With the large scale rollout of wireless Ethernet-based VoIP phones in the enterprise, VoIP technology has reached a level of maturity where residential VoIP cordless phones, wireless PBXs and, eventually, future cellular systems based on 802.11 are no longer just a possibility for VoIP over WLAN technology—they are inevitable.
The larger challenge of VoIP over WLANs will be how to handle handoffs of an active call between 802.11 APs or between an 802.11 AP and a cellular network's cell. VoIP residential cordless phones will be the only application that will not require handoff capabilities initially.
The following diagram illustrates the basic mobility challenge for WLAN implementations:
As the diagram illustrates, the hierarchical nature of IP network topology results in two types of mobility:
- Micro-Mobility
- Macro-Mobility
Micro-mobility is the simplest form of mobility. The subscriber is moving within a single domain, such as an enterprise, a set of hotspots owned by company A or some other sort of limited WLAN configuration. Micro-mobility essentially involves intra-domain handoffs. There is no need for external coordination. Issues of timing, call control and handoff control can be set (or bounded) by network design. The first wave of VoIP over WLAN services will be based on micro-mobility in the enterprise or in the residence.
Macro-mobility involves moving between two domains that fall under the administration of completely distinct organizations. For example, one hotspot could be run by carrier A and a second is administered by carrier B. The two domains must collaborate to complete the handoff and to conduct authentication, authorization and accounting (AAA) activities between the domains. These arrangements are similar to efforts in the cellular industry that have been developed over the last few years.
Given that micro-mobility solutions will be the first developed and deployed, micro-mobility solutions must consider the larger framework that includes macro-mobility capabilities as well as the eventual evolution to full macro-mobility.
There are two principal approaches for supporting mobility in VoIP services. They are:
- SIP (Session Initiated Protocol)
- Mobile IP
The following section gives a brief overview of Mobile IP and SIP for mobility applications and concludes with a discussion of cellular GPRS/WLAN integration for data services. Many in the industry anticipate that cellular data deployments like GPRS/WLAN will be implemented initially and a larger movement to full blown VoIP WLANs will follow later.
Mobile IP Overview
Several elements are needed to implement a WLAN incorporating Mobile IP capabilities for APto- AP handoffs. The following diagram illustrates the elements of a Mobile-IP system:
The elements shown above are defined as follows:
- Mobile Node: VoIP caller
- RFA: Regional Foreign Agent
- AAAF: Authentication, Authorization, Accounting - Foreign Network
- GFA: Gateway Foreign Agent
- HA: Home Agent
- AAAH: Authentication, Authorization, Accounting - Home Network
- CN: Corresponding Node (the network that a caller node is attached to)
- Caller Node: Another phone caller
When both caller and the mobile node are in the home network (i.e. when the switching occurs within one's own voice network), the PBX function is present. In this case, the VoIP call is routed through a self-contained enterprise network and no Internet domain resources are required.
Mobile IP is based on the concept that a mobile node has a home address associated with a home network. Each time the mobile node connects to a foreign network, it obtains a temporary address which is known as a Care of Address (CoA). The CoA is valid while the mobile node is attached to the foreign network domain. It is deleted or purged from the foreign network once the mobile node leaves the domain.
In Mobile IP WLANs, there are two mobility agents, Home Agents (HA) and Foreign Agents (FA), that coordinate, update and authorize the connections and associated CoAs for clients from foreign networks. When a call is set up between a Caller Node and the Mobile Node, a binding update message is sent by the Home Agent to the Corresponding Node. The binding message allows VoIP traffic and messages to be directly tunneled between the Caller Node and the Mobile Node. Messages need not be routed to the home network. Clearly, tunneling greatly increases the efficiency of Mobile IP.
Difficulty arises when roaming occurs between two foreign networks while a call is in progress between a Caller Node and a Mobile Node. Consider a Mobile Node that is leaving one foreign network and transitioning to a new foreign network. The first foreign network and associated FA must send binding warning messages to the caller's Home Agent (HA), alerting the HA that packets from the CN are arriving, but the Mobile Node is no longer in the first foreign network. The first foreign network redirects any received packets back to the caller's HA. Simultaneously, the Mobile Node will complete the association with the second foreign network. The HA then will establish the new CoA with the second Foreign Agent / foreign network. The HA then provides a new CoA to the second foreign network, and this allows direct tunneling of messages between the Mobile Node on the second foreign network and the caller node in the corresponding network. While this handoff is taking place, the HA acts as a middle man, maintaining the VoIP connection by forwarding packets from the first foreign network to the second foreign network. This stops once the new CoA and binding for direct tunneling are in place.
