What is MANET or why WiFi is not the solution to all telecommunication problems
Having read the next post on Habré on the topic of self-organizing networks, the author realized that there is still no clarity on this topic. In most cases, self-organizing radio networks are said to be “mesh”, “ad hoc”, “mobile ad hoc”, etc., implying that this is the same thing. And this applies not only to a simple Habro-philistine, even if technically savvy, but also to most engineers with a PhD. Moreover, such networks are often presented as a universal tool for all occasions [1] or even as a little or no single vector of development and evolution of the telecom industry [2]. I can assure you that without a clear understanding of the place and role of these networks, such statements are at least premature, although not far from reality.
So let's start by defining what networks like Mesh, Ad Hoc, mobile Ad Hoc are and what is the difference between them.
Mesh networks are mesh radio networks consisting of fixed wireless routers that create a wireless backbone and subscriber service area) and mobile / fixed subscribers having access (within a radio communications zone) to one of the routers. The topology is a star, with a random connection of support nodes.
Ad hoc networks - radio networks with random stationary subscribers that implement fully decentralized control in the absence of base stations or reference nodes. Topology - fixed with random connection of nodes.
MANET (Mobile Ad hoc NETworks) networks are radio networks with random mobile subscribers that implement fully decentralized control in the absence of base stations or reference nodes. Topology - rapidly changing with random connection of nodes.
To this we must add the WSN (Wireless Sensor Networks) - wireless sensor (telemetry) networks consisting of compact sensor nodes with integrated functions for monitoring certain environmental parameters, processing and transmitting data via radio channels. Depending on the task, they can be built as a mesh, ad hoc or MANET topology; vehicle networks VANET (Vehicle Ad hoc NETworks) - vehicle communication networks; and all kinds of hybrids of the above.
')
Today, the vast majority of terrestrial mobile wireless communication networks have a fixed infrastructure and are interconnected via various, usually wired or radio-relay, data transmission channels. In the last decade, great attention has been paid to the creation of mobile packet radio networks that do not have a fixed infrastructure — the fixed network (Ad Hoc) and mobile subscribers (MANET).
Such networks are self-organizing, since their nodes are not only end user terminals, but also are relay routers, relaying packets of other subscribers and participating in finding routes to them, therefore, these networks are capable of self-organization. Such networks can consist of tens, hundreds and even thousands of nodes. The scope of such networks is quite wide. For example, MANET networks are useful in search and rescue operations, at a tactical-level theater of operations, in crowded places (for example, to serve conference participants), and where there is no telecommunications infrastructure (for example, in expeditions to distant regions). ). It is possible that these networks in many cases can become an alternative to mass mobile communication networks.
In contrast to networks with a hierarchical structure and centralized management, peer-to-peer networks without infrastructure consist of nodes of the same type, where each node has a set of software and hardware tools that allow organizing the transfer of data from the source to the receiver directly in the physical presence of such a path and thereby distribute the load on the network and increase total network bandwidth. Data transmission from one subscriber to another can occur, even if these nodes are outside the direct radio visibility zone. In these cases, the data packets of these subscribers are relayed by other network nodes that have a connection with the corresponding subscribers. Networks with multiple retransmissions are called multi-hop or multi-hop (multihop). When developing such networks, the main problems are packet routing from the source node to the destination node, network scalability, endpoint addressing, maintaining connectivity in a variable topology.
Thus, the practical benefits of the implementation of such networks would be really huge. Starting from free calls within such a network to the restoration of communication in areas destroyed by the elements.
Most scientific papers and commercial projects (projects ) imply the use of IEEE 802.11 standards, popularly known as WiFi, IEEE 802.15.4 (ZigBee) or IEEE 802.15.1 (Bluetooth). This almost exhausts the proposed radio technologies on this topic. This non-diversity of approaches is explained simply: we already have WiFi and Bluetooth in every smartphone, is it not better to reuse them for the tasks of self-organization of radio networks. At the same time, a beautiful picture is drawn, as in [3], and the advantages of using them in the script are explained ... MANET.
