uDrop: Pushing Drop-Boxto the Edge of Mobile Network

Fangzhou Jiang∗†, Kanchana Thilakarathna∗†, Sirine Mrabet∗, Mohamed Ali Kaafar∗∗National ICT Australia, {firstname.lastname}@nicta.com.au

†University of New South Wales, Australia

Abstract—Content sharing and online collaboration have be-come extremely popular in recent years, in particular with thepervasiveness of cloud storage services such as Dropbox. However,these popular cloud services are not accessible when users areoffline and do not take the geographical location of the users intoaccount when distributing content. Despite years of research inopportunistic content dissemination, real users are yet to fullyexperience the advantages of content sharing in proximity. To thisend, we propose uDrop which literally pushes the cloud storageservices to the edge of the network whenever interested users arein proximity of each other. uDrop enables content sharing amongusers in the vicinity without relying on networking infrastructurewhile improving the user quality of experience and reducing thecost of bandwidth. In this demonstration, we show the practicalfeasibility of uDrop service on Android smartphones.

I. INTRODUCTION

Today’s pervasive mobile applications rely heavily on cloudbased infrastructure either for storage, computation or manage-ment in the process of mobile content distribution. Personalcloud storage have almost become a commodity along with thegrowing popularity of services such as Dropbox and GoogleDrive. All these popular services are centralized in naturewhere all the content is hosted at the service provider requiringall users to have Internet connectivity to access the service.The perceived advantages of this communication model arethe centralized control, simplification of device functionalityand the ability to connect with virtually any other deviceregardless of the device location. However, this makes theservice unavailable for offline users and inefficient in termsof networking resource usage as the service is required tocommunicate via the cloud service even if the end user devicesare physically co-located.

Despite the immense research effort in opportunistic com-munication and ad-hoc networking, centralized services aretypically being provided largely due to the benefits fromharvesting user data outweigh the expense of costs associatedwith storage and distribution of data. In this paper, we aimto demonstrate the feasibility of moving the cloud storageservices to the edge of the mobile network, and in particularto the smart mobile devices, such that users may take ad-vantage of co-location when sharing user generated content.The proximity based content sharing/hosting via short-rangenetworks such as WiFi and Bluetooth not only allows tooffload communications from the congested cellular networks,but also, and most importantly from a user perspective, reducesthe monetary bandwidth costs and latency. Distributing con-

tent locally often improves the user quality of experience inparticular when the content size is large and user density ishigh.

We propose uDrop, a distributed mobile cloud storageservice that takes advantage of the geographical co-locationof users to exchange content locally while still providing theconvenience of cloud storage services when users are notwithin each other vicinity. In general, it has been demonstratedthat users in proximity of each other are more likely tobe interested in similar content [1]. For instance, cohorts ofpeople in large events such as festival or concert are likely tobe interested in sharing their experience of the event. Photos orvideos shot from a better perspective could therefore be sharedamong people of interests. Along with the pervasiveness ofmobile hand-held devices and their increasing capabilities,uDrop exploits these local concentration of user interest, whichextends the reach of cloud storage to the users who are offline.Sharing in proximity expands users’ social network to thephysical proximity, and then blends the physical proximityback into the social network, which we believe would resultin a much improved mobile content distribution efficiency.We demonstrate the feasibility of uDrop through a prototypeimplementation on Android smartphones including a location-based social discovery feature and local content dropping andretrieving processes.

The rest of the paper is structured as follows. Section IIdescribes an overview of our system. Section III and Sec-tion IV provide the details of the prototype implementationand demonstration. Section V compares uDrop with relatedwork.

II. SYSTEM OVERVIEW

uDrop aims to provide the ability to seamlessly share con-tent with users in proximity irrespective of the availability ofnetworking infrastructure. uDrop achieves this by harnessingthe advanced capabilities of modern smart mobile devices viaa mobile application (uDrop-app). uDrop-app enables usersto create a local network and collaborate with other uDropusers in the vicinity. uDrop users will also have access toa cloud storage partition, which they can use to collaboratewhen they are not in the proximity. Any uDrop user caninitiate sharing content by activating a uDrop-Box on thedevice, which invokes WiFi tethering on the device to form auDrop-LAN. An overview of uDrop content sharing process isillustrated in Figure 1 which consist of two phases; 1) Social

2016 IEEE International Conference on Pervasive Computing and Communications Demonstrations

978-1-5090-1941-0/16/$31.00 ©2016 Crown

uDrop-Cloud

uDrop-LAN

uDrop-Box

Member1

Member2

Member3

Syncthumbnail

DropRequest

CloudSync

cloudpar99ons

Fig. 1. uDrop System Overview

discovery, and 2) Content dropping and retrieving as describedbelow.

