28 Apr 2017 |
Research article |
Information and Communications Technologies
Making LTE Networks More Efficient
This is one of the winning articles in the Ingenious Writers Contest organized by SARA and Substance ÉTS. It summaries an article co-written by Michel Kadoch, Professor in the Department of Electrical Engineering at the École de technologie supérieure (ÉTS), entitled: Enhanced Control for Adaptive Resource Reservation of Guaranteed Services in LTE Networks.
In its quest to facilitate everyday human activities through the promise of a more connected future, the network popularly referred to as the Internet of Things (IoT) continues to grow–bringing an ever increasing number and variety of things under its fold. Mobile devices are an important part of this connected network, and the consistent growth of services offered by mobile devices places unique challenges on the future sustainability of telecom networks within the framework of the IoT. Long-term evolution (LTE) telecom networks are faced with a set of increasingly difficult objectives–faster communication, higher data transmission, improved coverage and optimized radio access networks. It is therefore useful to evaluate possible changes to the design of currently deployed LTE networks so that they will continue to meet their future requirements resulting from an ever-growing IoT.
To meet these increasing requirements, the core of the LTE network–referred to as the Evolved Packet Core (EPC)–forms a critical design block towards realizing greater network efficiency, i.e. how efficiently the mobile network utilizes its resources to exchange information. However, the current design of the EPC system is not optimized with respect to ways that it allocates bandwidths to mobile services, and in particular to services referred to as guaranteed dedicated services. Fig. 1 below shows the current classification of mobile services in LTE systems.
When there is a request for a guaranteed service (such as a voice call) by the mobile device, the EPC allocates it a certain pre-determined dedicated bandwidth to ensure a promised Quality of Service (QoS). On the other hand, non-guaranteed services (such as video streaming) do not have any dedicated bandwidth allocated to them and can remain established for long periods of time. When there is a request for a guaranteed service, the already established non-guaranteed service may have to forego some of its bandwidth (and consequently suffer from a loss of QoS) so that the dedicated bandwidth serving the guaranteed service is available.
However, the entire bandwidth allocated to the guaranteed services may not be fully used. Moreover, there are currently no provisions to utilize this unused allocated bandwidth for other services that can benefit from utilizing it. All this translates into a mobile network that is forced to function below its actual capabilities due to a bandwidth allocation that is based only on a static implementation of identifying mobile services as guaranteed or non-guaranteed, and failing to identify the actual bandwidth requirement necessary to execute a guaranteed service. An example of such wasted resources is illustrated in Fig. 2.
The work presented by the authors in this paper is a step towards realizing a more efficient mobile network by offering a novel dynamic approach to the way resources are allocated for guaranteed services in LTE mobile networks. In contrast to the static bandwidth allocation that is currently used, the proposed technique is adaptive and it identifies the actual resource consumption by a guaranteed service, re-allocating any unused resources for other services that will benefit from utilizing this unused bandwidth. This prevents resource wastage and has a direct impact on improving the efficiency of the mobile network.
In the proposed approach (time series), guaranteed services in LTE mobile networks are further classified under two categories:
- Adaptive Guaranteed Services (also termed Contributing Services): These are guaranteed services that can contribute their unused resources to other guaranteed/non-guaranteed services.
- Pure Guaranteed Services: These are guaranteed services that cannot contribute their reserved resources even if they are unused. Examples of such services include critical services such as those for emergency calls.
Guaranteed and non-guaranteed services that are willing to utilize more resources than what they are currently utilizing are termed Acquiring services. The modified classification of services in the proposed approach is illustrated in Fig. 3 below.
Following this modified classification of services, the following five steps are executed:
- Step 1: Analyze the ongoing mobile traffic usage by adaptive services. This traffic usage data is a sequence of observations that are measured at consecutive points of time which are identically spread out.
- Step 2: Provide a time-series model that represents this analyzed traffic usage by adaptive services. Time series models represent the dynamic relation between the past/current observations of the analyzed data and their future values.
- Step 3: Forecast resource consumption by adaptive services using the time-series model. Time-series models allow forecasting of future resource consumption based on current and past values of the analyzed data.
- Step 4: Identify unused allocated resources to the adaptive services, with the help of data forecast using time-series model.
- Step 5: Allocate these unused resources to acquiring services.
The proposed technique was applied on a dataset representing the actual bit-rate in bits per second (b/ps or bps) for a conversational video service using the LTE network. The dataset is of 300s duration, and the service uses a guaranteed/reserved bit-rate (GBR) of 2.2Mb/s. A series of mathematical functions is first applied on this dataset to transform it in a way that is conducive to being modelled using a time-series. This is followed by experiments to determine the time-series model that best fits the experimental dataset. This step is critical, since an appropriate time-series model would allow forecasting future resource consumption of the guaranteed service based on current and past resource consumption with minimum error. Fig. 4 shows a comparison between the original dataset and the dataset obtained by forecasting using the optimized time-series model.
Based on this forecast data, the unused portion of the resources allocated to the conversational video service is now evaluated over a time period sample of 50s. This time sample was randomly selected from the entire 300s for which the original dataset was acquired. A safety margin based on the evaluated forecast error of the time-series model is also included so that the quality of the guaranteed service itself is not compromised by the proposed resource allocation. Fig. 5 shows the unused bit-rate over the 50s sample, and represents an average of 185.411 kb/s that can be released from the guaranteed 2.2Mb/s to be used by other mobile services that can benefit from utilizing it.
The proposed approach in this paper offers a novel approach to increase mobile network capacity by implementing a dynamic technique for resource allocation to guaranteed services. Such an improvement is extremely attractive for the future sustainability of mobile networks, in view of the increasing number of services that are being added by mobile devices due to the ever-widening scope of the Internet of Things.
For more information on this research, see the following reference article: Albasheir, Suliman and Kadoch, Michel. 2015. « Enhanced control for adaptive resource reservation of guaranteed services in LTE networks ». IEEE Internet of Things Journal, vol. 3, nº 2. p. 179-189.
Smarjeet Sharma is currently pursuing a Ph.D. degree at ÉTS in electrical engineering. He received the NTSE scholarship from NCERT, Government of India as well as the MITACS-Globalink Scholarship for undergraduate research in Canada.
Program : Electrical Engineering