Here is the "buried dog." The fact is that neither IEEE 802.15.4, and even less IEEE 802.11 were developed for MANET networks. They were designed for mesh networks (see definition above) and have a hierarchical design principle. The same ZigBee standard implies 3 types of devices on the network: coordinator, router (router) and end device. So what do you say? But the fact is that even if you have 2 end ZigBee devices on the table, say a floor lamp and a kettle , they will never “talk” directly with each other, but only through a router. In addition, the new device will never enter the network without a preliminary "OK" and a heap of other parameters from the coordinator. All this takes time, during which the user just waits. This is especially important for networks with a rapidly changing topology, such as VANET.
The same applies to Bluetooth technology, where there is always one coordinator who distributes time slots and controls access to the channel. Subscribers, as in ZigBee, always communicate through the coordinator. Coordinators can form scatternet among themselves, but in reality this does not work (see further why). Bluetooth 4.0, known as Bluetooth Low Energy, is designed to transmit, once per hour, packets of a maximum length of 20 bytes of application data [4]. So it is not physically suitable for high-speed services.
Now imagine a car that rushes along a highway and it needs to exchange information with a traffic light and a passing car and compare it with the time you spend entering a regular WiFi network or connecting via Bluetooth to a new device. Compare? Plus, the connection must be secure (with delivery confirmation), secure (encryption + authentication), and fast (bandwidth preferably> 1 Mbit / s).
Or another example - MANET. Imagine a soldier on the battlefield, where the situation changes every second, and he needs to report on the situation to the command in real time, get an order, load a tactical map, and so on. Such a connection, in addition to reliability and security, must also be resistant to topology changes, routing must have fast convergence, i.e. ensure that the route of a given quality is found in a reasonable time, ensure that there are no loops, and ensure multicast delivery. And if there are many such soldiers. Say company or battalion?
Now you can imagine what kind of problems are facing these networks.
So after all, why, with all the above requirements, you can not create a hybrid or fully distributed radio network type MANET on the standard IEEE 802.11 or ZigBee or Bluetooth? Can! But let's be honest, what this network will be.
1. MAC level
The most important for the efficient operation of packet-switched radio networks is the data link layer, more precisely, its MAC sublayer due to its conceptual complexity and global network influence, since the irrational organization of collective access to the radio channel can significantly reduce the speed of packet transmission over the network or even completely block its operation regardless of the quality of operation of other levels of the OSI reference model.
WiFi, like ZigBee, uses the Carrier Sense Multiple Access Protocol with Collision Avoidance - CSMA / CA. He suggests that "one says the rest are silent." In this case, the RTS / CTS control frames are exchanged to solve the hidden subscriber problem and the transmission medium is reserved for the transmitting station. I will not bore you with technical details on how this is done - who wants to know, there are tons of information on this topic on the Internet. Let me just say that reserving the environment using the CSMA / CA method requires strict symmetry of lines and certain elements of coordinated management, which is impossible / undesirable for MANET.
In summary, we have a degradation of performance parameters of the multiple access protocol with a high dynamics of topology changes and heavy traffic.
2. Addressing
Now only the lazy does not know that the IPv4 address space has been exhausted.
It is clear that raising DHCP and distributing addresses within the MANET network is an unsustainable idea, if only because it will take time to find the route to the server, and how to route the initial 0.0.0.0 address during a multi-hop packet retransmission?
It is believed that IPv6 is the protocol that will form the basis of the networks of the future. We agree with this. But then there is some confusion about the lack of universal support for the protocol by the routing developers. Even the fundamental ad hoc documents of IETF, such as RFC 3561, RFC 4728, do not offer specific IPv6 support mechanisms. Thus, this problem is left to the vendors of iron, and they in turn solve the problem as they can.