A complete description of this process is beyond the scope of this paper. However, it is obvious that when the problem is reduced to a single domain (such as an enterprise WLAN where there are no foreign networks), the solution is much more tractable.
In addition to the overall handoff process, controlling several timing issues is imperative for successful macro-mobility and micro-mobility handoffs. The following are some of the timing elements that must concern WLAN or handset designers:
- Ts: The period of time needed for a station to associate with an access point (probe and associate)
- Tf: The period of time needed by a handset to associate with a foreign network (inter-domain update)
- Th: The period of time to bind to a foreign network and create a new CoA
- Tmc: The period of time needed to send packets directly between the Mobile Node and a Caller Node
- Tno: The period of time needed to bind update messages from an old foreign network to a new foreign network
For macro-mobility (inter-domain) handoffs between two different foreign networks, Mobile IP has the following timing:
During a handoff, as the new tunnel connection is established between a Mobile Node and the Caller Node, a series of packets will be disrupted and may arrive out of order, causing them to be discarded. The period of time for this disruption is given by the same formula for both SIP and Mobil IP. It is:
For micro-mobility handoffs in the same domain, such as a handoff from one AP to another in an enterprise WLAN, Mobile IP has the following timing:
At this point, issues relating to intra-domain handoffs will be discussed. In the next section, the SIP approach and its associated timing issues will be described.
| Issues for Mobile IP Macro-mobility There are two major issues relating to the implementation of Mobile IP macro-mobility:
A more important issue is the fact that Mobile IP is not widely supported. The end-to-end deployment of Mobile IP on the Internet will take significant effort to achieve. |
SIP Overview
As an alternative to Mobile IP, SIP supports IP mobility for VoIP WLAN applications by providing handoff capabilities at the application layer.
SIP can make direct use of Dynamic Host Control Protocol (DHCP) when connecting to an 802.11 AP for binding an IP address. A number of proposed systems use DIAMETER as the AAA (authentication, authorization, accounting) protocol. SIP makes use of the concept of a visited registrar (VR) in the foreign network. The SIP VR combines some of the functions of a SIP proxy server, location server and user agent. The SIP proxy server concept allows SIP to handle both firewall functions and network address translations (NAT), which are pervasive in home network topologies. SIP was initially designed to support roaming (i.e. moving into a domain while the connection is disabled and then establishing service) so that a user could be found independently of location and network device. For example, with SIP a call on a handheld phone could be transferred to a computer SIP phone. SIP is being modified to support mobility as well as roaming applications.
Like Mobile IP, macro-mobility in an SIP implementation would be based on the concept of foreign networks and home networks. With SIP, the foreign agent of Mobile IP is replaced by an SIP VR in the foreign network. The Mobile IP home agent (HA) is replaced by an SIP home registrar (HR). The SIP HR is a combination of an SIP proxy server, a location server and a user agent server. The following diagram illustrates an SIP network:
The elements in this type of network are defined as follows:
- Mobile Node: VoIP caller
- DHCP: Dynamic Host Control Protocol
- AAAF: Authentication, Authorization, Accounting - Foreign Network
- SIP VR: Visited Registrar
- SIP HR: Home Registrar
- AAAH: Authentication, Authorization, Accounting - Home Network
- CN: Corresponding Node (the network that a caller node is attached to)
- Caller Node: Another phone caller
One of the principal differences between Mobile IP and SIP is the use of DHCP by 802.11 APs. DHCP doubles the number of transactions needed to associate with an access point. It also requires that the client perform an ARP (Address Resolution Protocol) to detect duplicate addresses in the sub-net. The advantage of DHCP is that no modification to the local network is needed.
While there are minor differences between Mobile IP and SIP, the handoff procedures are essentially identical. SIP has the advantage of using the existing IP network without modification. However, this comes at the expense of delays that are typically double those of Mobile IP in a macro-mobility environment.
To its advantage, SIP is fully supported today by the Windows environment (i.e. Windows XP), making possible a rapid deployment in the residential/SOHO marketplace.
Just as with Mobile IP, timing issues must be addressed if SIP handoffs are to be supported. For the most part, SIP's timing elements are identical to those for Mobile IP. The only exception is SIP's use of the Address Resolution Protocol (ARP). In the formulas below, Tarp is defined as the period of time needed for an ARP exchange.
