PaperlayerView exact revision on arXiv

Source-linked AI summary

A Performance Evaluation of Container Technologies on Internet of Things Devices

Roberto Morabito

arXiv:1603.02955v1cs.DCcs.PF

TL;DR

The paper evaluates Docker’s performance on a constrained device, comparing benchmark results with native execution on a Raspberry Pi 2. Overall, the reported impact of the container layer on performance is almost negligible.

  • Problem

    The paper assesses Docker’s performance in constrained environments as container use expands across settings including IoT.

    Source: §I · p. 1

  • Method

    The study runs benchmark tools on a Raspberry Pi 2, compares Docker with native execution, and measures power consumption.

    Source: §II · p. 1

  • Results

    The container layer has an almost negligible performance impact overall compared with native execution, though disk I/O degrades by roughly 10%.

    Source: A. Synthetic Benchmark · p. 1, §IV · p. 2

  • Takeaways & Limitations

    The reported results support an almost negligible overall performance impact for Docker containers on the evaluated Raspberry Pi 2.

    Source: §IV · p. 2

    The evaluation is on a Raspberry Pi 2; the authors identify testing additional single-board computer platforms as future work.

    Source: §IV · p. 2

Abstract

from arXiv · show

The use of virtualization technologies in different contexts - such as Cloud Environments, Internet of Things (IoT), Software Defined Networking (SDN) - has rapidly increased during the last years. Among these technologies, container-based solutions own characteristics for deploying distributed and lightweight applications. This paper presents a performance evaluation of container technologies on constrained devices, in this case, on Raspberry Pi. The study shows that, overall, the overhead added by containers is negligible.

I. INTRODUCTION

Containers offer a lightweight alternative to hypervisor-based virtualization by isolating processes at the host operating-system level, avoiding virtualized hardware and device-driver overhead. The paper assesses Docker’s performance in constrained environments such as Raspberry Pi.

I. INTRODUCTION

  • Containers isolate processes at the host operating-system level, avoiding the overhead of virtualized hardware and virtual device drivers.They package an application with the dependencies needed to run it.Source: §I · p. 1
  • The study aims to assess Docker in constrained environments through a detailed performance analysis.The motivation includes containers’ lightweight characteristics and increasing use across contexts such as cloud computing and IoT.Source: §I · p. 1

II. METHODOLOGY

The study compares Docker with native execution on a Raspberry Pi 2 using benchmark tools and measures power consumption. Its testbed setup is shown in Figure 1.

II. METHODOLOGY

  • The investigation evaluates Docker on a single-board computer using different benchmark tools.The passage introducing the goal is incomplete and does not specify further methodological details.Source: §II · p. 1
  • Docker performance is compared with native execution on a Raspberry Pi 2, with results averaged over 15 runs.The setup uses Docker 1.8.0 and a Hypriot Raspbian Jessie image with Linux 3.18.11; power is measured with an external meter.Source: §II · p. 1
  • Figure 1 presents the testbed environment setup.The supplied figure caption identifies it as the testbed setup.Source: Figure 1 · p. 1

A. Synthetic Benchmark

Synthetic benchmarks test CPU, memory, disk, and network behavior, with Docker results compared against native execution. CPU and memory differences are small, while disk I/O shows roughly 10% degradation; network results vary by protocol and direction.

A. Synthetic Benchmark

  • Roughly 10%: Docker disk I/O performance degrades relative to native execution.The study uses sysbench random read/write operations for this test.Source: A. Synthetic Benchmark · p. 1
  • 2.67%: Docker’s CPU performance difference from native execution is described as negligible.The CPU workload is run with sysbench.Source: A. Synthetic Benchmark · p. 1
  • 6.04%: The maximum memory-performance difference occurs during the memcpy test, with native and Docker performance otherwise described as comparable.The mbw tool tests memcpy, dumb, and mcblock by copying large arrays in memory.Source: A. Synthetic Benchmark · p. 1
  • 4.90%: Docker power consumption is higher than native when the Raspberry Pi 2 receives TCP traffic.For TCP client traffic, the study reports no power-consumption mismatch; at 80 Mbps UDP, power consumption is aligned in client and server cases, while 100 Mbps traffic has packet loss.Source: A. Synthetic Benchmark · p. 2
  • 13.13%: Docker performs fewer network read operations than native in the reported network benchmark.The table reports 107.062 MB for Docker and 123.25 MB for native; the accompanying text explains that the network analysis measures power during intensive traffic exchange.Source: A. Synthetic Benchmark · p. 2

B. Application Benchmark

Application benchmarks emulate real-world workloads using MySQL and Apache2. MySQL performance is similar between native and Docker, while Apache2 containerization uses less power at the expense of slightly lower performance.

B. Application Benchmark

  • MySQL shows no relevant performance difference between native and Docker, especially at high thread concurrency.The benchmark uses databases with one and two million rows and varies the number of concurrent threads; Figure 2 plots transactions per second and power consumption against concurrency.Source: B. Application Benchmark · p. 2, Figure 2 · p. 2
  • Apache2 containerization uses less power than native execution at the expense of a slight performance degradation.The study examines request rate and execution time as the number of requests increases; Figure 3 plots those metrics.Source: B. Application Benchmark · p. 2, Figure 3 · p. 2

IV. CONCLUSIONS AND FUTURE WORK

On Raspberry Pi 2, container virtualization has an almost negligible performance impact compared with native execution. Future work will test additional Single Board Computer platforms.

  • Container virtualization has an almost negligible performance impact on Raspberry Pi 2 compared with native execution.Source: §IV · p. 2
  • The paper evaluates container-based technologies running on a Raspberry Pi 2.Source: §IV · p. 2
  • Future work will conduct similar experiments on additional Single Board Computer platforms.Source: §IV · p. 2
Loading 1603.02955v1…