Another unpleasant fact for WiFi, in the context of self-organizing networks, is mandatory addressing at the data link level. It seems a trifle, but as practice shows, this trifle is capable of putting the whole network. I explain. If we are working in an IPv4 network, the channel frame header is formed using the ARP protocol, which determines the MAC address of the called party by its IP by periodically polling. In IP protocol version 6, ARP protocol is not. It is replaced by ICMP version 6, which involves the exchange of special messages "Neighbor solicitation" - "Neighbor Advertisement" to bind the MAC address to IPv6. Naturally, in the classic LAN these requests do not go beyond 1 router, since there all users are supposed to sit on one common bus. In the radio networks, due to their wireless nature, all subscribers cannot physically sit on the common channel and hear everyone else, especially in MANET. And filling the ARP or ICMPv6 network with requests leads to an increase in non-informative exchange between subscribers and, as a result, to a decrease in the actual throughput.
As you can see, from scratch they got the need to exchange channel frames between neighboring subscribers of the network.
3. Routing
It is customary to divide the routing protocols into proactive (tabular), reactive (probe) and their hybrids (there are still wave, but about them another time). At the dawn of the development of mesh networks, they tried to use the standard OSPF routing protocol. Nothing came of this, naturally, since It was designed for completely different operating conditions. As a result, a lot of work appeared, both scientific and not so much, where dozens of routing protocols for self-organizing radio networks are proposed. Who is interested in this topic, I advise the site www.ieee.org , section publications. The problem, however, lies in the fact that the actually developed routing protocols of the MANET networks are either not physically implemented in the C language, or are oriented towards achieving optimal use of network resources under quasi-static network conditions, when the topology changes slowly or does not change at all. The latter is relevant for table-oriented protocols such as OLSR, DSDV, WRP, BATMAN, Babel, etc.
Again, why is all this necessary in such detail? This is necessary because these protocols imply the presence in the route tables of information about the entire network at once. In short, they constantly build routes to all nodes they know, regardless of whether we need this subscriber or not. Got the second need to generate non-informative traffic on the network.
Probe protocols, as for example. AODV, DSR, SSR, TORA, suggest routing on demand, but not fully standardized (see above addressing problem). In addition, due to asymmetric channels, routing must support the construction of multiple routes both from the destination to the recipient and in the opposite direction. And this is only supported by the DSR and TORA protocols. Therefore, there is still a free field for the free thought of the developer. A developer is different ...
4. Transport level
The author, by the nature of his scientific work, happened to work on the adaptation of the TCP and the implementation of the new transport protocol for MANET. You can write on this topic very long and tedious. Therefore, I will give only the main conclusions from my experience. Standard TCP, in the implementation of RFC 793 and RFC 5681, does not work well with multi-hop relay and random access. The network is either underloaded or overloaded, this is due not to the absence of the actual bandwidth, but to large fluctuations in the parameters of the connection between the subscribers: delay, jitter, percentage of packet loss; and especially frequent changes in the actual transmission route. Moreover, it turned out that TCP should have access to lower levels - network and channel for a more adequate response to changes in such a network. And the concept of cross-layer exists only in articles of various smart people and nobody has been engaged in its practical implementation. Almost nobody. What to do with it? There is no other answer except RnD.
Now, with all this in mind, what off-the-shelf radio technologies can immediately meet all the requirements and cope with the problems described above? The answer is obvious - no. Conclusion: you can create a real self-organizing network like MANET on WiFi, but it will not work very well, to put it mildly. Exactly why we do not see such a boom in wireless p2p, as in the wired Internet, where there is Skype, TOR, torrents and many other interesting things.
I emphasize that this is not about sensor networks that generate 1 packet per hour, but about high-speed decentralized radio networks with dynamically changing topology, for example, for voice transmission.
But let's not exaggerate, something has been done for real MANET networks. We will also not engage in advertising, especially since it is prohibited by the rules of the habr-community. Let's just say that a certain team developed a software and hardware solution for embedded systems in the form of a radio module that allows you to organize access at speeds up to 2 Mbit / s (about 700 kbit / s in the application) in fully decentralized mobile radio networks that support reactive routing and protocol IPv6. Now the development of a developer kit is being completed, which includes from 3 radio modules with interface cards to which various sensors and other sources of information can be connected, a debugging board, a MANET-Ethernet gateway and a user's network management and configuration software. Based on this solution, it is also possible to implement a user terminal in the format of a secure smartphone or USB modem.