The time required to register and set up a VoIP call with SIP is the following:
(addition of 2Ts + Tarp vs. Mobile IP)
For macro-mobility (inter-domain) handoffs between two different foreign networks, SIP has the following timing:
The blackout time for SIP and Mobile IP is given by:
For micro-mobility (intra-domain) handoffs such as an enterprise-based AP-to-AP handoff in the same domain, Mobile IP has the following timing:
Tmip_intra = 4Ts + Tarp + 2Tf
(addition of 2Ts + Tarp vs. Mobile IP)
As these formulas indicate, by using the existing IP network without modification, SIP suffers from delays that can be 2x those of Mobile IP.
| SIP for Residential and SOHO Use SIP can be used on an existing netwok without modification. SIP is designed into Windows XP and will be in Windows CE SIP suffers from delays that can be 2x those of Mobile IP. In a larger network, these delays quickly become unacceptable. For cordless phone VoIP applications in the home, the delay in SIP is negligible and its ease-of-integration will greatly facilitate product introduction. |
IPv6 and Protocol Improvement
These discussions of Mobile IP and SIP have assumed WLAN deployments on IPv4 networks. Both SIP and Mobile IP would greatly benefit from the pervasive use of IPv6. In both cases this would allow direct addressing of a mobile node client. If Mobile IP and SIP were revised in light of IPv6, the Foreign Agent could be removed entirely and handoff times would improve.
For Mobile IP to succeed, it must be deployed pervasively. As a result, there is interest in the industry in SIP for consumer applications.
Wide Area Network Integration: WLAN and GPRS Inter-working
While the ultimate goal is to provide seamless IP mobility for all applications including voice, in the near term cellular carriers are planning to integrate data operations through a combination of 802.11 for hotspots and cellular telephony technology for wide area data networking.
This section briefly describes the integration of the cellular GPRS (General Packet Radio System) standard with WLAN technology in a seamless data network. This process will take several steps, including:
- Common billing and customer care but no inter-working of WLAN and GPRS networks.
- A 3GPP-based access control and charging system where all WLAN AAA (authentication, authorization, accounting) will be based on GPRS AAA procedures.
- Access to GPRS data services such as WAP are supported on the WLAN system, but there are no handoffs between WLAN and GPRS.
- Where jitter and time delay permit, there would be service continuity for the services described in item three above. These services would be provided across the WLAN and GPRS networks. The handoff of IP multimedia may not be supported, but other IP services would be.
- Seamless continuity where all services are supported transparently between WLAN and GPRS networks. There is no noticeable difference in the services.
- Access to 3GPP switched circuit services is provided and voice services are supported.
- Tightly coupled WLAN network
- Loosely coupled WLAN network
A loosely coupled network is illustrated in the following diagram. This loosely coupled scheme would be based on Mobile IP. Only minor modifications to installed WLAN networks would be required. However, cellular operators would need to install AAA servers for billing mediation and to support WLANs as well as to support interoperations between the GPRS network and WLANs. In a loosely coupled system, the Internet is used as the traffic backbone. Because service operators would not have complete control over the network, there is some concern that consistent quality might not be provided.
| Category | Tight Coupling | Loose Coupling |
| Authentication | GPRS authentication and cipher key encryption | SIM based authentication/Optional Radius based |
| Accounting | Reuse GPRS accounting | External Billing for common accounting |
| WLAN Cellular Mobility | SSGN Call Anchor, mobility by iner-SSGN handovers | Home Agent (HA) is the call anchor, mobile IP between access router and GGSN (gateway) |
| Context Transfer | Fine grain info on QoS, flows, etc. | limited information between GGSN and WLAN (IETF working on "seamboy") |
| System / Network Engineering | Impact on WLAN traffic to existing GSN bearer and signaling is an issue | WLAN and GPRS can be designed spearately |
| New Development | WLAN modification for GPRS for GPRS signaling, possible SGSN modes. | CAG for SIM-based authentication, Billing mediator for accounting |
| Standards | A new SSN interface | EAP-Sim and EAP-AKA authentication (IETF Ppext working group) |
| Target Usage | Applies to Cellular owned WLAN or affliated WISP | Broad Application |
Table 13
The deployment of WLAN hotspots has taken on a life of its own. It may not be practical to require conformance to certain standards or to modify existing WLAN equipment with cellular security and signaling. The motivation of carriers is clear. Carriers require control over network quality. They also are concerned that Mobile IP has not been widely deployed and, because of this, seamless interfaces may be delayed. Carriers also are concerned that when QoS capabilities are deployed, they may be haphazard at best. Clearly, all indications from the marketplace would suggest that the deployment of seamless networks of WLAN and cellular technology will happen in the near future. Indeed, certain cellular carriers already have acquired and are supporting WLAN hotspot networks, merging the billing and customer care operations for the two technologies.
