From eyh5@cornell.edu Mon Sep 3 16:27:01 2001 Return-Path: Received: from travelers.mail.cornell.edu (travelers.mail.cornell.edu [132.236.56.13]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f83KR0q01615 for ; Mon, 3 Sep 2001 16:27:00 -0400 (EDT) Received: from travelers.mail.cornell.edu (travelers.mail.cornell.edu [132.236.56.13]) by travelers.mail.cornell.edu (8.9.3/8.9.3) with SMTP id QAA07084 for ; Mon, 3 Sep 2001 16:26:59 -0400 (EDT) From: eyh5@cornell.edu Date: Mon, 3 Sep 2001 16:26:59 -0400 (EDT) X-Sender: eyh5@travelers.mail.cornell.edu To: egs@CS.Cornell.EDU Subject: 615 Paper # 1 Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII This paper, published in 1987, provides a detailed introduction to the packet radio network (PRNET), first developed in the 1970s by DARPA. The authors present the principles of the PRNET system, its mechanism of tracking continuously changing network topology by means of creating and maintaining several data tables, the routing protocols, and the requirements of the PRNET hardware capabilities. Today, PRNET is generally regarded as the harbinger of modern-day mobile ad-hoc networks (MANETs). According to the PRNET design concept, initially, a packet radio (PR) in a PRNET establishes potential packet-switching routes by proactively announcing its presence to all other PRs in the network before any packet transmission takes place. This advertisement is done in the form of Packet Radio Organization Packet (PROP). In the meantime, it receives PROPs from all other PR nodes in the network, thereby establishing its own perspective of the network topology. Such a proactive route-searching approach has the advantage of immediately transmitting a packet to a destination via a pre-selected route readily available. On the other hand, it requires a high volume of control information (i.e., PROP) to be exchanged amongst its peers, a result of which is the large consumption of network bandwidth. Therefore, the scope of such a network may be limited to a small scale. The idea of pacing during packet transmission is very similar to that of the Random Query Process Delay (RQPD) in ad-hoc networks, although the former applies to the data packets whereas the latter is concerned with the routing query control information. Also, the PRNET does address the issue of packet collision by implementing CSMA, a technique more commonly used in the bus arbitration-based Ethernet technology. It can be inferred that in designing the PRNET, the authors of the paper assume that network resources such as bandwidth are abundant and therefore are of a secondary concern (an assumption probably true in the 80s). This philosophy can be seen in their approach of transmitting packets to a destination PR. Here, the idea of multicasting a packet to several neighboring PR nodes is introduced, but only one (or more if the Alternate Routing feature is activated) PR will carry on with forwarding the packet, with the others simply discarding it. Such an approach may not be a practical solution in today's networking world due to the scarcity of the licensed bands. This philosophy also underscores the authors' primary concern of reliable, corruption-free transmission of data packets, at the expense of longer delay due to overhead processing. Each node of the PRNET has three separate tables containing information on its neighbors, their respective proximities (by hops) and the attached device(s). A future improvement can be made to consolidate these three separate tables into one. Doing so would allow reduced time by the node CPU to access relevant routing information, and would also free up some memory space in the node to be used elsewhere. In fact, in the subsequent research work following the publication of this paper, proposals have been made to efficiently use a single routing table in the mobile node. From wbell@CS.Cornell.EDU Mon Sep 3 23:03:04 2001 Return-Path: Received: from postoffice.mail.cornell.edu (postoffice.mail.cornell.edu [132.236.56.7]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84332q07303; Mon, 3 Sep 2001 23:03:02 -0400 (EDT) Received: from [192.168.1.100] (syr-66-24-16-64.twcny.rr.com [66.24.16.64]) by postoffice.mail.cornell.edu (8.9.3/8.9.3) with ESMTP id XAA24254; Mon, 3 Sep 2001 23:03:00 -0400 (EDT) Subject: 615 PAPER #1 From: Walter Bell To: egs@CS.Cornell.EDU Cc: wbell@CS.Cornell.EDU Content-Type: text/plain Content-Transfer-Encoding: 7bit X-Mailer: Evolution/0.12.99+cvs.2001.08.21.23.41 (Preview Release) Date: 03 Sep 2001 23:02:38 -0400 Message-Id: <999572583.2221.1.camel@brute> Mime-Version: 1.0 1) The DARPA Packet Radio Network Protocols This paper presents a high level overview of the evolution of the DARPA work for packet radio (PR) networks, where the goal was to use packet-switched, store-and-forward radio communications to provide reliable comminucations channel for mobile hosts. The goal was for automatic management and dynamic configuration of the network as well as dynamic routing, given a deployment of at least one packet radio within line-of-sight of another on the network. Routing and configuration is determined proactively on the network via PROP (Packet Radio Organization Packets) messages, which are sent out at a known interval, which summarizes each hosts' routing and connectivity information to each other using a distance vector approach. They present solutions for alternate routing in case of node failures as well as congestion and pacing protocols to ease network load and provide fairness, via simple protocols such as single-threading messages per node-node channel and non-strict fifo ordering on forwarded messages. The paper does a good job of the high level summary of how the network works, including simple and comprehensive examples of packets flowing through the network, which illustrates some of the nice optimizations that have been done to avoid active acknowledgment of all messages. The network has a solid design which works well for the small numbers of expected nodes [the designed limit was for 138 hosts], but fails to scale well because of the PROP message structure, which broadcasts each host's state of the world. There are some comments about varying the transmission energy to provide smaller neighborhoods for better scalability, but this seems to not work properly in a lot of cases [highly dense clusters of nodes wouldn't be able to find a transmission energy where an acceptable number of nodes was present.] This research was definitely groundbreaking at the time and most definitely affected and shaped a large body of future work, as future configuration and routing designs use this research as the comparison point. While this network organization would work on current hardware, this paper begs the question of if it's a reasonable assumption to have each node be aware of each other node in the network, much less get proactive routing information for each one. The scalability aspects of this system, as well as the energy consumption issues with a large amount of proactive overhead cast doubt on it's utility in the mobile, battery powered scenario. From viran@csl.cornell.edu Tue Sep 4 00:40:54 2001 Return-Path: Received: from cray.csl.cornell.edu (cray.csl.cornell.edu [132.236.71.70]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f844esq16453 for ; Tue, 4 Sep 2001 00:40:54 -0400 (EDT) Received: from localhost (viran@localhost) by cray.csl.cornell.edu (8.9.3/8.9.2) with ESMTP id AAA73330 for ; Tue, 4 Sep 2001 00:40:50 -0400 (EDT) (envelope-from viran@cray.csl.cornell.edu) X-Authentication-Warning: cray.csl.cornell.edu: viran owned process doing -bs Date: Tue, 4 Sep 2001 00:40:50 -0400 (EDT) From: "Virantha N. Ekanayake" To: Subject: 615 Paper 1 Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII The DARPA Packet Radio Network Protocols The paper describes an early ad-hoc wireless network based on packet radios (PRs). The PRs communicate via a common radio channel broadcast system (which is half-duplex at 400kbit/s or 100kbit/s), and autonomously figures out the network topology; information about changes to the network such as adding new PRs and PRs dropping out or moving around are propagated to every node. The combined number of PRs and devices in the network cannot exceed 138, and each PR can only handle 16 immediate neighbors. The