In this regard, I would like to conduct a small survey on what service would be most interesting and useful to users.
Thank you for your attention and I will be grateful for the feedback!
[1] - www.russianelectronics.ru/leader-r/review/doc/59295
[2] - habrahabr.ru/company/infomania/blog/102814
[3] - habrahabr.ru/post/79360
[4] - e2e.ti.com/support/low_power_rf/f/538/t/71381.aspx
So let's start by defining what networks like Mesh, Ad Hoc, mobile Ad Hoc are and what is the difference between them.
Mesh networks are mesh radio networks consisting of fixed wireless routers that create a wireless backbone and subscriber service area) and mobile / fixed subscribers having access (within a radio communications zone) to one of the routers. The topology is a star, with a random connection of support nodes.
Ad hoc networks - radio networks with random stationary subscribers that implement fully decentralized control in the absence of base stations or reference nodes. Topology - fixed with random connection of nodes.
MANET (Mobile Ad hoc NETworks) networks are radio networks with random mobile subscribers that implement fully decentralized control in the absence of base stations or reference nodes. Topology - rapidly changing with random connection of nodes.
To this we must add the WSN (Wireless Sensor Networks) - wireless sensor (telemetry) networks consisting of compact sensor nodes with integrated functions for monitoring certain environmental parameters, processing and transmitting data via radio channels. Depending on the task, they can be built as a mesh, ad hoc or MANET topology; vehicle networks VANET (Vehicle Ad hoc NETworks) - vehicle communication networks; and all kinds of hybrids of the above.
')
Why precisely MANET?
Today, the vast majority of terrestrial mobile wireless communication networks have a fixed infrastructure and are interconnected via various, usually wired or radio-relay, data transmission channels. In the last decade, great attention has been paid to the creation of mobile packet radio networks that do not have a fixed infrastructure — the fixed network (Ad Hoc) and mobile subscribers (MANET).
Such networks are self-organizing, since their nodes are not only end user terminals, but also are relay routers, relaying packets of other subscribers and participating in finding routes to them, therefore, these networks are capable of self-organization. Such networks can consist of tens, hundreds and even thousands of nodes. The scope of such networks is quite wide. For example, MANET networks are useful in search and rescue operations, at a tactical-level theater of operations, in crowded places (for example, to serve conference participants), and where there is no telecommunications infrastructure (for example, in expeditions to distant regions). ). It is possible that these networks in many cases can become an alternative to mass mobile communication networks.
In contrast to networks with a hierarchical structure and centralized management, peer-to-peer networks without infrastructure consist of nodes of the same type, where each node has a set of software and hardware tools that allow organizing the transfer of data from the source to the receiver directly in the physical presence of such a path and thereby distribute the load on the network and increase total network bandwidth. Data transmission from one subscriber to another can occur, even if these nodes are outside the direct radio visibility zone. In these cases, the data packets of these subscribers are relayed by other network nodes that have a connection with the corresponding subscribers. Networks with multiple retransmissions are called multi-hop or multi-hop (multihop). When developing such networks, the main problems are packet routing from the source node to the destination node, network scalability, endpoint addressing, maintaining connectivity in a variable topology.
Thus, the practical benefits of the implementation of such networks would be really huge. Starting from free calls within such a network to the restoration of communication in areas destroyed by the elements.
About technologies and standards
Most scientific papers and commercial projects (
Here is the "buried dog." The fact is that neither IEEE 802.15.4, and even less IEEE 802.11 were developed for MANET networks. They were designed for mesh networks (see definition above) and have a hierarchical design principle. The same ZigBee standard implies 3 types of devices on the network: coordinator, router (router) and end device. So what do you say? But the fact is that even if you have 2 end ZigBee devices on the table, say a floor lamp and a kettle , they will never “talk” directly with each other, but only through a router. In addition, the new device will never enter the network without a preliminary "OK" and a heap of other parameters from the coordinator. All this takes time, during which the user just waits. This is especially important for networks with a rapidly changing topology, such as VANET.
