Leveraging knowledge-as-a-service (KaaS) for QoS-aware resource management in multi-user video transcoding
- PDF / 1,939,210 Bytes
- 16 Pages / 439.37 x 666.142 pts Page_size
- 112 Downloads / 214 Views
Leveraging knowledge‑as‑a‑service (KaaS) for QoS‑aware resource management in multi‑user video transcoding Luis Costero1 · Francisco D. Igual1 · Katzalin Olcoz1 · Francisco Tirado1
© Springer Science+Business Media, LLC, part of Springer Nature 2020
Abstract The coexistence of parallel applications in shared computing nodes, each one featuring different Quality of Service (QoS) requirements, carries out new challenges to improve resource occupation while keeping acceptable rates in terms of QoS. As more application-specific and system-wide metrics are included as QoS dimensions, or under situations in which resource-usage limits are strict, building and serving the most appropriate set of actions (application control knobs and system resource assignment) to concurrent applications in an automatic and optimal fashion become mandatory. In this paper, we propose strategies to build and serve this type of knowledge to concurrent applications by leveraging Reinforcement Learning techniques. Taking multi-user video transcoding as a driving example, our experimental results reveal an excellent adaptation of resource and knob management to heterogeneous QoS requests, and increases in the amount of concurrently served users up to 1.24 × compared with alternative approaches considering homogeneous QoS requests. Keywords Resource management · Heterogeneous Quality of Service · Reinforcement Learning · Multi-core architectures · HEVC video transcoding
* Luis Costero [email protected] Francisco D. Igual [email protected] Katzalin Olcoz [email protected] Francisco Tirado [email protected] 1
Departamento de Arquitectura de Computadores y Automática, Universidad Computense de Madrid, Madrid, Spain
13
Vol.:(0123456789)
L. Costero et al.
1 Introduction and motivation The integration of intelligent policies for resource management and application tuning in shared computing systems is becoming a field of paramount interest to efficiently exploit the potential of the underlying architectures without human intervention. In situations where external limitations in terms of Quality of Service (QoS), tight per-application SLA (service level agreements) or energy consumption are imposed, the development and application of such policies become a hurdle difficult to be automatically addressed [14]. Two of the main properties of any generic autonomous system, including resource managers in the fields of cloud computing or HPC, are self-configuration (ability to adapt to environmental changes) and self-optimization (capability to improve performance and reduce overloading or underloading the underlying resources) [10]. Resource managers can be actually considered in terms of the IBM autonomic model [7], which encompasses four main generic steps. This sequence of steps is repeated in a control loop that usually features sensoring and acting capabilities towards the underlying architecture or application. These actions are commonly cast in terms of selecting values for architectural knobs (e.g. core frequency) or application-specific knobs
Data Loading...