routing algorithm is proactive or table-driven. It appears that each node or PR broadcasts its complete picture of the network in a PROP packet to its neighbors every 7.5 seconds. The PRs provides error control and hardware CRC, and alternate route forwarding after failed route transmissions. There's also dynamic delay estimation to time the broadcast and retransmission of packets, and an attempt at fairness by overriding the FIFO ordering of packet transmission with packets coming across weak radio-links. CSMA is used to avod collisions on the half-duplex channel. Some drawbacks in the DARPA PR network scheme include the need to keep a complete map of the network topology at every node and broadcast it periodically, which limits the network's scalability (as probably seen by the 138 device limit). These PROP packets need to be sent out regardless of whether there were any changes to the network topology, and are thus quite expensive. Also, in the paper, it's not clear how the PRs maintain a loop-free routing table in the place of updates. Another draw-back arises from the physical implementation: The half-duplex nature of the system limits its throughput as well as the fact that a PR can only receive from one radio; this leads to the triple delay timing used at a node to transmit packets. Say A to B to C is a path; then after A transmits, it needs to wait for B to transmit to C, and for C to acknowledge B before A can transmit another packet to B. Another drawback, although possibly unfair to direct at this paper considering its early roots, is the lack of security -- a rogue PR could capture packets and/or attempt to disrupt the network topology. It's also amusing to note that although the packet radios are touted as being low-cost, the actual cost is never mentioned. However, all in all, this was a pretty comprehensive description of an ad-hoc network over a radio broadcast link, and was probably one of the earliest descriptions of proactive routing. It's interesting to note that they implemented a special type of routing for speech transmissions (or other data types requiring low-latency and tolerant of packet losses). From teifel@csl.cornell.edu Tue Sep 4 06:31:58 2001 Return-Path: Received: from disney.csl.cornell.edu (disney.csl.cornell.edu [132.236.71.87]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84AVwq19463 for ; Tue, 4 Sep 2001 06:31:58 -0400 (EDT) Received: from localhost (teifel@localhost) by disney.csl.cornell.edu (8.11.3/8.9.2) with ESMTP id f84AVs442396 for ; Tue, 4 Sep 2001 06:31:54 -0400 (EDT) (envelope-from teifel@disney.csl.cornell.edu) X-Authentication-Warning: disney.csl.cornell.edu: teifel owned process doing -bs Date: Tue, 4 Sep 2001 06:31:54 -0400 (EDT) From: "John R. Teifel" To: Subject: 615 PAPER 1 Message-ID: <20010904062346.H81767-100000@disney.csl.cornell.edu> MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII PRnet: This paper presents an overview of packet radio networks and effective protocols to control dataflow over these distributed, wireless systems. The research conducted from 1973-87, pioneered the field of wireless computer networks in mobile environments. A low-cost packet radio (LPR), which communicate via line-of-sight with other LPRs, is the main hardware component of these types of networks. The main challenges for packet radio networks are efficiently maintaining topology information about a dynamically changing network and time-multiplexing transmission/receiving packets for neighbor LPRs. The main contributions of this paper are a) algorithms that learn and maintain network connectivity information and b) transmission protocols that efficiently operate on shared and broadcast channels. The s/w in the LPRs not only automatically figures out its network neighborhood, but also continuously gathers quality of service statistics about various channels. Considering this paper's date of publication, the system appears to have fairly robust error detection as well as reliable packet delivery. For a low-cost packet radio of this generation, I was impressed with the hardware capabilities of this system which included spread-spectrum, FDMA support, and 24dB power attenuation. The most important thing missing from this paper was performance data from real LPR networks. Without real-world data on network throughput, it is difficult to evaluate how well the author's algorithms and protocols actually work. For instance, I would be interested in how large neighborhoods affect performance of this system and what the latency is for a LPR to learn its environment and route data efficiently. I would have liked to see a comparison of this system to say, a system composed of local hosts communicating with base-stations and base-stations interfacing with each other. From gupta@CS.Cornell.EDU Tue Sep 4 10:23:26 2001 Return-Path: Received: from exchange.cs.cornell.edu (exchange.cs.cornell.edu [128.84.97.8]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84ENQq11723 for ; Tue, 4 Sep 2001 10:23:26 -0400 (EDT) content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" X-MimeOLE: Produced By Microsoft Exchange V6.0.4712.0 Subject: 615 PAPER 1 Date: Tue, 4 Sep 2001 10:23:27 -0400 Message-ID: <404A3A4758DDCC4C8A5C9A537384F9D6043A7B@opus.cs.cornell.edu> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: 615 PAPER 1 Thread-Index: AcE1TSYDmEhHkZ2MEdW1ugCgydyP2Q== From: "Indranil Gupta" To: "Emin Gun Sirer" Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by sundial.cs.cornell.edu id f84ENQq11723 ----START---- Reviewer: Indranil Gupta The DARPA Packet Radio Network Protocols, J. Jubin and J.D. Tornow. Summary: This paper describes the first prototype of an ad-hoc network (called a packet radio network or PRN by the authors) with 138 entities, developed under a DARPA project initiated in 1973. The system is engineered for multi-hop network routing and automated route management in the face of flaky wireless links. The system supports remote monitoring and debugging. The authors use a proactive routing protocol. Each packet radio (PR or a 'node') maintains a list of neighboring (1 hop) PRs and the quality of the links to them (using periodic PROP messages and exponential smoothing over the collected data). PROP messages are also propagated further on for 'good' links, with increasing hop numbers (called 'tiers' in the paper) and shortest routes are calculated using these hop numbers. All these are maintained in tables where entries time out, so if a link goes 'bad', it is eventually erased from all tables. Packets are acked at each forwarded hop - this requires symmetric connectivity. After three (failed) forwarding tries at a hop, a flag is set so that all neighbors will try to alternate route it (forward it) - to prevent infinite alternate routing, neighbors do this only if their hop number to the destination is <= that of the intended next hop of the packet. Duplicate packets are filtered using sender sequence numbers. Packets may contain QoS flags (Speech) that give them priority over other packets. Congestion control is done by having a window size of 1 at the sender. The congestion is pushed out from the low bandwidth network to the end points. At each forwarding hop, some notion of weighted fair queuing across different quality incoming links at that PR. Carrier sense multiple access is used at the link layer. Half-rate coding and CRCs are used for error control. FEC is used, with three different coding rates; dynamic algorithms are used to select the best bandwidth usage vs. error rate tradeoff depending on the current link quality. Chip modulation and bit stuffing are also used. Two bandwidth rates are used 400 and 100 kbps, with the former used on good links and the latter used on lossy links. Adaptive power control is used to solve the near-far problem. 