The same applies to Bluetooth technology, where there is always one coordinator who distributes time slots and controls access to the channel. Subscribers, as in ZigBee, always communicate through the coordinator. Coordinators can form scatternet among themselves, but in reality this does not work (see further why). Bluetooth 4.0, known as Bluetooth Low Energy, is designed to transmit, once per hour, packets of a maximum length of 20 bytes of application data [4]. So it is not physically suitable for high-speed services.
About problems
Now imagine a car that rushes along a highway and it needs to exchange information with a traffic light and a passing car and compare it with the time you spend entering a regular WiFi network or connecting via Bluetooth to a new device. Compare? Plus, the connection must be secure (with delivery confirmation), secure (encryption + authentication), and fast (bandwidth preferably> 1 Mbit / s).
Or another example - MANET. Imagine a soldier on the battlefield, where the situation changes every second, and he needs to report on the situation to the command in real time, get an order, load a tactical map, and so on. Such a connection, in addition to reliability and security, must also be resistant to topology changes, routing must have fast convergence, i.e. ensure that the route of a given quality is found in a reasonable time, ensure that there are no loops, and ensure multicast delivery. And if there are many such soldiers. Say company or battalion?
Now you can imagine what kind of problems are facing these networks.
So after all, why, with all the above requirements, you can not create a hybrid or fully distributed radio network type MANET on the standard IEEE 802.11 or ZigBee or Bluetooth? Can! But let's be honest, what this network will be.
1. MAC level
The most important for the efficient operation of packet-switched radio networks is the data link layer, more precisely, its MAC sublayer due to its conceptual complexity and global network influence, since the irrational organization of collective access to the radio channel can significantly reduce the speed of packet transmission over the network or even completely block its operation regardless of the quality of operation of other levels of the OSI reference model.
WiFi, like ZigBee, uses the Carrier Sense Multiple Access Protocol with Collision Avoidance - CSMA / CA. He suggests that "one says the rest are silent." In this case, the RTS / CTS control frames are exchanged to solve the hidden subscriber problem and the transmission medium is reserved for the transmitting station. I will not bore you with technical details on how this is done - who wants to know, there are tons of information on this topic on the Internet. Let me just say that reserving the environment using the CSMA / CA method requires strict symmetry of lines and certain elements of coordinated management, which is impossible / undesirable for MANET.
In summary, we have a degradation of performance parameters of the multiple access protocol with a high dynamics of topology changes and heavy traffic.
2. Addressing
Now only the lazy does not know that the IPv4 address space has been exhausted.
It is clear that raising DHCP and distributing addresses within the MANET network is an unsustainable idea, if only because it will take time to find the route to the server, and how to route the initial 0.0.0.0 address during a multi-hop packet retransmission?
It is believed that IPv6 is the protocol that will form the basis of the networks of the future. We agree with this. But then there is some confusion about the lack of universal support for the protocol by the routing developers. Even the fundamental ad hoc documents of IETF, such as RFC 3561, RFC 4728, do not offer specific IPv6 support mechanisms. Thus, this problem is left to the vendors of iron, and they in turn solve the problem as they can.
Another unpleasant fact for WiFi, in the context of self-organizing networks, is mandatory addressing at the data link level. It seems a trifle, but as practice shows, this trifle is capable of putting the whole network. I explain. If we are working in an IPv4 network, the channel frame header is formed using the ARP protocol, which determines the MAC address of the called party by its IP by periodically polling. In IP protocol version 6, ARP protocol is not. It is replaced by ICMP version 6, which involves the exchange of special messages "Neighbor solicitation" - "Neighbor Advertisement" to bind the MAC address to IPv6. Naturally, in the classic LAN these requests do not go beyond 1 router, since there all users are supposed to sit on one common bus. In the radio networks, due to their wireless nature, all subscribers cannot physically sit on the common channel and hear everyone else, especially in MANET. And filling the ARP or ICMPv6 network with requests leads to an increase in non-informative exchange between subscribers and, as a result, to a decrease in the actual throughput.
