This article presents the potential of using Multipath Transmission Control Protocol for limiting the energy consumption in 5G network. The number of errors occurring during packet transmissions and in effect the number of retransmissions affect the consumption of energy by the devices in the network. The paper analyzes the potential energy savings from implementing an algorithm for detecting problems and predicting the future retransmissions. Although this is the main object of the paper, it must be emphasized that the proposed method also allows increasing the speed of transmission and improving the security of the data and it is easy to implement in 5G networks.
A general opinion on the 5G network is that this technology could not be used in the near future, mostly due to the fact that the business segment remains unprepared for its implementation. Lots of GSM/3G/LTE operators have invested much in their infrastructure and what they now focus on is the return on this investment.
In a technical aspect, 5G networks are defined by the following parameters [ 1 millisecond end-to-end round trip delay (latency). 1000x bandwidth per unit area. 10–100x number of connected devices. (Perception of) 99.999% availability. (Perception of) 100% coverage. 90% reduction in network energy usage. Up to ten-year battery life for low power, machine-type devices.
Only a network meeting these provisions can be called a 5G network. An additional parameter is a new radio interface [
One of the existing technologies which can be taken into account for cross-layer optimization for 5G network communications and which is able to use MIMO as well as the existing infrastructure, is MultiPath Transmission Control Protocol (MPTCP in short). MPTCP technology allows using all the existing links to provide one stable and fast connection between two points of communication. The fact of using more than one connection in 5G network is something which does not neglect the MPTCP technology but can lead to limiting the energy consumption, which can be achieved by reducing the amount of retransmissions.
No less important than the reliability of 5G network is its speed. In order to eliminate the problems with delays in MPTCP, fuzzy logic can be applied, especially Ordered Fuzzy Numbers (OFN in short, called in some papers Kosinski’s Fuzzy Numbers), for predicting problems in the network [
The following section of the paper describes MPTCP technology in regard to its features and implementation status. Section
Prior to presenting the MPTCP technology, it is necessary to first introduce the concept of TCP [
TCP header.
Before the application starts transmitting the information, it is necessary to exchange the initialization data, as presented in Figure
Three-way handshake.
TCP connections cannot move from one IP address to another. When a PC switches from Ethernet to Wi-Fi, it is assigned a different IP address. Thus, all the existing TCP connections must be shut down and reestablished.
MPTCP is a set of extensions to the specification of TCP which allows the client to establish multiple connections using different network cards with the same destination host. This way fault-tolerant and efficient data connections are maintained between the hosts that are compatible with the existing network infrastructures.
The main goal of this solution is to enable using multiple network paths for a single connection as presented in Figure
MPTCP is located at the transport layer and it aims to be transparent to both higher and lower layers, as presented in Figure
MPTCP in the stack.
New connection of MPTCP is established in the same way as a standard TCP, which is presented in Figure
Establishing connection.
Nowadays, establishing a new connection is complicated by middle boxes (switches, routers). A pair (IP, port) of source and destination hosts is not sufficient for identifying the connection. Therefore, MPTCP extended its functionality with an additional option—MP_JOIN. Adding a new subflow is made in three steps presented in Figure
Adding a new subflow into MPTCP.
First, MP_JOIN option provides a token which is generated with the key (truncated hash of the key), formed during the initial connection. The exchange of HMAC (hash-based message authentication code) is the second step.
Now that the subflows have been established, MPTCP can use them to exchange data. Each host can send data over any of the established subflows. Furthermore, Figure
Error control in MPTCP.
Standard TCP “subflow sequence number” supports the reception of a single subflow and ensures detecting any loss of data. MPTCP uses “data sequence number” to sort the received data before passing them to the application [
MPTCP header (in short).
In general, the users connect to the Internet on their smartphones via Wi-Fi or 3G, but not by both at the same time. If TCP’s connection fails for some reason, it must be reestablished. MultiPath TCP avoids this situation by dynamically switching to the link, so the user does not waste time for reconnecting. It may also select the optimum speed.
The first mobile system that has been supporting MPTCP [
MPTCP is able to increase the security level of the transmitted data by using many different links to reach the destination, in contrast to the present methodology, which is based only on network protection [ Step Step Step Step Step Step Step Step Wi-Fi in three speed variants: 54, 22, and 11 Mbps, with 3G as a secondary link, Wi-Fi in three speed variants: 54, 22, and 11 Mbps, with LTE in three different speed variants: 150, 50, and 5 Mbps, as a secondary link.
Table
Data transfer results.
