INTRODUCTION

The availability of what can be termed an ideal MAC (Medium Access Control) for sharing a common communications channel suggests that virtually every system extant that required three or more entities to communicate could be redesigned for better performance and/or economics. These entities would include ICs, boards, backplanes, LANs, MANs, WANs, WiFI, cell communications, satellite networks plus everything under the IoT (Internet of Things) umbrella.  There is in fact a MAC that comes close enough to the ideal, it is based on work carried out by Prof Graham Campbell and his students at the Illinois Institute of Technology.

The seminal work on the “close-to-ideal MAC” is DQRAP, described by Xu and Campbell [1]. DQRAP supports full utilization of a channel when the offered traffic consists of fixed-size segments, e.g., ATM cells or single readings from RFID tags.  XDQRAP, Wu and Campbell [2], is a major enhancement that while still utilizing fixed-length segments supports variable length traffic, e.g., Ethernet frames, that are segmented for transmission but most importantly with no further overhead. An important feature of DQ is the ability to support both packet and fixed-bandwidth services, this feature is described by Wu and Campbell [3].  DQ supports true priorities and is based on work described by Lin and Campbell [4].

However,  the term DQ (Distributed Queueing) is used to described implementations of this family of MACs, a term that is apropos as the main feature of this family of MACs is that the queue, instead of residing in a central physical location, i.e., a router, is distributed amongst the transmitting nodes.

Thus at least theoretically virtually all systems extant could possibly benefit from DQ. However, while the author feels that DQ should be applied in specific applications where feasible it is suggested that conversion to DQ switching should start with the backbone of the Internet, the optical fibre infrastructure.  In essence DQ would enable optical-fibre circuits to serve as distributed switches instead of acting as conduits carrying packets between routers.

The current packet switching infrastructure employs routers that in turn utilize an underlying circuit infrastructure that is an optical fibre version of the POTS (plain old telephone system) in use for almost a hundred years. POTS started by serving first an individual community, a Local Service Area (LSA), with voice service from a central office by connecting each of the hundreds or thousands of local customers via copper wires in a star topology. The individual copper lines ofttimes had multiple customers on a single line, a “party line”.

The invention of the electron tube enabled the interconnection of LSAs on a country-wide basis. Direct-dialing eliminated manual operators and solid-state logic led to digital voice circuits and the introduction of computers. For the first time entities, albeit not human, could transmit meaningful amounts of information in microseconds, challenging a circuit-switching infrastructure designed to switch circuits that were in use for minutes. The solution was the introduction of  the packet-switching router. But the existing circuit-switched infrastructure was still used to interconnect those routers, the same infrastructure that today in the main utilizes optical fibre circuits and optical switches.

Unfortunately, packet-switching is challenged by the very type of traffic that it was designed to support - a surfeit of packets leads to congestion, and in addition routers are not amenable to providing the equivalent of fixed-bandwidth circuits, the type of circuit most suitable for the delivery of streaming video.  The result is that over-engineering is necessary for effective operation.

DSA (Distributed Service Areas)

As stated above theoretically virtually all communications systems could benefit from DQ, including all WiFi and cell operations. However even though the utilization of an individual WiFi hotspot or cell tower would be increased the full benefit of DQ, i.e., support for both packet and fixed-bandwidth services would not be available till the “backbone”, i.e., the optical fibre infrastructure supporting the Internet, was converted to DQ. In addition billions of devices would have to be modified.

Therefore it is suggested that there be a return to the original POTS concept of an LSA by utilizing the existing circuit-switched infrastructure to provision Distributed Service Areas (DSAs). A DSA utilizes a modern version of the “party line” that was in use for a hundred years, modern in the sense that the MAC employs an efficient algorithm and multiplexing to serve thousands or even hundreds of thousands of customers using but a single circuit. In place of the restricted geographical area served by an LSA,  a DSA can serve a community of users distributed over thousands of Km, e.g., a business with 100 branches nationwide or even a branch of the military, say all the army bases in the country.

