Analyzing Real-Time Multimedia Content From Network Cameras Using CPUs and GPUs in the Cloud

TL;DR

This paper presents a cloud resource management approach that optimizes the use of CPUs and GPUs for analyzing real-time multimedia from network cameras, significantly reducing costs while maintaining performance.

Contribution

It introduces a resource manager that models cloud resource allocation as a multiple-choice vector bin packing problem and demonstrates cost savings through experimental validation.

Findings

Cost reduction of up to 61% compared to other strategies

Effective estimation of resource requirements for CPU and GPU analysis

Application of bin packing algorithms to cloud resource allocation

Abstract

Millions of network cameras are streaming real-time multimedia content (images or videos) for various environments (e.g., highways and malls) and can be used for a variety of applications. Analyzing the content from many network cameras requires significant amounts of computing resources. Cloud vendors offer resources in the form of cloud instances with different capabilities and hourly costs. Some instances include GPUs that can accelerate analysis programs. Doing so incurs additional monetary cost because instances with GPUs are more expensive. It is a challenging problem to reduce the overall monetary cost of using the cloud to analyze the real-time multimedia content from network cameras while meeting the desired analysis frame rates. This paper describes a cloud resource manager that solves this problem by estimating the resource requirements of executing analysis programs using CPU or GPU, formulating the resource allocation problem as a multiple-choice vector bin packing problem, and solving it using an existing algorithm. The experiments show that the manager can reduce up to 61\% of the cost compared with other allocation strategies.