Data length [MB] | Transfer time [s] for different connection types | |||||
---|---|---|---|---|---|---|
Wi-Fi 54 & 3G | Wi-Fi 22 & 3G | Wi-Fi 11 & 3G | Wi-Fi 54 & LTE 150 | Wi-Fi 22 & LTE 50 | Wi-Fi 11 & LTE 5 | |
1 | 0.02 | 0.05 | 0.09 | 0.01 | 0.02 | 0.07 |
10 | 0.2 | 0.48 | 0.92 | 0.05 | 0.15 | 0.69 |
100 | 2 | 4.78 | 9.17 | 0.54 | 1.53 | 6.88 |
1000 | 20 | 47.83 | 91.67 | 5.39 | 15.28 | 68.75 |
These results demonstrate the possibilities of transferring data using the proposed algorithm in many common situations, where mobile user does not get his full transfer speed. Such situation may occur in many places due to the specificity of building construction and network overload.
Table
Packet transfer results.
Connection type | Number of packets transferred over Wi-Fi | Number of packets transferred over 3G/LTE | Transfer time [s] |
---|---|---|---|
Wi-Fi 54 & 3G | 1080 | 20 | 0.02 |
Wi-Fi 22 & 3G | 1052 | 48 | 0.05 |
Wi-Fi 11 & 3G | 1008 | 92 | 0.09 |
Wi-Fi 54 & LTE 150 | 291 | 809 | 0.01 |
Wi-Fi 22 & LTE 50 | 336 | 764 | 0.02 |
Wi-Fi 11 & LTE 5 | 756 | 344 | 0.07 |
In common sense, the term “fuzzy” refers to the situation where some precise data, like metric values or exact quantity of objects, cannot be given. Instead, words like “less,” “a little,” and “some” are used. Such verbal description of the reality is specific for the branch of artificial intelligence called fuzzy logic. The concept was developed by American professor of Columbia University in New York City and Berkeley University in California, Zadeh, who published a paper “Fuzzy Sets” in “Information and Control” journal in 1965 [
The first attempt to describe the new operations on fuzzy sets was undertaken in early 1990s by Kosiński and Słysz [
An ordered fuzzy number
Respective parts of the functions are called part up and down and are presented in Figure
Ordered Fuzzy Number—function up and down.
Ordered Fuzzy Number presented in a way referring to fuzzy numbers.
The continuity of the two parts called UP and DOWN shows that they are limited by a specific range. This range has been defined by the following values:
Fuzzy number in OFN notation where the order is (a) positive or (b) negative.
Functions
A membership function of an ordered fuzzy number
Reversal of the orientation of the ordered fuzzy number
One of the most remarkable notations interpreting ordered fuzzy number is a set of key points: addition: scalar multiplication: subtraction: multiplication:
As it was presented in Section
When the number of retransmissions rises, the energy consumption is increasing as well. The application of OFN could speed up the decision to change the data transmission path, which allows decreasing the number of transmission errors. OFN allow predicting data loss in the currently used channel and can also help make a decision to retransmit the packet faster and limit the number of retransmissions as well. Lowering the number of retransmissions allows limiting the energy consumption.
In this section the algorithm which uses OFN for detecting potential problems is described. The algorithm measures a TCP retransmission in all the used channels during the transmission and provides it in a percentage value of the transmitted packets. This measure is made both in a given period of time—a timeslot—and continuously. Four timeslots of the continuous measure could be defined in time:
These four results together give a fuzzy number in OFN notation, where
This fuzzy number in OFN notation is presented in Figure
Fuzzy number in OFN notation.
The definition of fuzzy observation of the connection used is as follows.
Fuzzy observation of
One has
According to this definition, the counters of retransmissions in the links during the connection observation should give positive order of OFN when the packet retransmission count negative order of OFN when the packet retransmission count
The four measurements performed during the data transfer allow preparing fuzzy number in OFN notation. This gives a fuzzy observation of the MPTCP connections, defined as follows.
Fuzzy observation of the MPTCP connections is described by the following formula:
According to Definition
The algorithm with application of OFN for predicting transmission errors in MPTCP transmission consists of four steps.
The administrator declares the start value for
The amount of packets
During the transmission the fuzzy observation of connection
When the calculated
When the calculated
The
In order to check a MPTCP scheduler with OFN, a simulation was performed.
The system was designed with two connection links: Connections 1 and 2, labeled the corrector for the links acceptance level load balance on start for connection load balance on start for connection the timeslots used were 60 seconds.
Table
Fuzzy observation of the connections.
Time | % Error on connections | | ||||
---|---|---|---|---|---|---|
| | | | |||
1 | 1 | 3 | ||||
2 | 2 | 4 | ||||
3 | 2 | 6 | ||||
4 | 1 | 9 | | | | |
5 | 2 | 13 | | | | |
6 | 1 | 12 | | | | |
7 | 1 | 11 | | | | |
8 | 2 | 12 | | | | |
9 | 1 | 8 | | | | |
10 | 2 | 7 | | | | |
11 | 1 | 6 | | | | |
12 | 2 | 5 | | | | |
13 | 1 | 3 | | | | |
14 | 2 | 2 | | | | |
15 | 1 | 1 | | | | |
As it can be observed, after the 4th measure the OFN numbers is generated and
Load balance for the connections.