A DQ-based DSA utilizes broadcast/multicast to replace the mesh topology currently used by the Internet in effect moving the routing function from the Network Layer to the MAC/Data Link layer, eliminating the need for routers for that segment of the traffic and thus reducing the number of routers required to support the entire Internet. DQ  in effect augments the “rough” granularity of optical switches with a “fine” granularity that supports the switching of 64/128/256-byte segments thus enabling packets to be transported efficiently from source to destination(s) utilizing only the PHY and MAC/Data Link layers. It is similar in operation to the well known add-drop function in synchronous circuits but whereas the granularity of conventional sync add-drop is usually days, weeks or semi-permanent DQ enables as little as a 64-byte segment to be “added” or “dropped”. The router function is satisfied without a router.

The term “Distributed Ethernet” would be applicable but at some time in the future it may be decided to eliminate the Ethernet frame structure and utilize higher layer formats -- DQ supports any type frame structure.

Roadmap to DQ Switching

DQ will have universal application but a three-stage plan is proposed to implement DQ, the first two stages address the core systems that support the Internet -- the terrestrial network; the final stage addresses WiFi hotspots and the cell phone network that in turn encompass all the IoT.

These stages are:

1. DSAs - Connect all terrestrial users to one or more DSAs depending upon geographic and/or functional interest.  Inter-DSA traffic will be supported by conventional routers in this stage; intra-DSA benefits, e.g., isochronous channels and congestion-free, will not be available to inter-DSA traffic. Backhaul service can be provided to WiFi hotspots and cell towers using conventional routers.

2. Interconnect DSAs - Initially with QHubs that support disparate DSAs along a dual-bus. These are then replaced with QSwitches that support universal interconnection of DSAs. The Q-Switches, operating at the MAC/Data Link layer,  and integrated into the optical-fibre infrastructure will provide DQ benefits to all inter-DSA traffic. This stage eliminate the need for conventional routers, excepting where interfacing to wireless systems.

3. Convert all WiFi and cell service, and IoT devices to DQ.  This step allows all hot spots, towers, etc., to be seamlessly integrated into the DQ network.

Stage 1 - The DSA

It is fortunate that while DQ can be described as a disruptive technology it can be implemented in a non-disruptive manner. VPNs (Virtual Private Networks) are used to describe how and why a DSA will function. VPNs are in general use by organizations that desire to control traffic flow in/out of their own facilities.  A VPN uses software to restrain all traffic to a closed group of nodes thus achieving the economics of utilizing Internet resources while maintaining security. It is counterintuitive to suggest “shared” switching for the Internet since it is thought of as a mesh, but typically most traffic in a VPN is destined for a common point, i.e., HQ or some central data base. In addition to VPNs there are natural aggregation points in a typical packet network thus a carrier could analyze their traffic patterns and implement a DSA wherever economically viable. The aforementioned underlying synchronous network was and still to a great extent utilizes a hub-and-spoke technology thus the actual traffic flow in a network, probably upwards of 50%, could be amenable to DQ-based DSAs.   

Three figures illustrate why and how DQ switching will work in the Internet.  Figure 1 illustrates a typical packet network supporting general traffic and multiple VPNs, color coded; expand this figure to tens of thousands of routers and we have the Internet.  Figure 2 illustrates the traffic flow in just one VPN, most traffic flows between branch nodes and a central site typically following tree-and-branch topology. Figure 3 illustrates the same traffic now assigned to a DSA. Each DSA utilizes a separate physical network rather than sharing a physical circuit as do routers but there are two reasons why this does not lead to underutilization of the physical circuit:

1. A DQ-based DSA can be designed for over 90% average use -- offered traffic surges exceeding 100% of capacity cause no problems -- thus leading to greater overall utilization of the physical circuits..

2. A DSA can itself be “virtual”, i.e., central nodes of dozens of VPNs can be co-located so as to gain the advantage of utilizing a single “fat” channel rather than multiple smaller channels.