20 RF frequencies are reused within clusters of PRs. Comments: - Many of the ideas currently 'hot' topics in today's ad-hoc networking research seem to originate from this work. This is the first proactive multi-hop routing algorithm. - The protocol classifies a hop as 'good' or 'bad' depending on the loss characteristics, and propagate tier messages only for good links. In reality, even the goodness of good links vary (SIR, or signal to interference ratio, on that link). If this information is propagated in the tier messages as well, routes can be chosen through links that are 'better' than other 'good' links. This could also potentially avoid the route having flaky links that just happened to be in a good phase during the route discovery. - The authors (using pacing) push congestion out of the wireless networks to the end points. Given today's wireless link speeds and hardware, is this still a viable approach to congestion control ? How does this end to end argument apply to different applications and workloads ? - A major part of the protocol is devoted to handling link level contention among neighboring PRs that can hear each other (section V.C on CSMA). Can this contention not be eliminated completely by using, within each cluster, the large number channels available today (large number of frequencies in FDMA, infinite number of codes in CDMA) and today's hardware (multiple receivers at a single PR) ? - The hop numbering restriction in alternate routing could cause packets to reach a dead-end. The alternative is to let packets circulate within the network (possibly forever), perhaps with a guarantee on eventual delivery (eg., when the connectivity is better). Does this make sense for any application ? Has it been explored ? - Hidden collisions are not avoided at the link layer. This could be a significant source of packet losses. An RTS-CTS dialogue could help avoid this. ----END---- From avneesh@csl.cornell.edu Tue Sep 4 10:49:39 2001 Return-Path: Received: from capricorn.ds.csl.cornell.edu (capricorn.csl.cornell.edu [132.236.71.92]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84Encq14309 for ; Tue, 4 Sep 2001 10:49:38 -0400 (EDT) Subject: 615 PAPER #1 MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" Date: Tue, 4 Sep 2001 10:53:38 -0400 Message-ID: <97C142C1212ED545B0023A177F5349C4053A4A@capricorn.ds.csl.cornell.edu> Content-Class: urn:content-classes:message X-MimeOLE: Produced By Microsoft Exchange V6.0.4712.0 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: 615 PAPER #1 Thread-Index: AcE1UWF8Hsqt/zybTaGrUy2nBqFUpg== From: "Avneesh Bhatnagar" To: Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by sundial.cs.cornell.edu id f84Encq14309 The DARPA Packet Radio Network Protocols Summary: This paper explains the features and implementational issues of the Packet Radio Network PRNET, which uses broadcast radio as the communication medium and works on a low network level. This allows for mobility, rapid deployment and ease of configuration. However the constraints in this design are that packet radios (PRs) may not be able to transmit and reciev at the same time, and the protocols need to accomodate transmission only when the receiving PR is not receiving from another PR.Network management is done dynamically based upon link connectivity (summarized later). The PRNET subnet comprises of (1)PRs (implemented as Low cost Prs containing digital and RF components), which engage in radio transmission and are used by (2) devices which are connected by a wire medium to these PRs. Devices could be host computers connected via NIUs, LANs etc which run a conventional protocol. The network size supported by the described system is that of 138 entities (including PRs and devices), with each PR being able to support about 16 neighbors. Congested PRs use logical and actual neighbors to alleviate the problem. Other features of the PRs include a facility for remote debugging and presence of network monitors to analyze the network behavior. Link connectivity and network routes are determined through: (a) Neighbor tables, which are created by each PR which broadcasts a Packet radio organization packet (PROP) every 7s to announce its existence to its neighbors. Link quality is determined by taking a ratio of packets received (hence acked) by the packets sent. This is important when determinig the route to a neighboring node. (b) Teir tables: These mantain the knowledge of the best PR to forward a packet to, depending upon how many hops away that PR is from the sender/forwarder. This hence helps to calculate the shortest route to a destination. It is also important to discover route loops when calculating new routes or updating route information as a result of failed routes. Finally, with the help of device tables, which mantain information through logical addressing of the device associated to the particular PR, the realtionship of the device to the PR can be altered dynamically and hence computation of the new address of the attached device is made easier. The device table allows the PR to know where to forward the received packet. The protocols (forwarding and transmission) provide various facilities with the help of: (a) Packet headers, which are the end to end (ETE) to denote type of service and routing headers to denote nodes in the route to the destination. (b) Forwarding acknowledgement, which could be active (as a routing header transmitted to the source), or as passive acknowledgement, which comes as a result of the broadcast nature of communication. Other facilities are error correction and detection (using FEC and CRC), determining alternative routes( by setting bits in the header), and eliminating duplicates( by using UPIs). (c) Flow and congestion control is provided through pacing, CSMA and fairness queueing. Pacing consists of single threading, which requires acks to be received before transmission to the same PR; calculation of forwarding delay as to how much delay should be provided b/w retransmits; and evaluation of a pacing function which actually helps to reduce congestion due to retransmitted packets by allowing data to be forwarded and acknowledged by further nodes before seding more data. Fairness queueing tries accomodate packets moving in a fast direction from being stuck being those in a slow direction (as evaluated by pacing), due to FIFO constraints. It also reduces unfairness due to pacing in case of weaklinks thus enabling them to move information in a FIFO environment. Finally CSMA helps reduce contention in the medium by detecting the business of the channel before transmissionand using policies such as backoff during retransmits. The paper concludes by describing the hardware support provided for error control, modulation criteria, RF frequency range and receive signal power measurement for efficient route calculation. Critique: The paper clearly describes the PRNET system and the software, hardware components. In some places, it correctly anticipates future applications and problems, some of which are relevant in today's domain also, primarily the calculation of an efficient and fair route. Couple of points to note though, first of all the reason or how the network size is 138 is not explained. Secondly, it seems that each node stores a lot of routing information, infact information about the whole network. This might be a concern. From daehyun@csl.cornell.edu Tue Sep 4 10:57:49 2001 Return-Path: Received: from postoffice2.mail.cornell.edu (postoffice2.mail.cornell.edu [132.236.56.10]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84Evmq15259 for ; Tue, 4 Sep 2001 10:57:49 -0400 (EDT) Received: from atlas (syr-66-24-28-91.twcny.rr.com [66.24.28.91]) by postoffice2.mail.cornell.edu (8.9.3/8.9.3) with SMTP id KAA00652 for ; Tue, 4 Sep 2001 10:57:49 -0400 (EDT) Message-ID: <000a01c13551$ac3c7170$5b1c1842@atlas> From: "Daehyun Kim" To: Subject: 615 PAPER (1) Date: Tue, 4 Sep 2001 10:55:43 -0400 MIME-Version: 1.0 X-Security: MIME headers sanitized on sundial.cs.cornell.edu See http://www.impsec.org/email-tools/procmail-security.html for details. $Revision: 1.129 $Date: 2001-04-14 20:20:43-07 X-Security: The postmaster has not enabled quarantine of poisoned messages. Content-Type: multipart/alternative; boundary="----=_NextPart_000_0007_01C13530.24952120" X-Priority: 3 X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook Express 5.50.4522.1200 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4522.1200 This is a multi-part message in MIME format. ------=_NextPart_000_0007_01C13530.24952120 Content-Type: text/plain; charset="ks_c_5601-1987" Content-Transfer-Encoding: base64 VGhpcyBwYXBlciBkZXNjcmliZXMgREFSUEEgUGFja2V0IFJhZGlvIE5ldHdvcmtzLg0KVGhlIGZp cnN0IHR3byBzZWN0aW9ucyAtIHNlY3Rpb24gMSBhbmQgMiBleHBsYWluIHRoZSBnZW5lcmFsIGlk