Accordingly, the mobile device is free to switch in the office's wired local area network and wireless wide area network, and during this process, the established connection will continue without interrupts, of course, this type of mobile needs to have mobile IP addresses (Mobile IP [RFC2002]).
1.2 Conversion of Wireless Connections (Assumptions About the Radio Link)
Most of the system structure mentioned above is that it is not sufficient to reflect the characteristics of wireless connections based on a wireless connection (may also be more than one radio level). There is also the following considerations:
- What is the lack of wireless media? Because of burst errors and disconnections, the wireless connection is higher than the wired network. The following techniques typically do not list all types of errors, corresponding, perfect solutions should integrate the essence of all recommendations. However, in this article, we will pay more attention to most typical cases and provide solutions, these cases are:
(1) Higher 'Dislike Rate' (BER) due to random errors. These random errors mean link layers error correction and relay will result in longer and uncertain delays.
(2) Due to system conversion (HANDOFF) or disconnected service interruption.
The latter is equally important, in this survey, we also include corresponding recommendations.
- Wireless service is based on a dataginary or based on virtual circuit? At present, the virtual circuit switch is also used to be more common, but the information package network has begun to appear, for example, Metricom's Starmode [CB96], CDPD [CDPD], integrated information wireless service, the GPRS (General Packet Radio Service) and GSM [Bw97].
- What reliability can be provided in wireless connection? The feature of the wireless service is that it is relays the packet (FRAME) until it is received by the destination. Of course, this feature can be turned off. For example, GSM allows you to close RLP [RLP] protocol. METRICOM has a similar "lightweight" mode. In the RLP of GSM, the frame can be relay before reaching the maximum number of relay (protocol parameters). When this limit is reached, what happens is fully determined by the telecom operator: or disconnects the physical connection, or re-resetting the sequence digital resynchronization, and due to the transfer and receiving buffers, it will also cause data lost. Some wireless services, such as CDMA IS95-RLP [CDMA, Karn93], in order to limit the latency of wireless connection, the specified only twice in the frame twice. This approach effectively reduces the residual misalignment rate, but does not provide a completely reliable connection service.
Montenegro, et al. Informational [Page 6]
Mobile devices can transmit and receive data at the same time? To do this, this will definitely increase the electronic cost of the mobile device. Of course, this is not the main reason. In this article, we assume that the mobile device performs transmission and reception of data.
- Can mobile devices address multiple points on wireless connections? The path walking through the information packages sent to different locations is separate, and therefore, the path connected to each point also exhibits different characteristics. In fact, it is very common, at any given time location, mobile devices can only be connected to one point (intermediate node). Of course, with the development of technology, these issues can be solved after the Channel-State Dependent Packet Scheduling application (see "Packet Scheduling" section later). 2 should it be a IP protocol? (SHOULD IT BE IP or Not?)
In the first conclusion, use IP protocols as the foundation network protocol. In particular, some of the data protocols developed by wireless phones are not always (may sometimes) at the upper part of the IP [MOWGLI, WAP] layer, which are derived from the concept of providing adaptation services between wireless and wired portions.
For mobile devices, communication is always reasonable through proxy. However, we expect many wireless mobile devices to utilize wired networks as much as possible when conditions allow. This mode is close to the use of laptop: The device can use a local area network (LANS), and after the office, you can only take a dial-up mode.
For these mobile devices, the IP architecture is the best way because this mode is required to communicate with the intermediate node (referred to as 10m network to the W-LAN or office).
2.1 Potential Network Error Characteristics (Underlying NetWork Error Characteristics)
Use IP-based network protocols require robust underlying connection support, which is also the case where wireless connection is concerned.
IP protocol, other protocols of IP packets, from head to tail, is guaranteed by checksums, this mechanism has been behind, and in some cases are optional (see [Stevens94, Paxson97) ]). For most Internet applications, the checksum mechanisms can basically meet the requirements; and in the wireless environment, the RAW mode is more fragile than other ends. Therefore, for wireless connectivity, it is unimaginable to detect or correct transmission errors like these channels. In this regard, it should be completed by a local connection mechanism.
Montenegro, et al. Informational [Page 7]
In addition, the damaged IP package will still spread on the network until the host arrives at the target. For example, an intermediate router requires a verification of the IP packet, but regardless of UDP or TCP checksum. Correspondingly, when the payload in the IP package is damaged until the package arrives at the final destination will be discovered.
In order to improve the characteristics of the wireless connection, it is to provide more reliable services, a preferred method is to use the link-layer mechanism, such as FEC, relay, and so on. This method has been used by CDPD, RicoChet, and CDMA.
This method is generally similar to successful point-to-point protocol (PPP), a robust group frame and 16-bit check, can replace the Serial Line Protocol, SLIP on Wired Networks (Serial Line Interface Protocol, SLIP) ), Using single-by-segment frame, no verification.
