Today's cellular networks are built with``macro cell'' basestations connected to the Internet via a rigid, complicated backhaul. Even with state-of-art technologies like LTE, users get limited throughput and high latency, with high variance. Performance enhancing IP boxes are deployed in the cellular operator's datacenters, far from the user. As a result, the most compelling cloudlet applications are difficult to realize on such networks and cloudlet researchers have thus far focused on Wi-Fi networks only.
We argue that the cloudlet community should consider small cell networks in addition to Wi-Fi networks. Small cells, such as femtocells and picocells, are relatively new additions to the cellular standards. By reducing the cell size compared to the traditional macro cells, they increase spatial reuse of precious licensed frequencies. Users get higher bandwidth and lower latency, with relatively less variance. This architecture, where small cells are deployed simply with power and Ethernet connectivity, lends itself well to cloudlet augmentation. In this position paper, we describe why deployed macro cell basestations are unsuitable for cloudlet deployment. In contrast, we describe why a small cell architecture is amenable for cloudlet deployments. Our experience from operating a small cell testbed in licensed frequencies matches that reported by equipment vendors. The applications we care about require high throughput and low latency. In a cellular network this can be achieved today by augmenting small cells with powerful cloudlets.
[1]
Ming Yang,et al.
DeepFace: Closing the Gap to Human-Level Performance in Face Verification
,
2014,
2014 IEEE Conference on Computer Vision and Pattern Recognition.
[2]
Seungyeop Han,et al.
The Case for Onloading Continuous High-Datarate Perception to the Phone
,
2013,
HotOS.
[3]
Paramvir Bahl,et al.
The Case for VM-Based Cloudlets in Mobile Computing
,
2009,
IEEE Pervasive Computing.
[4]
Alec Wolman,et al.
MAUI: making smartphones last longer with code offload
,
2010,
MobiSys '10.
[5]
Mahadev Satyanarayanan,et al.
Towards wearable cognitive assistance
,
2014,
MobiSys.
[6]
Vincent Lepetit,et al.
View-based Maps
,
2010,
Int. J. Robotics Res..
[7]
Feng Qian,et al.
A close examination of performance and power characteristics of 4G LTE networks
,
2012,
MobiSys '12.
[8]
Geoffrey E. Hinton,et al.
ImageNet classification with deep convolutional neural networks
,
2012,
Commun. ACM.
[9]
Ramesh Govindan,et al.
Odessa: enabling interactive perception applications on mobile devices
,
2011,
MobiSys '11.
[10]
Tadahiro Kuroda,et al.
A versatile recognition processor employing Haar-like feature and cascaded classifier
,
2009,
2009 IEEE International Solid-State Circuits Conference - Digest of Technical Papers.
[11]
David L. Tennenhouse,et al.
Proactive computing
,
2000,
Commun. ACM.
[12]
Michael I. Hill,et al.
Generalizeability of Latency Detection in a Variety of Virtual Environments
,
2004
.