ZWEgYW5kIHN0cnVjdHVyZS4NClNlY3Rpb24gMyBhbmQgNCBzaG93IHRoZSBuZXR3b3JrIGxheWVy IHByb3RvY29scy4gU2VjdGlvbiA1IGFuZCA2IHRhbGsgYWJvdXQgdGhlDQpEYXRhIGxpbmsgbGF5 ZXIgcHJvdG9jb2xzLiBUaGUgbGFzdCBzZWN0aW9uIC0gc2VjdGlvbiA3IGNvbmNsdWRlcy4NCklu IHRoaXMgY2xhc3MsIHdlIGFyZSBpbnRlcmVzdGVkIGluIHRoZSByb3V0aW5nIGFsZ29yaXRobXMs IHNvIEknbGwgZm9jdXMgb24gdGhlDQpzZWNvbmQgcGFydC4NCg0KRm9yIHJvdXRpbmcsIHRocmVl IHRhYmxlcyBhcmUgbWFpbnRhaW5lZC4NCg0KMS4gTmVpZ2hib3IgdGFibGUgOiB0aGlzIHRhYmxl cyBjb250YWlucyBuZWlnaGJvciBQYWNrZXQgUmFkaW9zIChQUi1zKS4NCjIuIFRpZXIgVGFibGUg OiB0aGlzIHRhYmxlcyBtYWludGFpbnMgdGhlIHJvdXRpbmcgaW5mb3JtYXRpb24uIFRoZSBiZXN0 IG5laWdoYm9yDQogICAgICAgICAgICAgICAgUFIgdG8gZm9yd2FyZCBwYWNrZXRzIGFuZCB0aGUg dGllciBmb3IgZXZlcnkgcHJvc3BlY3RpdmUgZGVzdGluYXRpb24NCiAgaXMgc3RvcmVkLg0KMy4g RGV2aWNlIFRhYmxlIDogdGhpcyB0YWJsZSBwcm92aWRlcyB0aGUgZGV2aWNlLXRvLVBSIG1hcHBp bmcgZm9yIHRoZSBkeW5hbWljDQogICAgICAgICAgICAgICAgICBhZGRyZXNzaW5nLg0KDQpUbyBi dWlsZCBhbmQgdXBkYXRlIHRoZSB0YWJsZXMsIGEgUGFja2V0IFJhZGlvIE9yZ2FuaXphdGlvbiBQ YWNrZXQgKFBST1ApIGlzDQpicm9hZGNhc3QgZXZlcnkgNy41IHNlYy4gDQoNCkEgcGFja2V0IGlz IHRyYW5zZmVycmVkIGZyb20gYSBkZXZpY2UgdG8gYW5vdGhlciBkZXZpY2UuIEZpcnN0LCB0aGUg cGFja2V0IGlzIHBhc3NlZA0KdG8gYSBQUiBhdHRhY2hlZCB0byB0aGUgZGV2aWNlLiBUaGUgUFIg Zm9yd2FyZHMgdGhlIHBhY2tldCBiYXNlZCBvbiB0aGUgdGllciBUYWJsZS4NClRoZSBkZXN0aW5h dGlvbiBQUiBwYXNzZWQgdGhlIHBhY2tldCB0byB0aGUgZGVzdGluYXRpb24gZGV2aWNlLg0KDQpT ZXZlcmFsIHRlY2huaXF1ZXMgYXJlIHVzZWQgdG8gY29wZSB3aXRoIGVycm9ycy4NCg0KMS4gUmV0 cmFuc21pc3Npb24gOiBJZiBhIFBSIGRvZXMgbm90IHJlY2VpdmUgYW4gYWNrbm93bGVkZ21lbnQs IGl0IHJldHJhbnNtaXRzLiAgIA0KMi4gRXJyb3IgQ29udHJvbCA6IEZFQyBhbmQgQ1JDIGFyZSB1 c2VkLg0KMy4gQWx0ZXJuYXRlIFJvdXRpbmcgOiBJZiBhIHJvdXQgZmFpbHMgdGhyZWUgdGltZXMs IGFsdGVybmF0ZSByb3V0ZXMgYXJlIHVzZWQuDQo0LiBEdXBsaWNhdGUgRmlsdGVyaW5nIDogRXZl cnkgcGFja2V0IGhhcyBhIHVuaXF1ZSBJRCAoVVBJKSBhbmQgZHVwbGljYXRlZCBwYWNrZXRzDQog ICAgICAgICAgICAgICAgICAgICAgICAgYXJlIGRpc2NhcmRlZC4NCg0KSW4gbXkgb3Bpbmlvbiwg dGhlIG1haW4gd2Vha25lc3Mgb2YgdGhlIGFsZ29yaXRobSBpbiB0aGlzIHBhcGVyIGlzIHRoYXQg aXQgbmVlZHMgdG8NCnVwZGF0ZSB0aGUgdGFibGVzIHBlcmlvZGljYWxseS4gSSB0aGluayB0aGlz IGRvZXMgbm90IHJlZmxlY3QgdGhlIGR5bmFtaWNzIG9mIHRoZQ0KYWQtaG9jIG5ldHdvcmsgd2Vs bC4NCg0KMS4gRWFjaCBQUiBuZWVkcyB0byBtYWludGFpbiB0aGUgY29tcGxldGUgcm91dGluZyBp bmZvcm1hdGlvbiBmb3IgdGhlIGVudGlyZSBuZXR3b3JrLg0KICAgSSB3b25kZXIgaWYgdGhpcyBj YW4gYmUgZWZmaWNpZW50IGZvciB0aGUgYWQtaG9jIG5ldHdvcmsuIFRoZSBuZXR3b3JrIHRvcG9s b2d5DQogICBjaGFuZ2VzIGZyZXF1ZW50bHkgYW5kIHRoaXMgaW5mb3JtYXRpb24gY2FuIGJlIHBy b3BhZ2F0ZWQgYXQgdGhlIHNwZWVkIG9mIG9uZSB0aWVyDQogICBwZXIgNy41IHNlYy4gIFRoaXMg bWF5IG5vdCBwcm9wYWdhdGUgdGhyb3VnaCB0aGUgbmV0d29yayBlbm91Z2ggZmFzdC4gUFIxIG1h eSBoYXZlDQogICBhIHdyb25nIHJvdXRpbmcgaW5mb3JtYXRpb24gZm9yIFBSMiwgYmVjYXVzZSBQ UjIgaGFzIGFscmVhZHkgY2hhbmdlZCBhZ2FpbiB3aGVuIFBSMQ0KICAgZ2V0cyB0aGUgaW5mb3Jt YXRpb24gYWJvdXQgdGhlIHByZXZpb3VzIGNoYW5nZS4gVGhlcmVmb3JlLCB0aGlzIHdpbGwgbGlt aXQgdGhlIHRvdGFsDQogICBzaXplIG9mIHRoZSBuZXR3b3JrIGFuZCByZWR1Y2UgdGhlIHNjYWxh YmlsaXR5LiAgIA0KDQoyLiBJbiB0aGUgb3Bwb3NpdGUgY2FzZSBvZiAxLCBpZiB0aGUgbmV0d29y ayB0b3BvbG9neSBjaGFuZ2VzIG5vdCBtdWNoLCBQUk9QDQogICBvdmVyaGVhZCB3aWxsIGJlIGJp Zy4gVGhvdWdoIGFkLWhvYyBuZXR3b3JrIGlzIGR5bmFtaWMsIHNvbWUgcGFydHMgb2YgdGhlIG5l dHdvcmsgbWlnaHQNCiAgIGJlIHN0YXRpYyBmb3Igc29tZSB0aW1lLiBCdXQgUFJPUCBuZWVkcyB0 byBiZSBzZW50IGV2ZXJ5IDcuNSBzZWMgZm9yIGV2ZXJ5IFBSLiBJbg0KICAgd2lyZWxlc3MgbmV0 d29yaywgc2VuZGluZyBhbmQgcmVjZWl2aW5nIHBhY2tldHMgYXJlIGV4cGVuc2l2ZSwgc28gdGhl IG51bWJlcg0KICAgb2YgdHJhbnNtaXNzaW9ucyBzaG91bGQgYmUgbWluaW1pemVkLg0KDQpSZXZp ZXdlciA6IERhZWh5dW4gS2ltDQo= ------=_NextPart_000_0007_01C13530.24952120 Content-Type: text/html; charset="ks_c_5601-1987" Content-Transfer-Encoding: base64 PCFET0NUWVBFIEhUTUwgUFVCTElDICItLy9XM0MvL0RURCBIVE1MIDQuMCBUcmFuc2l0aW9uYWwv L0VOIj4NCjxIVE1MPjxIRUFEPg0KPE1FVEEgaHR0cC1lcXVpdj1Db250ZW50LVR5cGUgY29udGVu dD0idGV4dC9odG1sOyBjaGFyc2V0PWtzX2NfNTYwMS0xOTg3Ij4NCjxNRVRBIGNvbnRlbnQ9Ik1T SFRNTCA1LjUwLjQ2MTYuMjAwIiBuYW1lPUdFTkVSQVRPUj4NCjxTVFlMRT48L1NUWUxFPg0KPC9I RUFEPg0KPEJPRFkgYmdDb2xvcj0jZmZmZmZmPg0KPERJVj48Rk9OVCBzaXplPTI+VGhpcyBwYXBl ciBkZXNjcmliZXMgREFSUEEgUGFja2V0IFJhZGlvIE5ldHdvcmtzLjxCUj5UaGUgZmlyc3QgDQp0 d28gc2VjdGlvbnMgLSBzZWN0aW9uIDEgYW5kIDIgZXhwbGFpbiB0aGUgZ2VuZXJhbCBpZGVhIGFu ZCANCnN0cnVjdHVyZS48QlI+U2VjdGlvbiAzIGFuZCA0IHNob3cgdGhlIG5ldHdvcmsgbGF5ZXIg cHJvdG9jb2xzLiBTZWN0aW9uIDUgYW5kIDYgDQp0YWxrIGFib3V0IHRoZTxCUj5EYXRhIGxpbmsg bGF5ZXIgcHJvdG9jb2xzLiBUaGUgbGFzdCBzZWN0aW9uIC0gc2VjdGlvbiA3IA0KY29uY2x1ZGVz LjxCUj5JbiB0aGlzIGNsYXNzLCB3ZSBhcmUgaW50ZXJlc3RlZCBpbiB0aGUgcm91dGluZyBhbGdv cml0aG1zLCBzbyANCkknbGwgZm9jdXMgb24gdGhlPEJSPnNlY29uZCBwYXJ0LjwvRk9OVD48L0RJ Vj4NCjxESVY+PEZPTlQgc2l6ZT0yPjwvRk9OVD4mbmJzcDs8L0RJVj4NCjxESVY+PEZPTlQgc2l6 ZT0yPkZvciByb3V0aW5nLCB0aHJlZSB0YWJsZXMgYXJlIG1haW50YWluZWQuPC9GT05UPjwvRElW Pg0KPERJVj48Rk9OVCBzaXplPTI+PC9GT05UPiZuYnNwOzwvRElWPg0KPERJVj48Rk9OVCBzaXpl PTI+MS4gTmVpZ2hib3IgdGFibGUgOiB0aGlzIHRhYmxlcyBjb250YWlucyBuZWlnaGJvciBQYWNr ZXQgDQpSYWRpb3MgKFBSLXMpLjxCUj4yLiBUaWVyIFRhYmxlIDogdGhpcyB0YWJsZXMgbWFpbnRh aW5zIHRoZSByb3V0aW5nIGluZm9ybWF0aW9uLiANClRoZSBiZXN0IA0KbmVpZ2hib3I8QlI+Jm5i c3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7 Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7IA0KUFIgdG8gZm9yd2FyZCBwYWNrZXRzIGFu ZCB0aGUgdGllciBmb3IgZXZlcnkgcHJvc3BlY3RpdmUgDQpkZXN0aW5hdGlvbjxCUj4mbmJzcDsm bmJzcDtpcyBzdG9yZWQuPEJSPjMuIERldmljZSBUYWJsZSA6IHRoaXMgdGFibGUgcHJvdmlkZXMg DQp0aGUgZGV2aWNlLXRvLVBSIG1hcHBpbmcgZm9yIHRoZSANCmR5bmFtaWM8QlI+Jm5ic3A7Jm5i c3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7 Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7Jm5ic3A7IA0KYWRkcmVzc2luZy48L0ZPTlQ+ PC9ESVY+DQo8RElWPjxGT05UIHNpemU9Mj48L0ZPTlQ+Jm5ic3A7PC9ESVY+DQo8RElWPjxGT05U IHNpemU9Mj5UbyBidWlsZCBhbmQgdXBkYXRlIHRoZSB0YWJsZXMsIGEgUGFja2V0IFJhZGlvIE9y Z2FuaXphdGlvbiANClBhY2tldCAoUFJPUCkgaXM8QlI+YnJvYWRjYXN0IGV2ZXJ5IDcuNSBzZWMu IDwvRk9OVD48L0RJVj4NCjxESVY+PEZPTlQgc2l6ZT0yPjwvRk9OVD4mbmJzcDs8L0RJVj4NCjxE SVY+PEZPTlQgc2l6ZT0yPkEgcGFja2V0IGlzIHRyYW5zZmVycmVkIGZyb20gYSBkZXZpY2UgdG8g YW5vdGhlciBkZXZpY2UuIA0KRmlyc3QsIHRoZSBwYWNrZXQgaXMgcGFzc2VkPEJSPnRvIGEgUFIg YXR0YWNoZWQgdG8gdGhlIGRldmljZS4gVGhlIFBSIGZvcndhcmRzIA0KdGhlIHBhY2tldCBiYXNl ZCBvbiB0aGUgdGllciBUYWJsZS48QlI+VGhlIGRlc3RpbmF0aW9uIFBSIHBhc3NlZCB0aGUgcGFj a2V0IHRvIA0KdGhlIGRlc3RpbmF0aW9uIGRldmljZS48L0ZPTlQ+PC9ESVY+DQo8RElWPjxGT05U IHNpemU9Mj48L0ZPTlQ+Jm5ic3A7PC9ESVY+DQo8RElWPjxGT05UIHNpemU9Mj5TZXZlcmFsIHRl Y2huaXF1ZXMgYXJlIHVzZWQgdG8gY29wZSB3aXRoIGVycm9ycy48L0ZPTlQ+PC9ESVY+DQo8RElW PjxGT05UIHNpemU9Mj48L0ZPTlQ+Jm5ic3A7PC9ESVY+DQo8RElWPjxGT05UIHNpemU9Mj4xLiBS ZXRyYW5zbWlzc2lvbiA6IElmIGEgUFIgZG9lcyBub3QgcmVjZWl2ZSBhbiANCmFja25vd2xlZGdt ZW50LCBpdCByZXRyYW5zbWl0cy4mbmJzcDsmbmJzcDsgPEJSPjIuIEVycm9yIENvbnRyb2wgOiBG RUMgYW5kIENSQyANCmFyZSB1c2VkLjxCUj4zLiBBbHRlcm5hdGUgUm91dGluZyA6IElmIGEgcm91 dCBmYWlscyB0aHJlZSB0aW1lcywgYWx0ZXJuYXRlIA0Kcm91dGVzIGFyZSB1c2VkLjxCUj40LiBE dXBsaWNhdGUgRmlsdGVyaW5nIDogRXZlcnkgcGFja2V0IGhhcyBhIHVuaXF1ZSBJRCAoVVBJKSAN CmFuZCBkdXBsaWNhdGVkIA0KcGFja2V0czxCUj4mbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJz cDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsm bmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJzcDsmbmJz cDsgDQphcmUgZGlzY2FyZGVkLjwvRk9OVD48L0RJVj4NCjxESVY+PEZPTlQgc2l6ZT0yPjwvRk9O VD4mbmJzcDs8L0RJVj4NCjxESVY+PEZPTlQgc2l6ZT0yPkluIG15IG9waW5pb24sIHRoZSBtYWlu IHdlYWtuZXNzIG9mIHRoZSBhbGdvcml0aG0gaW4gdGhpcyANCnBhcGVyIGlzIHRoYXQgaXQgbmVl ZHMgdG88QlI+dXBkYXRlIHRoZSB0YWJsZXMgcGVyaW9kaWNhbGx5LiBJIHRoaW5rIHRoaXMgZG9l cyANCm5vdCByZWZsZWN0IHRoZSBkeW5hbWljcyBvZiB0aGU8QlI+YWQtaG9jIG5ldHdvcmsgd2Vs bC48L0ZPTlQ+PC9ESVY+DQo8RElWPjxGT05UIHNpemU9Mj48L0ZPTlQ+Jm5ic3A7PC9ESVY+DQo8 RElWPjxGT05UIHNpemU9Mj4xLiBFYWNoIFBSIG5lZWRzIHRvIG1haW50YWluIHRoZSBjb21wbGV0 ZSByb3V0aW5nIGluZm9ybWF0aW9uIA0KZm9yIHRoZSBlbnRpcmUgbmV0d29yay48QlI+Jm5ic3A7 Jm5ic3A7IEkgd29uZGVyIGlmIHRoaXMgY2FuIGJlIGVmZmljaWVudCBmb3IgDQp0aGUgYWQtaG9j IG5ldHdvcmsuIFRoZSBuZXR3b3JrIHRvcG9sb2d5PEJSPiZuYnNwOyZuYnNwOyBjaGFuZ2VzIGZy ZXF1ZW50bHkgYW5kIA0KdGhpcyBpbmZvcm1hdGlvbiBjYW4gYmUgcHJvcGFnYXRlZCBhdCB0aGUg c3BlZWQgb2Ygb25lIHRpZXI8QlI+Jm5ic3A7Jm5ic3A7IHBlciANCjcuNSBzZWMuJm5ic3A7IFRo aXMgbWF5IG5vdCBwcm9wYWdhdGUgdGhyb3VnaCB0aGUgbmV0d29yayBlbm91Z2ggZmFzdC4gUFIx IG1heSANCmhhdmU8QlI+Jm5ic3A7Jm5ic3A7IGEgd3Jvbmcgcm91dGluZyBpbmZvcm1hdGlvbiBm b3IgUFIyLCBiZWNhdXNlIFBSMiBoYXMgDQphbHJlYWR5IGNoYW5nZWQgYWdhaW4gd2hlbiBQUjE8 QlI+Jm5ic3A7Jm5ic3A7IGdldHMgdGhlIGluZm9ybWF0aW9uIGFib3V0IHRoZSANCnByZXZpb3Vz IGNoYW5nZS4gVGhlcmVmb3JlLCB0aGlzIHdpbGwgbGltaXQgdGhlIHRvdGFsPEJSPiZuYnNwOyZu