The past years have witnessed the significant development of the Internet. Numerous emerging network architectures and protocols have triggered the demand for traffic generators which stand in stark contrast to previous schemes. Namely, fixed test content is inefficient in the presence of such a dynamic and realistic demand. Moreover, the requirement of high-performance has raised the stakes on developing a new concurrent system. In this paper, we present a hierarchical parallel design for a

more »

... traffic generator on a TILERAGX36 processor, called TGMP. We discuss the challenges in developing its hierarchical architectural design, and elaborate on its implementation details. Specifically, in order to generate a realistic network workload over a long and large time scale, we propose a user-control scheme based on cubic spline interpolation. To better improve the scalability of the system and satisfy the required flow rate, we adopt techniques, including optimization of parameters under the Linux kernel, event-driven concurrency, and parallel architectures of a TILERAGX36 processor. The experimental results demonstrate that TGMP is able to create real traffic and simulate 50,000 users accessing the Web server simultaneously. Appl. Sci. 2017, 7, 154 2 of 23 Interface). However, these kinds of tools are inefficient and poor in performance, so it is hard to satisfy the requirements of a future network. To be specific, take Geist [7] for example, which is a tool that generates realistic traffic for exercising web-servers and e-commerce front-end servers but the insufficient performance allows limited access. In a nutshell, common instrumentation usually provides limited flexibility and available software generators have poor performance. In this paper, we join both high-performance equipment and programmable software design. In literature, a huge amount of work exists on the characterization, modeling and simulation of the network workload such as, live streaming media [8] , and YouTube traffic [9] . Unfortunately, we cannot state the same for the generation of a realistic network workload based on those methods. Many other studies also have contributed in the field of Web traffic generation. Zinke [10] points out that the Web traffic generator has two requirements. A real workload is defined as a sequence of requests which are received from a real world web server. A representative workload is defined as a generated workload which has the same characteristics as a given real workload. When it comes to the realistic network workload, numerous researches have been done. In order to formulate different kinds of traffic, Cheng [11] proposes a HTTP traffic generator to generate user-defined HTTP traffic and analyze the performance under different network characteristics. Given the representative workload, based on the characteristics of the present representative workload, Botta [12] proposes a method for the realistic network workload to study the emerging networking scenarios with multidimensional heterogeneity and scale. However, most of the related work cannot generate a realistic network workload over a long and large time scale. In this paper, aiming to design a Web traffic generator with the characteristics of verifying the future network architectures, we propose a flexible framework to support arbitrary models. Firstly, the system should generate realistic web traffic over a large timescale; secondly, a large number of users can certainly be described; finally, multiple operators can use the platform simultaneously. The generic Many-Core architectures with lots of cores [13], a remarkable new field, are providing new opportunities for a high-performance traffic generator. We develop a high-volume parallel design for a Web traffic generator on Many-Core processors, called TGMP. Unlike other traffic generators, we concentrate on the hierarchical architectural design, which allows for better control of the traffic and a more scalable generation. In order to generate a realistic network workload over a large time scale, we combine the method of user behavior with the user-control method by cubic spline interpolation. We first explore how the TGMP can be leveraged to efficiently overcome the limitations in current Web traffic generator architectures with low flexibility, low scalability and high cost. After that, to break this stalemate, we have tackled two main technical issues. First, we perform large-time-scale flow simulation. Specifically, we use cubic spline interpolation to handle the traffic self-similarity in one hour or one day, which keeps the corresponding traffic flow that reflects the real network conditions. Compared to solutions performed by traffic replay, our solution does not need to capture the packet in advance. Second, we explore new parallel opportunities provided by Many-Core processors. More specifically, we propose a parallel design for implementing TGMP on the TILERAGX36. Our solution is inspired by Jiang [14], who designs a NIDS (Network Intrusion Detection System) on such processors on which tasks are split and sub-tasks are allocated to run concurrently through decomposition techniques. The contribution of our work can be listed as follows: