Reliability, timeliness and load reduction at the edge for cloud gaming

Abstract

In this paper we study reliability, timeliness and load reduction in an hybrid Mobile Edge Computing (MEC)/Cloud game streaming infrastructure. In our scenario, a user plays a game streamed by the producer to their handheld deviceor User Equipment (UE). The UE communicates user actions to the edge/cloud via the mobile communication infrastructure; the object is to retrieve the latest game status, of which the most important information is the rendered frame. Particularly, we study reliability through replication in a number of MECservers, timeliness through Age of Information (AoI) and load reduction by leveraging the X2 interface at the edge, in order to abort useless frame rendering computations. We translate it as a scenario where a sink-representing the UE-is interested in the freshest possible update from distributed nodes. Each node sends updates following a Last Come First Served (LCFS) policy with preemption. We consider two scenarios; the first is n parallel LCFS systems sending updates, and the second adds a feedback loop aimed at decreasing the number of jobs sent per second by the nodes, thus decreasing the load per node. We analyze the number of jobs sent per second and average peak Age of Information at the sink, showing that the second scheme achieves a significantly lower rate of jobs compared with the first, while maintaining constant AoI, thus reducing the load at the edge. We also find that using the feedback loop, we achieve the maximum saving in transmitted jobs per second when the average arrival rate per system is equal to the inverse of the average busy time in every node. Index Terms-Age of information; Queuing theory; Cloud gaming; Edge computing.

An Overview of the Networking Issues of Cloud Gaming

“…Based on the identified issues and challenges, 20 pieces of literature were selected to discuss solutions. From the literature, the identified solutions include lag or latency compensation [17], [29], compression with encoding techniques [20], [32], [40], [41], the use of client computing power [29], [35], [42], edge computing [23], [25], [31], [43], machine learning [36], [38], [44], [45], frame adaption [46], [47], and GPU-based server selection [48], [49]. Figure 3 presents the identified solutions along with the number of references cited to review them.…”

“…Encoding Techniques [20], [32], [40], [41] ✓ ✓ ✓ Client Computing Power [29], [35], [42] ✓ Edge Computing [23], [25], [31], [43] ✓ ✓ ✓ ✓…”

“…However, they stated that the cost could become higher and that stability may be lower as more edge nodes are needed and added. On the other hand, for mobile users, Franco et al [43] stated that a hybrid mobile edge and cloud computing system could increase reliability and reduce load. Integrating machine learning into cloud gaming is a strategy that can also be used to reduce delay.…”

Peak Age of Information Distribution for Edge Computing With Wireless Links

“…represents an upper bound to the age, and can be used in reliability contexts. These metrics have been extensively used in XR [20], [21] and vehicular [22] use cases.…”

Load Reduction for Timely Applications in the Cloud

“…The first work found the exact expression of the average AoI, while the latter found its entire distribution. Timely applications in the edge/cloud were studied in [9]- [12]. Particularly, in the latter, a feedback loop is inserted as a means of reducing the load at the server side, but only the percentage of aborted tasks is considered as a metric of the advantage of having feedback.…”