YnNwOyBzaXplIG9mIA0KdGhlIG5ldHdvcmsgYW5kIHJlZHVjZSB0aGUgc2NhbGFiaWxpdHkuJm5i c3A7Jm5ic3A7IDwvRk9OVD48L0RJVj4NCjxESVY+PEZPTlQgc2l6ZT0yPjwvRk9OVD4mbmJzcDs8 L0RJVj4NCjxESVY+PEZPTlQgc2l6ZT0yPjIuIEluIHRoZSBvcHBvc2l0ZSBjYXNlIG9mIDEsIGlm IHRoZSBuZXR3b3JrIHRvcG9sb2d5IGNoYW5nZXMgDQpub3QgbXVjaCwgUFJPUDxCUj4mbmJzcDsm bmJzcDsgb3ZlcmhlYWQgd2lsbCBiZSBiaWcuIFRob3VnaCBhZC1ob2MgbmV0d29yayBpcyANCmR5 bmFtaWMsIHNvbWUgcGFydHMgb2YgdGhlIG5ldHdvcmsgbWlnaHQ8QlI+Jm5ic3A7Jm5ic3A7IGJl IHN0YXRpYyBmb3Igc29tZSANCnRpbWUuIEJ1dCBQUk9QIG5lZWRzIHRvIGJlIHNlbnQgZXZlcnkg Ny41IHNlYyBmb3IgZXZlcnkgUFIuIEluPEJSPiZuYnNwOyZuYnNwOyANCndpcmVsZXNzIG5ldHdv cmssIHNlbmRpbmcgYW5kIHJlY2VpdmluZyBwYWNrZXRzIGFyZSBleHBlbnNpdmUsIHNvIHRoZSAN Cm51bWJlcjxCUj4mbmJzcDsmbmJzcDsgb2YgdHJhbnNtaXNzaW9ucyBzaG91bGQgYmUgbWluaW1p emVkLjwvRk9OVD48L0RJVj4NCjxESVY+PEZPTlQgc2l6ZT0yPjwvRk9OVD4mbmJzcDs8L0RJVj4N CjxESVY+PEZPTlQgc2l6ZT0yPlJldmlld2VyIDogRGFlaHl1biBLaW08L0ZPTlQ+PC9ESVY+PC9C T0RZPjwvSFRNTD4NCg== ------=_NextPart_000_0007_01C13530.24952120-- From mh97@cornell.edu Tue Sep 4 11:09:46 2001 Return-Path: Received: from postoffice.mail.cornell.edu (postoffice.mail.cornell.edu [132.236.56.7]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84F9jq16495 for ; Tue, 4 Sep 2001 11:09:46 -0400 (EDT) Received: from mars (dhcp4.csl.cornell.edu [132.236.71.51]) by postoffice.mail.cornell.edu (8.9.3/8.9.3) with ESMTP id LAA15810 for ; Tue, 4 Sep 2001 11:09:47 -0400 (EDT) From: "ming hao" To: "'Emin Gun Sirer'" Subject: 615 PAPER 1 Date: Tue, 4 Sep 2001 11:09:22 -0400 Message-ID: <000201c13553$95007f40$3347ec84@mars> MIME-Version: 1.0 Content-Type: text/plain; charset="us-ascii" Content-Transfer-Encoding: 7bit X-Priority: 3 (Normal) X-MSMail-Priority: Normal X-Mailer: Microsoft Outlook, Build 10.0.2616 X-MimeOLE: Produced By Microsoft MimeOLE V5.50.4807.1700 Importance: Normal In-Reply-To: <200109022045.f82Kj9409239@hoho.cs.cornell.edu> The DARPA Packet Radio Network Protocols This paper presents a packet communication network consisting Of mobile nodes with wireless connection. This paper details The network's automatic management, routing protocol, hardware Components and its effect on the design and potential applications. In general, PRNET project is probably the first project exploring The ad hoc network as far as I know and is really pioneer work In the field, though there are some apparent places where improvement Are necessary. Because I think their main goal was to just implement this type of network and performance and optimization is not their priority. The main part of this paper talks about the routing protocol and automatic management. This is not only the most difficult part for ad hoc network because of mobile nature of nodes and constant changing of topology of the network, but also the biggest advantage of this kind of network, without central management, flexible, easy deployment. The routing info is Set up by rippling the PROP through the network. The paper ignored one problem that how to prevent the infinite broadcasting of PROP. There must be A mechanism for PR to stop forwarding PROP at some point. the nodes constantly check the quality of links and broadcasting news of bad link to change to the routing. Further, data packet is also used to set up routing setup. packet is forwarded though net along single path with forwarding-ack-retransmission protocol. Passive acknowledgement takes broadcasting feature of radio. In case of there times transmission failure, alternate routing is needed by setting one bit in the header. Transmission protocol mainly focuses on how to choose interval delay. Also, in order to avoid multiple transmission at the same time, CSMA is used. Queue fairness is also considered. As I mentioned, almost no performance data is mentioned. There is a huge Improvement room there. Such as periodically broadcasting wastes bandwidth. Can full-duplex channel be used to accelerate the three step transmission Protocol? What problems come up if promiscuous mode and bidirectionally good Channel are not available? What is the most efficient way to store the table And how to accelerate the table lookup? -ming From daehyun@csl.cornell.edu Tue Sep 4 11:14:05 2001 Return-Path: Received: from wilkes.csl.cornell.edu (wilkes.csl.cornell.edu [132.236.71.69]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84FE4q17046 for ; Tue, 4 Sep 2001 11:14:04 -0400 (EDT) Received: (from daehyun@localhost) by wilkes.csl.cornell.edu (8.9.3/8.9.2) id LAA02934 for egs@cs.cornell.edu; Tue, 4 Sep 2001 11:14:01 -0400 (EDT) (envelope-from daehyun) From: Daehyun Kim Message-Id: <200109041514.LAA02934@wilkes.csl.cornell.edu> Subject: 615 PAPER (1) To: egs@CS.Cornell.EDU Date: Tue, 4 Sep 2001 11:14:01 -0400 (EDT) X-Mailer: ELM [version 2.4ME+ PL54 (25)] MIME-Version: 1.0 Content-Type: text/plain; charset=US-ASCII Content-Transfer-Encoding: 7bit This paper describes DARPA Packet Radio Networks. The first two sections - section 1 and 2 explain the general idea and structure. Section 3 and 4 show the network layer protocols. Section 5 and 6 talk about the Data link layer protocols. The last section - section 7 concludes. In this class, we are interested in the routing algorithms, so I'll focus on the second part. For routing, three tables are maintained. 1. Neighbor table : this tables contains neighbor Packet Radios (PR-s). 2. Tier Table : this tables maintains the routing information. The best neighbor PR to forward packets and the tier for every prospective destination is stored. 3. Device Table : this table provides the device-to-PR mapping for the dynamic addressing. To build and update the tables, a Packet Radio Organization Packet (PROP) is broadcast every 7.5 sec. A packet is transfered from a device to another device. First, the packet is passed to a PR attached to the device. The PR forwards the packet based on the tier Table. The destination PR passed the packet to the destination device. Several techniques are used to cope with errors. 1. Retransmission : If a PR does not receive an acknowledgment, it retransmits. 2. Error Control : FEC and CRC are used. 3. Alternate Routing : If a rout fails three times, alternate routes are used. 4. Duplicate Filtering : Every packet has a unique ID (UPI) and duplicated packets are discarded. In my opinion, the main weakness of the algorithm in this paper is that it needs to update the tables periodically. I think this does not reflect the dynamics of the ad-hoc network well. 1. Each PR needs to maintain the complete routing information for the entire network. I wonder if this can be efficient for the ad-hoc network. The network topology changes frequently and this information can be propagated at the speed of one tier per 7.5 sec. This may not propagate through the network enough fast. PR1 may have a wrong routing information for PR2, because PR2 has already changed again when PR1 gets the information about the previous change. Therefore, this will limit the total size of the network and reduce the scalability. 2. In the opposite case of 1, if the the network topology changes not much, PROP overhead will be big. Though ad-hoc network is dynamic, some parts of the network might be static for some time. But PROP needs to be sent every 7.5 sec for every PR. In wireless network, sending and receiving packets are expensive, so the number of transmissions should be minimized. From ramasv@CS.Cornell.EDU Tue Sep 4 11:22:09 2001 Return-Path: Received: from exchange.cs.cornell.edu (exchange.cs.cornell.edu [128.84.97.8]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84FM9q18080 for ; Tue, 4 Sep 2001 11:22:09 -0400 (EDT) content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="iso-8859-1" X-MimeOLE: Produced By Microsoft Exchange V6.0.4712.0 Subject: 615 PAPER 1 Date: Tue, 4 Sep 2001 11:22:10 -0400 Message-ID: <706871B20764CD449DB0E8E3D81C4D4301E7F264@opus.cs.cornell.edu> X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: 615 PAPER 1 Thread-Index: AcE1VPolIWti2ZUTEdWTbQCQJ5m7oA== From: "Venu Ramasubramanian" To: "Emin Gun Sirer" Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from quoted-printable to 8bit by sundial.cs.cornell.edu id f84FM9q18080 The DARPA: Packet Radio Network Protocols This invited paper published in 1987 suveys the state of research on packet radio networks. Despite being more than a decade old, it is often cited in todays papers on ad-hoc