Another advantage of DQ is that once a packet enters the queue, the exact time it will reach the destination is known.

Implementing a DSA  

As mentioned above a DSA can be described as a modern version of the original shared Ethernet LAN that was implemented on a coax cable using a linear or tree-and-branch topology. The difference is that the DQ MAC is close to ideal in performance, the service area is not limited, and the circuit is a standard carrier circuit, thus standard carrier circuit interface technology is utilized. All that is required when utilizing carrier circuits is a single NIC at each node and the equivalent of an “or” gate at junction points. The goal of all circuit switching systems since the time of Bell has been to maintain separation between circuits, i.e., to create a separate physical circuit for each user. The goal of a DQ switch is the opposite, i.e., allow as many users as is possible to occupy the same physical circuit - the DQ MAC ensures that only a single frame will arrive at any junction point.

It should be pointed out that this concept is not new, it is behind all wireless communications, e.g., all signals back to the “big bang” are available to a single cell phone - time division, frequency division, filtering and encryption ensure that data is delivered to the correct destination.  All will be used by DQ.  

Extra DQ Benefits:

1. Fibre Server

A DQ-based DSA could be custom-designed to in some instances replace conventional servers.  Content is repeatedly transmitted to replicate carousel operation on the circuit. For instance:

• A DQ 1 terabit/s circuit could, every two seconds, make available the initial pages from approximately 1 million sites. (Google estimate of ~300,000 bytes per site).

• A DQ 1 petabit/s circuit could make available hundreds of the more popular videos using near Video ­on­ Demand (NVOD) that along with minimal buffering would appear as true VOD to the consumer. If the 80/20 rule applies where 80% of viewers are watching 20% of movies this would reduce bandwidth requirements for companies like NetFlix.

• Many sites could make virtually all their material available via DQ fibre server.

• Popular videos/movies could be made available in the same manner.

One of the big advantages, aside from reducing the number of satellite servers, is that it would be difficult for anyone to mount a “denial of service attack” when virtually millions of users, spread across the country, could “log in” almost simultaneously.

2. Video Services

DQ provides the option of using unicast, multicast or broadcast when transmitting the videos that are predicted to consume 80% of the capacity of the Internet.  It would be possible to monitor the tens of thousands of video streams and using “smart” broadcasting techniques to identify duplicate transmissions and collapse them into a single transmission. Even where two transmissions are not synchronized it would be possible with minimal buffering to bring them into synchronicity and then dropping one transmission.

Stage 2 (a)- Interconnecting DSAs, Cascaded Distributed Queueing (CDQ)

From the start it was understood that it was necessary for what we now term a DSA should be able to communicate with another DSA and/or the Internet. In the case of DQRAP where single segments are processed an ATM switch would be ideal.  In hindsight if DQ had been available when ATM was conceived then ATM might have been successful, in fact this would still be an interesting avenue of research given the amount of ATM assets still in use. But as it was expected that XDQRAP would dominate then it was assumed that conventional routers would be used, with usually the Ethernet frames being assembled/disassembled at a junction point.  

From the start there was an ongoing search by the author for a means of DSAs communicating with each other that would maintain the benefits of DQ, i.e., mixed circuit and packet services, no congestion, etc.  After seven years the concept of CDQ (Cascaded Distributed Queue) was developed. Whereas implementing a DSA can be accomplished with close to off-the-shelf hardware CDQ, see Chang and Campbell [5], requires new hardware, a QHub. CDQ  enables DSAs to be interconnected at the MAC/Data Link layer using these QHubs that in effect allow separate DSAs to be cascaded along a dual-bus thus eliminating the need for routers. This is accomplished by the QHub serving the dual purpose as a central hub for one DSA and as a node on a succeeding DSA.  Chang and Campbell [5] demonstrate through simulation how a series of DSAs distributed across the US, Fig 4, can be interconnected using QHubs, supporting both local and cross-country traffic. 

