay solution (less expensive edge equipment, less maintenance).
-
The public Internet occasionally suffers congestion. Although this may someday change with the introduction of MLPS, [3] for the moment it is a significant concern for organizations that must have extremely high uptime. Some vendors offer fail-over capabilities that allow traffic to be sent over an alternative link (e.g., dial-up ISDN) if the Internet is congested.
-
The level of available VPN encryption, while certainly adequate for any domestic U.S. commercial needs, may not be available for some international traffic due to government restrictions. However, this may be changing, at least for some countries. France, for example, has long required that encryption be no stronger than that afforded by a 40-bit key. Recently, the maximum permitted length has been increased to 128 bits, a considerable increase in security levels.
-