networks. Basically, the structure of packet radio networks described in this paper also forms the classical definition for ad hoc networks, a widely researched topic today. More importantly, the core issues identified in this paper at different layers such as physical layer, data link layer, media access layer, network (routing) layer continue to remain critical issues in implementing ad hoc networks of the present generation. Although the technology described in this paper has been greatly improved upon today, this paper continues to remain a pioneering work in this field. Packet radio networks (PRNETs), consists of radios under the control of host computers and terminals that receive and generate data. The fundamental characteristics of a PRNET is its ability to be mobile and self configuring. As a consequence, these networks use peer to peer routing protocols, wherein the PRNET devices themselves act as intermediate routers. High mobility, low bandwidth wireless propagation medium, low cost (therefore not sophisticated) equipment form some of the limiting constraints in operating such networks. This paper describes protocols at different network levels that have then been designed and implemented in PRNETs. This paper describes a table-driven proactive routing protocol that employs peer-peer routing in order to deliver packets. The PRNET routing protocols consists of maintaining at each node a lookup table that indicates the next hop neighbor used to reach all the nodes in the network. This table is built and maintained by periodically broadcasting (neighbor-casting) the contents of this table at each node. It employs both passive and active acknowledgements at hop level to ensure reliability. An aspect of the periodic messaging (PROPs) activity also includes maintaining neighbor information and detecting quality of links. Although this proactive routing protocol could be dismissed as being expensive, it still forms the first of its kind in this field of research. Several important ideas such as the need to identify and use links of good quality (usage of hysteresis), topology control, multi-path forwarding have been identified quite early. Another pioneering idea (in this field) expressed in this paper involves congestion and flow control. The pacing protocol described here is very interesting and its applicability to todays ad hoc networks has not been adequately reserached yet. The need for queuing strategies in order to prevent starvation has also been presented in this paper and forms another important aspect of todays ad hoc networks. CSMA, that has been chosen as the MAC protocol, is not known to be a very efficient protocol and suffers from hidden-terminal problem. Better MAC protocols with RTS-CTS scheme have been designed later on. CSMA forces the need for using hop-level acknowledgements at routing layer and this increases the control overhead of the routing protocol. The important characteristics of the physical layer such as CDMA, forward error correction, adaptive power control etc. have led to the production of sophisticated ad hoc network and sensor network radio hardware. This is a survey of fundamental research work in the field of ad hoc and packet radio networks. Although this paper may be technologically out-dated today, it still describes a pioneering work and hence is a must-read for people doing research on ad hoc networks. From ms103@cornell.edu Tue Sep 4 11:40:33 2001 Return-Path: Received: from travelers.mail.cornell.edu (travelers.mail.cornell.edu [132.236.56.13]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84FeWq19791 for ; Tue, 4 Sep 2001 11:40:33 -0400 (EDT) Received: from travelers.mail.cornell.edu (travelers.mail.cornell.edu [132.236.56.13]) by travelers.mail.cornell.edu (8.9.3/8.9.3) with SMTP id LAA28470; Tue, 4 Sep 2001 11:40:32 -0400 (EDT) From: ms103@cornell.edu Date: Tue, 4 Sep 2001 11:40:32 -0400 (EDT) X-Sender: ms103@travelers.mail.cornell.edu To: egs@CS.Cornell.EDU Subject: 615 PAPER 1 Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII Jubin and Tornow describe the protocols, algorithms and hardware used in an experimental packet radio network. The network operates in a fully distributed manner, and has a high degree of mobility and dynamism. The algorithms used for routing, identification and link management seem fairly simplistic - but this is understandable considering that this was the first network of it's kind. The traffic generated by control structures (in particular, the routing and link connectivity features) could be excessive and network congestion by control traffic seems a very real possibilty. Little quantitative information is given about simulation or experimental results using this packet radio network. Therefore it is difficult to examine it's performance critically. Another concern is that the problem of node identification is glossed over. The authors seem to take for granted that each node will always have it's own unique identifier. From ranveer@CS.Cornell.EDU Tue Sep 4 11:42:31 2001 Return-Path: Received: from exchange.cs.cornell.edu (exchange.cs.cornell.edu [128.84.97.8]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84FgUq20187 for ; Tue, 4 Sep 2001 11:42:31 -0400 (EDT) content-class: urn:content-classes:message MIME-Version: 1.0 Content-Type: text/plain; charset="utf-8" X-MimeOLE: Produced By Microsoft Exchange V6.0.4712.0 Subject: 615 PAPER 1 Date: Tue, 4 Sep 2001 11:42:32 -0400 Message-ID: X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: 615 PAPER 1 Thread-Index: AcE1WDYhA47TxyoJTx+ipJ/Z+8dbhQ== From: "Ranveer Chandra" To: "Emin Gun Sirer" Content-Transfer-Encoding: 8bit X-MIME-Autoconverted: from base64 to 8bit by sundial.cs.cornell.edu id f84FgUq20187 The DARPA Packet Radio Network Protocols This paper describes a work in progress on packet radio networks. In the fifteen years since this paper was written, a lot of research has been on done on these problems, amd today we have more efficient radio, MAC and transport protocols. However, this does not undermine the importance of this paper. This paper would definitely have been one of the building blocks for what ad hoc networks are today. Some of the ideas proposed in this paper are still used today, although at a different level. The heartbeat mechanism proposed for neighbor discovery is still used my most routing protocols(although slightly modified). The idea of using the quality of links for routing is also interesting. This idea coupled with more information such as the traffic pattern could prove extremely useful in ad hoc networks. It is quite easy to delve into the disadvantages of different schemes proposed in this paper. There has been a lot of research on each of the problems that this paper deals with. For example with the development of 802.11 we have a more efficient MAC protocol, which proposes a more efficient RTS/CTS/Unicast/ACK technique for sending messages. Besides, the maximum wireless bandwidth is no longer 400Kbps.. 802.11b offers 11Mbps and 802.11a promises to give a maximum bandwidth of 54Mbps. The routing protocol proposed in this paper is also extremely expensive in the number and size of messages sent. Despite all the disadvantages, one cannot ignore the contribution of this paper. It has proposed the ideas on which work is still being done to provide a feasible packet radio network. It is also interesting to note that this paper has looked into different aspects of packet radio networks, but has overlooked security!!! This is one aspect of ad hoc networks that needs a lot of work to make them deployable.. From samar@ece.cornell.edu Tue Sep 4 11:56:36 2001 Return-Path: Received: from memphis.ece.cornell.edu (memphis.ece.cornell.edu [128.84.239.87]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84Fuaq21756 for ; Tue, 4 Sep 2001 11:56:36 -0400 (EDT) Received: from ockham.ee.cornell.edu (ockham.ee.cornell.edu [128.84.236.58]) by memphis.ece.cornell.edu (8.11.2/8.11.2) with ESMTP id f84FubV19891 for ; Tue, 4 Sep 2001 11:56:37 -0400 Date: Tue, 4 Sep 2001 11:56:37 -0400 (EDT) From: Prince Samar X-Sender: samar@ockham.ee.cornell.edu To: egs@CS.Cornell.EDU Subject: 615 PAPER 1 Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII 1. The DARPA Packet Radio Network Protocols This paper discusses the basic concepts of packet radio