Simulation results show that when subject to 90% traffic load following a Poisson Distribution, with mix of traffic between coast-to-coast and that travelling two or three hops the delay at each QHub is less than 1.5 slot times. Thus no congestion or packet drop.

Basically CDQ works by classifying all incoming traffic into just two classes, local and remote. The local traffic can be controlled by modifying the conventional DQ feedback such that it can be throttled to provide priority for the remote traffic, over which there is no control of the originating station. Simulation results indicate that a buffer size of 5% of the link length suffices [5].

Stage 2 (b)- Interconnecting DSAs, The QHub and the QSwitch

Figure 4 shows how separate DSAs could be connected along a dual bus and will raise by a substantial percentage the amount of traffic that can travel between source and destination on the Internet without having to pass through a conventional router.  Universality will be accomplished with what can be described as the QSwitch that will support at least three ports, as opposed to the two ports of the QHub and thus support the mesh topology of the Internet. This will require a major research project but certain features of the QHub suggest that it could be successful.

Stage 3 -  Universal DQ Service

Countless wireless applications will have availed themselves of the benefits of DQ, there will be many instances were it will be feasible to convert a wireless operation to DQ even it would only have access to a DSA.  But now the remainder of such devices, numbering in the billions, can be converted. It will be a gradual replacement but possibly when it becomes apparent that DQ will dominate individual devices will be programmed with both the existing MAC and DQ thus simplifying and speeding up the conversion.

Observations

Stage 1 offers no real challenges to implementation, there are only three components (1) existing circuits, (2) a standard Ethernet NIC that contains the DQ algorithm, which incidentally in its simplest form is implemented with a four-state machine plus two binary counters, and  (3) the DQ algorithm. The reader is referred to Laya, Kalalas, et al [6] where the DQ algorithms are compared with virtually all other MACs.

Stage 2 requires probably two more years of research to get CDQ and the QHub to the stage where it can be implemented.  Several more years of research will be required to develop the QSwitch. However if the techniques used in CDQ can be used for the QSwitch then that time will be reduced.

Stage 3 will commence almost immediately but in a piecemeal manner where it develops that the availability of just a DSA justifies the use of DQ in the user application.

Conclusions

Bell introduced circuit switching in the 1870s,  packet switching was introduced via the ArpaNet in the 1970s.  Each provided a solution for certain types of traffic so the Internet today is supported by essentially two distinct networks, circuit and packet, operating in conjunction, but with the problems as described above due in the main to the interface between the two systems.  DQ, including the proposed QSwitch, in effect a hybrid of the two systems, overcomes the problems and is proposed as a switching system that will satisfy all the requirements of the Internet of Things.

REFERENCES

1. W. Xu and G. Campbell "DQRAP - A Distributed Queueing Random Access Protocol for a Broadcast Channel", presented at SIGCOMM '93, San Francisco. Computer Communication Review, Vol 23, No. 4, Oct 1993, pp. 270-278

2. C.T. Wu and G. Campbell, "Extended DQRAP (XDQRAP): A Cable TV Protocol Functioning as a Distributed Switch", Proceedings of 1st International Workshop on Community Networking, July 1994, San Francisco. 

3. C. T. Wu and G. Campbell "CBR Channels on a DQRAP-based HFC Network", SPIE '95 (Photonics East), Philadelphia, PA Oct. 1995.

4. H.J. Lin and G. Campbell “PDQRAP - Prioritized Distributed Queueing Random Access Protocol”, Proceedings of 19th Conference on Local Computer Networks, Oct 1994, pp 82-91.

5. Andrew Chen-Hung Chang and G. Campbell "Cascaded Distributed Queue Switching", Unpublished.

6.  Andres Laya, Charalampos Kalalas, Francisco Vazquez-Gallego, Luis Alonso and Jesus Alonso-Zarate “Goodbye, ALOHA!” IEEE Access Year: 2016, Volume: 4 Pages: 2029 - 2044,