| Phil Karn. The Qualcomm CDMA digital cellular system. In Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pages 35--39. USENIX Association, August 1993. |
....time sensitive traffic (video, audio) must be delivered within a certain time limit beyond which the data is obsolete. Exhaustive retransmissions in this case merely succeed in wasting time in order to deliver data that will be discarded once it arrives at its destination. CDMA IS95 RLP [CDMA, Karn93] limits the latency on the wireless link by retransmitting a frame only a couple of times. This decreases the residual frame error rate significantly, but does not provide fully reliable link service. This behavior is also possible in GSM RLP [RLP] by properly configuring the maximum number of ....
Karn, P., "The Qualcomm CDMA Digital Cellular System" Proc. USENIX Mobile and Location-Independent Computing Symposium, USENIX Association, August 1993.
....CC systems provide wide area coverage at low bit rates, while WLAN systems support higher speeds at low error rates but within smaller areas. WLANs depict losses of up to 1. 5 for Ethernet size frames [3] while CC systems suffer from losses of 1 2 for their much shorter (voice optimized) frames [4]. The effects of these losses are dramatic: a 2 packet loss rate over a single WLAN link reduces TCP throughput by half [5] The performance of real UDP applications (as opposed to UDP based benchmarks) over wireless links has largely been ignored, due to the diversity of UDP applications. ....
....Splitting TCP connections is also incompatible with IP security which encrypts TCP headers [7] The alternative to transport layer modifications is local recovery at the link layer. The Radio Link Protocols (RLPs) provided by CC systems provide error control customized for the underlying link [4]. Since they only apply a single scheme though, they can be inappropriate for some traffic. For example, retransmissions delay real time traffic and may interfere with TCP recovery [8] One way to avoid 3 PHY LL IP TCP UDP Wireless Host A PHY LL IP Base Station A 10Mbps 1ms PHY LL ....
[Article contains additional citation context not shown here]
P. Karn, The Qualcomm CDMA digital cellular system, in: Proc. of the USENIX Mobile and Location-Independent Computing Symposium, 1993, pp. 35--39. 21
....least) encapsulate IP datagrams into link frames, thus isolating higher layers from low level details. Minimalistic link layers however may be insufficient for wireless links. In voice telephony, random frame losses of 1 2 are considered reasonable as they do not cause audible speech degradation [1]. Since physical layer errors are usually clustered, randomization is achieved by interleaving and coding across several frames. Most Internet applications are not error tolerant though, thus wireless losses impose additional error recovery requirements. The traditional Internet approach is to ....
....Frames may carry either encoded voice or higher layer data. Compared to WLANs, CC systems exhibit higher delays due to the lower bit rates and longer distances involved. The outdoor CC environment is also harsher, with multipath fading caused by buildings and hills. Frame loss rates of 1 2 [1] are not detrimental to voice quality as long as they seem random. This is achieved by bit interleaving, which considerably increases processing delay. CC systems are interconnected to other networks using an Interworking Function (IWF) 5] The IWF provides digital to analog conversions to ....
[Article contains additional citation context not shown here]
P. Karn, "The Qualcomm CDMA digital cellular system," Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, Aug. 1993, pp. 35-39.
....error rates. Personal Communications Systems (PCS) the evolution of CT, will provide higher bit rates using smaller coverage cells. WLANs suffer from losses of up to 1. 5 for Ethernet size packets [2] while CT systems suffer from losses of 1 2 for their much shorter (voice oriented) frames [3]. The effects of such losses are dramatic: a 2 packet loss rate over a single WLAN link reduces TCP throughput by half [4] The performance of UDP applications over wireless links has largely been ignored, mainly due to the diversity of UDP applications. Considerable work has been devoted to TCP ....
....Split TCP connections are also incompatible with IP security which encrypts TCP headers [6] The main alternative to transport layer solutions is local error recovery at the link layer. The Radio Link Protocols (RLPs) provided by CT systems offer error control customized for the underlying link [3]. Since they always apply the same scheme though, they may not be appropriate for some traffic. For example, retransmissions may delay real time traffic or interfere with TCP recovery [7] One way to avoid such adverse interactions is to exploit transport layer information at the link layer. By ....
[Article contains additional citation context not shown here]
Karn, P.: The Qualcomm CDMA digital cellular system. Proc. of the USENIX Mobile and Location-Independent Computing Symposium (1993) 35--39
....to hide wireless losses from TCP by retransmitting lost packets over the wireless link using an automatic repeat request (ARQ) protocol. Examples of commercial systems that use the link layer approach are CDPD, which uses a LAPD derived protocol called MDLP [3] and the CDMA wireless system [4]. The main advantage of employing a link layer protocol for recovery of wireless losses is that its implementation is confined to the data link layer at the base stations and mobile hosts, so that it fits naturally into the layered structure of TCP IP without requiring changes to TCP. The main ....
P. Karn, "The Qualcomm CDMA Digital Cellular System", Proc. 1993 USENIX Symp. on Mobile and Location-Independent Computing, pp. 3540, Aug. 1993.
....value in the absence of losses, for local area transfers [1] The corresponding figure for wide area transfers is only 23 . Even worse, a cellular system that suffers a 2 loss over its short, speech optimized, frames, would suffer IP packet loss of 64 , thus reducing TCP throughput to nearly zero [2]. We believe that these problems present to the networking community both a challenge and an opportunity. The challenge is to overcome the limitations of existing protocols when employed over wireless media. The opportunity is to prohibit similar media dependencies from causing trouble in the ....