technology as evolved through the DARPA Packet Radio Network (PRNET) project, which has formed the foundation for the ongoing research on Mobile Ad hoc Networks (MANETs). The paper describes in detail the architecture, design, protocols, hardware capabilities and potential applications of the PRNET system. The PRNET system is designed to be a packet-switched, store-and-forward radio communication system which provides reliable network transport and datagram services, by dynamically and distributedly determining optimal routes, controlling congestion, and fairly allocating the channel irrespective of changing link conditions, varying traffic loads and mobility. The routing is done by proactively maintaining tables which determine the next hop node to reach any possible destination in the network at all times. A node broadcasts its view of the network topology to its neighbors at regular intervals, which in turn broadcast their own view - so that each node eventually has the "best" knowledge of how to route its data packets to the destination. Alternate routing can be used in case the link to the next hop node is broken. The pacing protocol provides flow and congesting control by single-threading and applying pacing function to separate the transmission to each neighbor. A not-so-strict FIFO is used to inject fairness. CSMA at the MAC layer provides multiple access, though the system is said to have the capability to use FDMA or CDMA too. FEC and CRC are used to reduce/detect errors. Direct Sequence Spread Spectrum is used for modulation, which does require known qualifiers and synchronization with the receiver. The maximum number of neighbors of a node is limited to sixteen. The PRNET is probably the first completely distributed, stand-alone and dynamic radio communication system in which nodes can move, enter and leave at their will without much affect on the performance on the system, provided that the network is not partitioned. The paper also introduces some nice optimizations like passive acks and CSMA randomization delay where needed. The table-driven approach has the advantage of having route to a destination available upon the arrival of a data packet and hence latency is very low. But at the same time, a lot of control traffic is never used by a node, wasting the network capacity and reducing efficiency. Also, each node transmitting its own perspective of the whole network topology is not the best way to perform proactive routing - the control traffic increases exponentially as the nodes in the network increases, severely limiting the scalability of the system (a maximum 138 entities). For example, link-state broadcast based routing (OLSR, FSR etc.) would be more scalable. The Packet Radio Organization Packet (PROP) is broadcasted every 7.5s. An update period based on the mobility of the PRs in the network would have been better suited to support a range of mobility in the network rather than this sacred number which restricts the kind of mobility that the network can support. The big PROP messages and acknowledgments can cause rapid drainage of a node's battery, thereby affecting the life of the whole network. The shortest route with good links is chosen, though it may not be the one with least congestion or nodes with high battery life. Alternate routing may not work if the destination has moved from its neighborhood until it broadcasts new PROP message. Also, the protocol seems to assume bidirectional links - a node sends out three messages before it leans to Alternate Routing for help. Convergence to new, stable routes may be slow after dynamic changes in the network. There are absolutely no simulation/performance figures in the paper, making it harder to believe some of the claims that they make. From c.tavoularis@utoronto.ca Tue Sep 4 12:03:35 2001 Return-Path: Received: from bureau6.utcc.utoronto.ca (bureau6.utcc.utoronto.ca [128.100.132.16]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84G3Zq22386 for ; Tue, 4 Sep 2001 12:03:35 -0400 (EDT) Received: from webmail3.ns.utoronto.ca ([128.100.132.26] EHLO webmail3 ident: IDENT-NOT-QUERIED [port 47381]) by bureau6.utcc.utoronto.ca with ESMTP id <241865-13389>; Tue, 4 Sep 2001 12:01:29 -0400 Received: by webmail3.ns.utoronto.ca id <414676-5759>; Tue, 4 Sep 2001 12:00:09 -0400 To: COM S 615 Subject: 615 PAPER 1 Message-ID: <999619209.3b94fa89a882c@webmail.utoronto.ca> Date: Tue, 04 Sep 2001 12:00:09 -0400 (EDT) From: c.tavoularis@utoronto.ca MIME-Version: 1.0 Content-Type: text/plain; charset=ISO-8859-1 Content-Transfer-Encoding: 8bit User-Agent: IMP/PHP IMAP webmail program 2.2.3 Summary Paper #1 – Christina Tavoularis This paper describes the progress of the DARPA packet radio network (PRNET) in 1987. PRNET was developed to provide communication between computers in a wireless and mobile environment, such that a device can be connected to a packet radio (PR) to communicate as part of the network. All levels of the implementation of PRNET are described including routing, error control, modulation and hardware. PRNET successfully automates the management of a medium-sized (up to 138 entities), dynamic, wireless network and can adapt to changes in the network. PRNET exploits the broadcasting capabilities of wireless communications by transmitting packets to all PRs within one hop. It queries the status of the network topology by having each PR periodically send Packet Radio Organization Packets (PROPs) to its neighbors. An advantage of combining a device with a PR is that a device will only need to know the ID of the other device to communicate with it, no matter which PR it is connected to, or where it is physically located. All routing is concealed from the device and user. A PR routes information by maintaining facts on the network topology that it has gathered from receiving PROPs. A PR will keep track of its neighbors (within 1 hop) and paths to other PRs and devices, including all hops and corresponding tiers. It will even associate a link quality between PRs, and will label a link as ‘bad’ if communication severely deteriorates between two PRs. However, the calculation of the link quality seems to be too complex. The performance of this system is slow compared to networks today. Periodic updates every 7.5 seconds would not be able to keep up with fast moving PRs in the network. PRNET avoids the issue of network partitioning by ensuring that each PR has at least one line-of-sight path to another PR in the network. Overall, issues mentioned in this article such as congestion in densely packed sub-networks, and network partitioning, are still valid today. From gleason@CS.Cornell.EDU Tue Sep 4 12:06:12 2001 Return-Path: Received: from postoffice.mail.cornell.edu (postoffice.mail.cornell.edu [132.236.56.7]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84G6Bq22559 for ; Tue, 4 Sep 2001 12:06:11 -0400 (EDT) Received: from cypher.cs.cornell.edu (dhcp-147.rover.cornell.edu [128.84.24.147]) by postoffice.mail.cornell.edu (8.9.3/8.9.3) with ESMTP id MAA03983 for ; Tue, 4 Sep 2001 12:06:11 -0400 (EDT) Message-Id: <5.0.2.1.2.20010904083706.00a774a8@postoffice.mail.cornell.edu> X-Sender: gleason@pop.cs.cornell.edu (Unverified) X-Mailer: QUALCOMM Windows Eudora Version 5.0.2 Date: Tue, 04 Sep 2001 12:05:12 -0400 To: egs@CS.Cornell.EDU From: Sunny Gleason Subject: 615 Paper # 1 Mime-Version: 1.0 Content-Type: text/plain; charset="us-ascii"; format=flowed The DARPA Packet Radio Network Protocols The PRNET paper presents a fairly extensive description of the operation of the DARPA Packet Radio Network. The authors discuss a variety of issues, ranging from the physical layer (i.e. RF modulation) to the transport layer (i.e. packet delivery/QoS). I was the most interested in the performance limitations of the following three design decisions: the use of CSMA to detect collisions, the active broadcast/computation of routing tables, and the use of the fair queuing algorithm. The authors do not really discuss the tradeoffs inherent in the use of CSMA to detect collisions. As stated in "MACAW: A media access protocol for wireless LANS" (Bharghavan et al, 1994), CSMA is useful to detect collisions on the sender side, not the receiver. Using RTS/CTS packets seems to be a more efficient way of avoiding collisions at the receiver end. The active broadcast of