As you can see, from scratch they got the need to exchange channel frames between neighboring subscribers of the network.
3. Routing
It is customary to divide the routing protocols into proactive (tabular), reactive (probe) and their hybrids (there are still wave, but about them another time). At the dawn of the development of mesh networks, they tried to use the standard OSPF routing protocol. Nothing came of this, naturally, since It was designed for completely different operating conditions. As a result, a lot of work appeared, both scientific and not so much, where dozens of routing protocols for self-organizing radio networks are proposed. Who is interested in this topic, I advise the site www.ieee.org , section publications. The problem, however, lies in the fact that the actually developed routing protocols of the MANET networks are either not physically implemented in the C language, or are oriented towards achieving optimal use of network resources under quasi-static network conditions, when the topology changes slowly or does not change at all. The latter is relevant for table-oriented protocols such as OLSR, DSDV, WRP, BATMAN, Babel, etc.
Again, why is all this necessary in such detail? This is necessary because these protocols imply the presence in the route tables of information about the entire network at once. In short, they constantly build routes to all nodes they know, regardless of whether we need this subscriber or not. Got the second need to generate non-informative traffic on the network.
Probe protocols, as for example. AODV, DSR, SSR, TORA, suggest routing on demand, but not fully standardized (see above addressing problem). In addition, due to asymmetric channels, routing must support the construction of multiple routes both from the destination to the recipient and in the opposite direction. And this is only supported by the DSR and TORA protocols. Therefore, there is still a free field for the free thought of the developer. A developer is different ...
4. Transport level
The author, by the nature of his scientific work, happened to work on the adaptation of the TCP and the implementation of the new transport protocol for MANET. You can write on this topic very long and tedious. Therefore, I will give only the main conclusions from my experience. Standard TCP, in the implementation of RFC 793 and RFC 5681, does not work well with multi-hop relay and random access. The network is either underloaded or overloaded, this is due not to the absence of the actual bandwidth, but to large fluctuations in the parameters of the connection between the subscribers: delay, jitter, percentage of packet loss; and especially frequent changes in the actual transmission route. Moreover, it turned out that TCP should have access to lower levels - network and channel for a more adequate response to changes in such a network. And the concept of cross-layer exists only in articles of various smart people and nobody has been engaged in its practical implementation. Almost nobody. What to do with it? There is no other answer except RnD.
Now, with all this in mind, what off-the-shelf radio technologies can immediately meet all the requirements and cope with the problems described above? The answer is obvious - no. Conclusion: you can create a real self-organizing network like MANET on WiFi, but it will not work very well, to put it mildly. Exactly why we do not see such a boom in wireless p2p, as in the wired Internet, where there is Skype, TOR, torrents and many other interesting things.
I emphasize that this is not about sensor networks that generate 1 packet per hour, but about high-speed decentralized radio networks with dynamically changing topology, for example, for voice transmission.
About prospects
But let's not exaggerate, something has been done for real MANET networks. We will also not engage in advertising, especially since it is prohibited by the rules of the habr-community. Let's just say that a certain team developed a software and hardware solution for embedded systems in the form of a radio module that allows you to organize access at speeds up to 2 Mbit / s (about 700 kbit / s in the application) in fully decentralized mobile radio networks that support reactive routing and protocol IPv6. Now the development of a developer kit is being completed, which includes from 3 radio modules with interface cards to which various sensors and other sources of information can be connected, a debugging board, a MANET-Ethernet gateway and a user's network management and configuration software. Based on this solution, it is also possible to implement a user terminal in the format of a secure smartphone or USB modem.
In this regard, I would like to conduct a small survey on what service would be most interesting and useful to users.
Thank you for your attention and I will be grateful for the feedback!
[1] - www.russianelectronics.ru/leader-r/review/doc/59295
[2] - habrahabr.ru/company/infomania/blog/102814
[3] - habrahabr.ru/post/79360
[4] - e2e.ti.com/support/low_power_rf/f/538/t/71381.aspx
Source: https://habr.com/ru/post/197860/
All Articles