....recovery at the link layer at the expense of greatly varying delays [7] These delays however can cause TCP to timeout and trigger congestion recovery end to end. Another approach tries to avoid such adverse interactions by only offering limited recovery, leaving complete recovery to higher layers [2], if it is required. Although this scheme works well with TCP, it suffers from the same problem as the traditional RLP approaches: it is inflexible since it only offers a single type of service, regardless of higher layer protocol and or application needs. III. The Case for a Flexible, Adaptive ....
P. Karn, "The Qualcomm CDMA digital cellular system," in Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, August 1993, pp. 35--39.
....introduced by the wireless portion of the link. One method of hiding these errors is by using link level retransmissions. Another method is to use some form of error correction such as forward error correction (FEC) These techniques are currently used in digital cellular systems such as CDMA [31] andTDMA. Another protocol that uses these techniques is the AIRMAIL protocol [32] Although the layered protocol approach would suggest that no cooperation from higher level protocols should be necessary to make this scheme work well, it has been shown that knowledge of the retransmission timeout ....
P. Karn, "The Qualcomm CDMA digital cellular system," in Proceedings of the USENIX Mobile and Location-Independent Computing Symposium: August 2--3, 1993, Cambridge, Massachusetts, USA (USENIX Association, ed.), (Berkeley, CA, USA), pp. 35--39, USENIX, Aug. 1993.
....With this aim, RLP provides efficient but partial recovery to reduce the losses seen by TCP. However, complete recovery by RLP would increase the variability of the round trip delay seen by TCP, and consequently reduce TCP throughput [1,7] This philosophy of partial RLP recovery is explained in [5]. The performance of Protocol T can be improved by suitable changes to RLP that simultaneously reduce the probability of RLP aborts as well as the variability of RLP delays. Although, RLP is quite efficient, it is possible that a partial recovery version of SRL is better suited underneath TCP. ....
P. Karn, The qualcomm CDMA digital cellular system, ftp://ftp. qualcomm.com/pub/cdma.
....increase due to collisions. Measurements of TCP performance over a single hop path (a WLAN link with 2 packet loss) showed that TCP throughput dropped to only 47 of its value in the absence of losses [4] CT links on the North American CDMA system exhibit 1 2 error rates over its 172 bit frames [5]. TCP IP datagrams must be segmented into multiple such frames, causing IP datagram error rate to increase dramatically and TCP performance to drop accordingly. CT systems use interleaving techniques to randomize frame losses and avoid degraded voice quality. Random frame losses cause more IP ....
....use retransmissions to provide full recovery, at the expense of variable delays. These may cause TCP to timeout in the process and trigger end to end loss recovery anyway [9] In order to avoid such adverse interactions limited recovery may be provided, leaving further recovery to higher layers [5]. Another variation is to provide transport protocol specific optimizations at the link layer [10] By exploiting transport layer information the link layer can avoid adverse interactions and economize on control overhead. This coupling however limits the applicability of such solutions to ....
[Article contains additional citation context not shown here]
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. of the USENIX Mobile and Location-Independent Computing Symposium , August 1993, pp. 35--39.
....framing schemes offer reasonable performance, which implies that end to end performance is limited by the worst link on the path. For voice systems, with their small frames, error rates of 1 2 are a reasonable design goal as they do not cause audible speech degradation when distributed uniformly [2]. Since physical layer bit errors are usually clustered, bit interleaving and coding across several link layer frames is used to randomize errors. For data applications where loss is not acceptable however, the lowest common denominator approach of IP necessitates additional error recovery. ....
....to the lack of link contention these are more predictable though. The outdoor cellular environment is harsher than that of indoor WLANs, with interference and multipath fading caused by buildings and hills. For optimal voice performance, CT systems use short frames that suffer from losses of 1 2 [2]. As long as the physical layer manages to randomize these losses, audible voice quality degradation is avoided. For connections terminating outside the CT system, an Interworking Function (IWF) is provided to interface with other networks [9] Communication with analog telephones or modems is ....
[Article contains additional citation context not shown here]
P. Karn. The Qualcomm CDMA digital cellular system. In Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pages 35--39, August 1993.
No context found.
Phil Karn. The Qualcomm CDMA digital cellular system. In Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pages 35--39. USENIX Association, August 1993.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. of the USENIX Mobile and Location-Independent Computing Symposium, 1993, pp. 35--39.
No context found.
P. Karn. The Qualcomm CDMA digital cellular system. In Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pages 35--39, August 1993.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. of the USENIX Mobile and Location-Independent Computing Symposium, 1993, pp. 35--39.
No context found.
P. Karn. The Qualcomm CDMA digital cellular system. In Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pages 35--39, August 1993.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. of the USENIX Mobile and Location-Independent Computing Symposium, 1993, pp. 35--39.
No context found.
P. Karn, "The Qualcomm CDMA Digital Cellular System," Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pp. 35-39, 1993.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. of the USENIX Mobile and Location-Independent Computing Symposium, 1993, pp. 35--39.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. of the USENIX Mobile and Location-Independent Computing Symposium, 1993, pp. 35--39.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, August 1993, pp. 35--39.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. 1993.
No context found.
P. Karn, "The Qualcomm CDMA digital cellular system," in Proc. of the USENIX Mobile and Location-Independent Computing Symposium, 1993, pp. 35--39.
No context found.
Karn, P., "The Qualcomm CDMA Digital Cellular System" Proc. USENIX Mobile and Location-Independent Computing Symposium, USENIX Association, August 1993.
No context found.
P. Karn. The Qualcomm CDMA digital cellular system. In Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pages 35--39, August 1993.
No context found.
P. Karn. The Qualcomm CDMA digital cellular system. In Proceedings of the USENIX Mobile and Location-Independent Computing Symposium, pages 35--39, August 1993.
First 50 documents
Online articles have much greater impact More about CiteSeer.IST Add search form to your site Submit documents Feedback
CiteSeer.IST - Copyright Penn State and NEC