routing tables seems to have inherent scalability limitations. Specifically, as the number (density) of nodes increases, the overhead increases superlinearly in the number of nodes (or linear in the number of pairs); although I might have missed a spot where they talked about optimizing the broadcast routing tables. In addition, the fair queuing algorithm would seem to imply drastic variances in the perceived speed of the network for each node, even when the nodes are within line of sight. Depending on the number of nodes (ignoring noise) the perceived bandwidth could range between 400kbps [for one node] and 2kbps (1/50 * 100kbps) [for 50 nodes]. I found the lack of performance measurements somewhat disappointing. The ideas set forth in the paper seem intuitive, but without benchmarks, it is most difficult to make any kind of conclusions about the robustness or scalability of the system. Despite these limitations, however, the collection of ideas set forth in this document are remarkable as one of the first solid implementations of ad-hoc networking. From jcb35@cornell.edu Tue Sep 4 13:47:33 2001 Return-Path: Received: from travelers.mail.cornell.edu (travelers.mail.cornell.edu [132.236.56.13]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84HlXq03801 for ; Tue, 4 Sep 2001 13:47:33 -0400 (EDT) Received: from travelers.mail.cornell.edu (travelers.mail.cornell.edu [132.236.56.13]) by travelers.mail.cornell.edu (8.9.3/8.9.3) with SMTP id NAA02216 for ; Tue, 4 Sep 2001 13:47:31 -0400 (EDT) From: jcb35@cornell.edu Date: Tue, 4 Sep 2001 13:47:31 -0400 (EDT) X-Sender: jcb35@travelers.mail.cornell.edu To: egs@CS.Cornell.EDU Subject: 615 PAPER 1 Message-ID: MIME-Version: 1.0 Content-Type: TEXT/PLAIN; charset=US-ASCII This paper describes the implementation of a PR (Packet Radio) network (called PRnet) in detail which outlined the idea of a network set up over a wireless connection. The paper details the hardware and link level operations as well as routing protocols and uses of a PRNet system. The paper states that the radio network they designed was meant for at most 138 entities in mind, but larger networks can be easily created by binding PRnets together with gateways. But for todays network, this system would not scale so well - Periodic updates at about every 7-8 seconds would not necessarily be quick enough to keep track of changing topology. The network was set up with the packet radios which are used by devices commented to those PRs. At a low level, the PR use CSMA to detect collisions with random delays imposed once a channel becomes free and backoffs in case of collisions. Since the paper, many other protocols have been developed that solve problems with CSMA such as MACAW. Routing is done through proactive means - The PRs send out radio on packets periodically to establish peer connectivity. The receivers then can compute what they perceive as the network topology. Each PR keeps three routing tables - one of the neighbors, one for other PRs one hop away, and another that keeps track of which devices are on what PR addresses. To send packets out, they construct a end-to-end (ETE) header that contains the source and destination along with a type of service flag. A routing header above that contains more state for routing between the source and destination, including PR ids which have transmitted the packet and other information. Comments: This paper looked at many central problems of ad-hoc networks and produced a working system with a medium number of nodes. It detailed every level of the solution from the link-level to the routing level. In some places, however, it did not anticipate some things that are central to ad-hoc networks today such as scalability (they only designed the system for 138 devices, and each node kept track of the whole network topology), security, and power conservation. From egs@CS.Cornell.EDU Tue Sep 4 14:47:12 2001 Return-Path: Received: from hoho.cs.cornell.edu (hoho.cs.cornell.edu [128.84.96.89]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84IlCq11144 for ; Tue, 4 Sep 2001 14:47:12 -0400 (EDT) From: Emin Gun Sirer Received: (from egs@localhost) by hoho.cs.cornell.edu (8.11.3/8.11.3/C-3.2) id f84IlAe22023 for egs; Tue, 4 Sep 2001 14:47:10 -0400 (EDT) Date: Tue, 4 Sep 2001 14:47:10 -0400 (EDT) Message-Id: <200109041847.f84IlAe22023@hoho.cs.cornell.edu> To: egs@CS.Cornell.EDU Subject: 615 PAPER 1 >From andre@CS.Cornell.EDU Tue Sep 4 11:47:53 2001 Date: Tue, 4 Sep 2001 11:45:50 -0500 From: =?iso-8859-1?Q?Andr=E9?= Allavena To: Emin Gun Sirer Subject: CS615 Summaries Content-Disposition: inline User-Agent: Mutt/1.3.20i DARPA IN this protocol, the nodes broadcast all their routes to everybody every 7.5s. There are different tables kept in memory: a list of best routes, the quality of the link with the neighbours, and the association node/device. Limitation to 16 neighbours and 138 nodes in that implementation. Why not dynamically change the 7.5s If there are no movement, this can be decreased with no harm, and increased again when there are things to say. IN both protocols there are ACK, and it seems they are waiting for the ACK before sending a new packet. What about a sliding windows as in TCP? -- Andri Allavena (local) 154 A Valentine Place Icole Centrale Paris (France) Ithaca NY 14850 USA Cornell University (NY) (permanent) 879 Route de Beausoleil PhD in Computer Science 06320 La Turbie FRANCE From egs@CS.Cornell.EDU Tue Sep 4 14:48:04 2001 Return-Path: Received: from hoho.cs.cornell.edu (hoho.cs.cornell.edu [128.84.96.89]) by sundial.cs.cornell.edu (8.11.3/8.11.3/M-3.7) with ESMTP id f84Im3q11201 for ; Tue, 4 Sep 2001 14:48:03 -0400 (EDT) From: Emin Gun Sirer Received: (from egs@localhost) by hoho.cs.cornell.edu (8.11.3/8.11.3/C-3.2) id f84Im2n22062 for egs; Tue, 4 Sep 2001 14:48:02 -0400 (EDT) Date: Tue, 4 Sep 2001 14:48:02 -0400 (EDT) Message-Id: <200109041848.f84Im2n22062@hoho.cs.cornell.edu> To: egs@CS.Cornell.EDU Subject: 615 PAPER 1 >From kewang@CS.Cornell.EDU Tue Sep 4 12:29:34 2001 content-class: urn:content-classes:message X-MimeOLE: Produced By Microsoft Exchange V6.0.4712.0 Subject: review for PRNET Date: Tue, 4 Sep 2001 12:29:35 -0400 X-MS-Has-Attach: X-MS-TNEF-Correlator: Thread-Topic: review for PRNET Thread-Index: AcE1XsfXqvYtIBXcTe2kSb0Js+3HEQ== From: "Ke Wang" To: "Emin Gun Sirer" X-MIME-Autoconverted: from quoted-printable to 8bit by sundial.cs.cornell.edu id f84GTXq25048 Review for " the DARPA packet radio network protocols", by J. Jubin and J. Tornow This paper introduces the DARPA packet radio network (PRNET) in detail, which is thought to be the origin of nowadays ad hoc network. In the paper the authors give description of the whole system, the algorithms for dynamic routing, controlling congestion, fairly allocating channels in PRNET. This is almost the earliest one considering the proactive or table-driven routing protocol in wireless network. The routing algorithm in PRNET is table-driven with an underlying routing table update mechanism that involves constant propagation of routing information. Each packet radio keeps three tables: neighbor table, tier table and device table. PRs can independently decide how to route upon the network initialization and update automatically as the topology changes. Each PR will broadcast a Packet Radio Organization Packet (PROP) every 7.5s to make the information about network topology up-to-date. Because routing information is constantly broadcast, the route to every node in PRNET is always available, without waiting to establish it when a node needs to use. And the tables only save the information about next hop instead of the whole route, so the memory overhead is not large. These are advantages of this routing protocol. However, this kind of table-driven mechanism will incur considerable bandwidth and power consumption, which is scarce in current ad hoc networks. One thought about this routing protocol is why not combine the three tables into one? And broadcasting every 7.5s is fast enough to reflect the change of the whole network as the size of the ad hoc network grows? To save bandwidth and battery power, maybe a better choice is combining the table-driven routing and on-demand routing together. Use on-demand within a smaller region, while using table-driven protocol to connect them in the large scale.