没有合适的资源?快使用搜索试试~ 我知道了~
资源推荐
资源详情
资源评论
Managed Edge Computing on Internet-of-Things
Devices for Smart City Applications
Yu-Chen Hsieh
1
, Hua-Jun Hong
1
, Pei-Hsuan Tsai
1
, Yu-Rong Wang
1
, Qiuxi Zhu
2
,
Md Yusuf Sarwar Uddin
2
, Nalini Venkatasubramanian
2
, and Cheng-Hsin Hsu
1
1
Department of Computer Science, National Tsing Hua University, Taiwan
2
Department of Computer Science, University of California Irvine, CA, USA
Abstract—We demonstrate a managed edge computing plat-
form for Internet-of-Things (IoT) devices, which supports dy-
namic deployment of virtualized containers running distributed
analytics. We build a model city, and install multiple Raspberry
Pis as minions, and a mini PC as the master. T hrough the
web dashboard on the master, we show how users can remotely
monitor, manage, and upgrade the IoT analytics and devices.
Multiple concrete IoT analytics, namely: (i) air quality monitor,
(ii) sound classifier, and (iii) image recognizer are demonstrated.
Several sample measurements on deployment speed, Quality-of-
Service (QoS) achievements, and event-driven mechanisms are
also carried out on the testbed.
Index Terms—Internet-of-Things, fog computing, edge com-
puting, virtualization
I. INTRODUCTION
Smart cities have become very popular in the past decade,
with a focus on efficiently collecting data from lots of Internet-
of-Things (IoT) devices. Several commercial IoT platforms
are offered by service providers, such as AT&T, Amazon, and
Microsoft. They support edge computing, which may also be
known as fog computing [1], to a certain degree. However,
their supports on distributed analytics and event-triggered
deployment are limited. There are a few academic studies
trying to realize managed, programmable, and virtualized edge
computing. For example, Patti et al. [3] adopt Smartlink to
manage a set of web services on multiple servers. Santoro et
al. [4] employ Kubernetes to deploy and manage Docker im-
ages executed on servers in data centers and edge clouds. None
of the above-mentioned work supports distributedly deploying
and managing IoT analytics on heterogeneous devices from
data center servers, edge cloud workstations, and embedded
IoT devices (such as smart sensors).
In this paper, we present an IoT platform to fill the gap
mentioned above. Realizing such a platform is quite challeng-
ing, because of the complex interplay among IoT devices,
dynamic networks, and on-demand requests. The challenges
are amplified by the facts that IoT devices and edge servers
are resource constrained, wireless networks are best-effort, and
device requests from individuals, companies, organizations,
and government are unpredictable. Our earlier work [2, 5]
tackles the problem of optimally deploying IoT analytics.
The current paper focuses on the testbed implementation. In
This work was supported by the Ministry of Science and Technology of
Taiwan under grant #105-2221-E-007-088. The authors also appreciate the
generous supports from Lite-On Technology Corporation.
particular, we will demonstrate three IoT analytics: (i) air
quality monitor, (ii) sound classifier, and (iii) image recognizer
in a model city with multiple Raspberry Pis as minions and a
mini PC as the master.
"
#
"
!
*
*
*
"
"
"
%
&
'
(
)
Fig. 1. The system architecture of our IoT platform. The containers encap-
sulate multiple sample components from our demo applications.
II. OUR IOTPLATFORM
Fig. 1 shows the overall system architecture, which consists
of the following components:
• Container. We run applications in Docker containers.
Container is a lightweight virtualization technology. Each
Docker container is launched with a Docker image, which
consists of one or more software components.
• Registry. The Docker images may come from online
servers, such as Docker Hub. We store these images in
registry on the master and minions to avoid downloading
duplicated images. Registry also enables version control,
and leverages the dependency among Docker images to
further reduce the resource consumption.
• Device manager. We enhance Kubernetes to monitor
heterogeneous devices for: (i) resource utilization (e.g.,
CPU and RAM), (ii) location (e.g., GPS), and (iii) sensor
list (e.g., cameras and gas sensors). Such information is
collected from Kubernetes agents on minions (
5
).
• Deployment manager. The deployment manager in Ku-
bernetes is enhanced for deploying, killing, and restarting
(
4
) containers. We add supports for event-triggered con-
tainer deployment using the MQTT (Message Queuing
Telemetry Transport) pub/sub protocol.
• Optimization algorithm. We develop a deployment al-
gorithm [5] to decide which applications are deployed
on which devices (
3
). We may add more optimization
978-1-5386-3416-5/18/$31.00
c
2018 IEEE
资源评论
帅气滴点C
- 粉丝: 1w+
- 资源: 7
上传资源 快速赚钱
- 我的内容管理 展开
- 我的资源 快来上传第一个资源
- 我的收益 登录查看自己的收益
- 我的积分 登录查看自己的积分
- 我的C币 登录后查看C币余额
- 我的收藏
- 我的下载
- 下载帮助
最新资源
- IMG_20240425_190941.jpg
- 基于stm32的pid控制算法附带串口控制台
- 实验项目一:线性表(模板 ).docx
- 基于LM317A设计交流220V输入可调5V-12V直流输出稳压电源电路multisim10仿真源文件.zip
- Screenshot_2024-04-25-17-04-18-55_2332cb9b27b851b548ba47a91682926c.jpg
- -基于AVR单片机的LNG加液机控制器设计.pdf
- ThePowerOfNow-EckhartTolle.mobi
- BLOCK_TYPE_HEARTBEAT_D70A3465D4EE4E9_046141_dump_1st.dmp
- 项目方法测试调用接口工具
- studyupdate
资源上传下载、课程学习等过程中有任何疑问或建议,欢迎提出宝贵意见哦~我们会及时处理!
点击此处反馈
安全验证
文档复制为VIP权益,开通VIP直接复制
信息提交成功