[AGS98] It is recommended to use FEC in a satellite environment. Note that the connecting layer can adjust its size to prevent the dislocation rate (BER), which can decompose and assemble the package itself, so that IP can still accommodate a large MTU (maximum transmission unit) size [LS98].
The main problem with IP is that its head is too large. At present, some basic connection layer protocols such as PPP [RFC1661], located above the IP layer, can allow the head to compress [IPHC, IPHC-RTP, IPHC-PPP], There is a great extent alleviated such problems.
2.2 Non-IP Selection (non-ip alternatives)
For wireless environments, many non-IP mode selection is proposed, and below is a representative advice:
2.2.1 WAP
Wireless Application Protocol, WAP specifies network protocol [WAP] of the application architecture and wireless device (mobile phone, pager, PDA). The system needs to make a proxy between mobile devices and servers. The WAP protocol stack layer is on the data report transmission service layer. Most wireless networks provide this datagram, such as: IS-136, GSM SMS / USSD, UDP (such as CDPD) and GSM GPRS in IP networks.
Montenegro, et al. Informational [Page 8]
The core of the WAP protocol is a binary HTTP / 1.1 protocol with additional characteristics. Additional features refer to buffering headers between requests, sharing status between customers and servers, and more.
2.2.2 Configuring Non IP Selection (Deploying Non-IP Alternatives)
IP can be said to be a deep-rooted, non-IP mode to be extended, and only because they are not based on IP architectures. Any non-IP mode must provide a gateway service to implement the corresponding IP address and non-IP address, and must terminate the security of the IP level at the gateway, and the Gateway cannot adopt an IP-based protocol (such as a dynamic host configuration protocol -Dynamic Host Configuration Protocol, Domain Name Server - Domain Name Services, Light Directory Access Protocol - LightWeight Directory Access Protocol, Service Location Protocol - Service Location Protocol, etc.
More complicated situations are that the device also supports wireless operations and wired operations. If the device uses IP to perform wireless operation, when switching to a wired network, the operation process does not interrupt - using mobile IP or may be implemented. But if the non-IP mode is used, this conversion is more difficult to implement.
Non-IP mode faces such a responsibility, that is, it must provide evidence to prove that IP cannot adapt to the wireless environment, which determines that IP is not feasible in the wireless environment.
2.3 IP-based considerations (IP-BASED CONSIDERATIONS)
Look at the application of IP globally, it is clear that it can become the basic protocol of network technology. Implementation of optimization techniques makes traditional Internet application protocols, including new protocols, including IP or UDP, thus benefited.
2.3.1 Select Maximum Transmission Unit (Choosing The MTU) [Stevens94, RFC1144]
In a slow network, the time used in the transmission allowable range is quite large. The interactive response time should not exceed the human factor limit - 100 to 200 milliseconds. This time should be considered in the maximum time budget:
(1) Send a package; (2) get a response. In the implementation of most network stacks, (1) is higher depending on the maximum transmission unit (MTU). In the worst case, in the interactive application, a small bag may wait for a large package of a batch transfer application before being sent. Therefore, a better rule is to select the MTU to see if its transmission time is less than 200 milliseconds (if it is just a little bit).
Montenegro, et al. Informational [Page 9]
Of course, the compression technology and service type queuing (interactive packets are given higher priority) can alleviate this issue. In particular, the latter can reduce the waiting time, which saves 1/2 MTU transmission time.
2.3.2 MTU channel inventions [RFC1191]
The MTU channel is invented to make any IP-based agreement benefited. It allows the sender to determine the maximum end-to-end transport unit to the given purpose. Without MTU channel inventions, the default IPv4 MTU size is 576. The benefits of using greater MTU are:
- Small header data overhead rate (Smaller Ratio of Header Overhead to Data)
- When the unit of the segment is increased, the TCP allows TCP to increase the window.
Of course, for a given dislocation rate (BER), the larger the MTU, the greater the probability of the corresponding specified segment error. However, if you use a more underlying technique, such as FEC and link layers, BER is possible to decrease. Due to additional relays, delays have become a problem, in practice, the package transfer time with large MTU is relatively long.
Recommendation: It is recommended to use channel MTU inventions. [AGS98] It is recommended to apply it to the satellite environment.
2.3.3 Non-TCP Protocol (Non-TCP Proposals)
Other protocols assume that IP datagram is based on the IP layer [NetBLT], or in the UDP layer [MNCP] on the UDP layer, an optimized transmission method is implemented. It is also possible to achieve the way to rely on TCP protocols, and related research and development has also accumulated many valuable experience. However, the Internet system did not collapse. In contrast, because of its main protocol (TCP) is very cautious when using the network, all problems caused by network blocking and other conditions are as black, that is, opaque, thereby avoiding further Clogging.
However, the wiring caused by the wireless medium, due to high-level error rate (BER) or signal attenuation, can cause packet loss. The optional protocol is:
- NetBLT [NetBLT, RFC1986, RFC1030]
- MNCP [mncp]
Montenegro, et al. Informational [Page 10]