Time | | |
---|---|---|
| | |
1 | 50 | 50 |
2 | 50 | 50 |
3 | 50 | 50 |
4 | 50 | 50 |
5 | 50 | 25 |
6 | 50 | 12,5 |
7 | 50 | 6,25 |
8 | 50 | 12,5 |
9 | 50 | 25 |
10 | 50 | 50 |
11 | 50 | 50 |
12 | 50 | 50 |
13 | 50 | 50 |
14 | 50 | 50 |
15 | 50 | 50 |
The total number of packets passed to the connections changes during the timeslots according to the load balance set at the beginning of the test. This affects the number of packets transferred through the connections and, depending on the error level on the link, provides different number of errors. In the simulation there were 40000 packets passed to the link. Table
Number of errors in packet transmission.
Time | Errors count | |||||
---|---|---|---|---|---|---|
MPTCP with OFN | MPTCP without OFN | |||||
| | | | | | |
1 | 400 | 1200 | 1600 | 400 | 1200 | 1600 |
2 | 800 | 1600 | 2400 | 800 | 1600 | 2400 |
3 | 800 | 2400 | 3200 | 800 | 2400 | 3200 |
4 | 400 | 3600 | 4000 | 400 | 3600 | 4000 |
5 | 1067 | 3467 | 4533 | 800 | 5200 | 6000 |
6 | 640 | 1920 | 2560 | 400 | 4800 | 5200 |
7 | 711 | 978 | 1689 | 400 | 4400 | 4800 |
8 | 1280 | 1920 | 3200 | 800 | 4800 | 5600 |
9 | 533 | 2133 | 2667 | 400 | 3200 | 3600 |
10 | 800 | 2800 | 3600 | 800 | 2800 | 3600 |
11 | 400 | 2400 | 2800 | 400 | 2400 | 2800 |
12 | 800 | 2000 | 2800 | 800 | 2000 | 2800 |
13 | 400 | 1200 | 1600 | 400 | 1200 | 1600 |
14 | 800 | 800 | 1600 | 800 | 800 | 1600 |
15 | 400 | 400 | 800 | 400 | 400 | 800 |
The main goal of this paper is to analyze the effect of the proposed solution on the energy consumption. As it can be noticed, the system of transmission itself consumes much power. For instance, sending data packets over radio network in smartphones drains their batteries, while nowadays there are many processors available described as Low Energy Usage.
The presented algorithm does not require any complex calculations, so it does not cause any significant power dissipation. Rather than that, the problem is the energy consumption by the system of transmission. The process of sending data is power-consuming not only for the sender device, but also for the entire IT system. This system in the case of 5G networks comprises numerous devices: transmitters, switchers, and routers. Each of these devices on the packet’s path uses energy. In a case where the packet is retransmitted the total energy consumption is even higher as the receiver has to realize that a problem with the received data occurred and, for example, recalculate the CRC checksum; the receiver has to send the information to the sender about the problem with transmission; the sender has to send the data again.
As it was described above, in the case of retransmission, the data is sent twice, and, moreover, additional data concerning the problem is transmitted over the network, which means that the system is used thrice.
The analysis of the proposed algorithm allows concluding that implementing simple calculations and making optimal decisions by the sender faster contributes to more efficient power management in the whole IT ecosystem.
In this article, a special algorithm for mobile connections, applicable in 5G network, was introduced. For this algorithm a MPTCP concept was used and the energy consumption as well as the security aspect were developed by a special usage of fuzzy observation of the errors in the transmissions.
The presented new concept of MPTCP scheduler combined with OFN was tested during a data transmission simulation. As it was presented in the previous section, the proposed algorithm allows reducing the number of retransmissions, as less packets are transferred over the connection link in which some problems are detected. Such application of MPTCP allows limiting the energy consumption by decreasing the amount of data which has to be transmitted. This is a possible usage of OFN for improving the existing solutions like MPTCP in a simple way, without any complicated algorithm requiring a high level processor performance. This is very important for the 5G network, in which the speed and power consumption is a given parameter.
As it was mentioned in the introduction, there are numerous configurations in which not only 5G, but also the existing infrastructure will be used. In this situation there will be more than one connection to the same destination. For this reason, the MPTCP with OFN could give a valuable solution in critical network connections. The presented algorithm is able to provide a better energy management and higher transmission security and reliability without generating higher costs in the implementation process [
The author declares that there is no conflict of interests regarding the publication of this paper.