A. Social Discovery

At the time of installation of uDrop-app, each user is askedto register with the uDrop content management server (CMS)to gain access to the uDrop cloud storage partition (uDrop-Cloud), which can be used to host content and/or as a databackup service. uDrop-app shares the user’s locations with theuDrop-Server periodically to assist in discovering the nearbyuDrop users in the order of geographic proximity to the user’scurrent location or in a map as shown in Figure 2a. uDropusers can invite nearby users to form a uDrop-LAN to startsharing content locally as illustrated in Figure 2b. uDrop-app also notifies the user to move closer to the uDrop-Box,if the user is outside the uDrop-LAN reach. When Internetconnectivity is not available, nearby users can be discoveredonly if; 1) there is an active uDrop-LAN in the proximityor 2) by activating a uDrop-LAN on the device. Further, thenearby user list is limited to the users within the uDrop-LANcommunication range. In addition, uDrop allows its users toremember its previously connected users and add them asfriends forming a location-based content sharing network.

B. Content Dropping and Retrieving

Each member of the uDrop-LAN can drop any contentinto the connected uDrop-Box. More specifically, the devicehosting the uDrop-Box receives the meta information of anycontent drops by its member including a thumbnail for thecontent. uDrop-Box makes these information available to allthe connected members. Members in uDrop-LAN are also ableto exchange content directly with the assistance of uDrop-Boxfor a minimum latency, and these contents are to be pushedto uDrop-Box later. The actual content will be uploaded tothe uDrop-Box, if a member drop a particular content or thecontent is pushed after retrieving. The local data transmissionsimproves the user’s quality of experience while saving themobile bandwidth for the user.

(a) (b) (c)

Fig. 2. uDrop User Interface

All uDrop-apps updates the CMS with the meta informa-tion of connected uDrop-LANs and uDrop-Boxes periodically,whenever they are connected to Internet. This enables the CMSto create a virtual uDrop-LAN by setting up proper accesspermissions among uDrop-Cloud partitions. Users may alsocollaborate on the same content, where CMS will backupmultiple versions of the content. When uDrop users areconnected to their uDrop-Cloud via a low-cost network, i.e.WLAN, uDrop-app uploads the content that are stored onthe local uDrop-Box to make the content available online tomembers of the virtual uDrop-LAN.

III. IMPLEMENTATION

uDrop system consists of two main entities; 1) uDrop mobileapp and 2) uDrop cloud services

A. uDrop Mobile App

The prototype of uDrop-app has been implemented onAndroid smartphones as an standalone application. uDrop-appcan be developed as a standalone application or an extensionto popular cloud storage services. The basic components of theuDrop-app are illustrated in Figure 3. The Android uDrop-appconsists of two main activities socialDiscovery and fileEx-plorer that interact with the user and one Android service ser-verService which runs in the background. socialDiscoveryinitiates the relevant intents to discover nearby uDrop memberseither communicating with the CMS when online or sendingand listening to the social discovery probes locally whenoffline. fileExplorer handles user requests related to contentdataManager. The database access-control permissions aremanaged by the requestHandler, while dataManager keepstrack of the history of modifications, record usage statisticsand update the cloud partitions. Finally, thumbnailManagerand filesManager update and manage thumbnail and file metainformation in local uDrop-Box via database queries.B. uDrop Cloud Services

uDrop cloud services consist of the CMS and individualcloud storage partitions (uDrop-Cloud). The CMS assist in thesocial discovery phase by collecting and hosting user locations

uDrop-LAN

FileExplorer Activity

Request Handler

FilesManager

ThumbnailManager

Socket 1 Socket 2 Socket 1Server Service

uDrop-Box

uDrop-Member

Database

Data ManagerVersion Control

Usage Statistics

Storage Manager

SocialDiscovery Activity

uDrop-Cloud

CloudSync

Fig. 3. uDrop-app Implementation

and authenticating uDrop-app users. In addition, the CMSkeeps an updated database of uDrop friends. uDrop system isdesigned in such a way that it can be easily integrated with anexisting cloud service provider. Therefore, the uDrop-Cloudstorage partition simply acts as an online storage which ismanaged by the CMS.

IV. DEMONSTRATION AND TECHNICAL REQUIREMENTS

In this demonstration, we run uDrop-app on three co-located Android smartphones. In particular, we demonstrateits core functionality; 1) peer discovery, 2) content droppingand retrieving, as described earlier. One of the Android smart-phones acts as the uDrop-Box and hosts the uDrop-LAN. Othertwo Android devices are used to demonstrate dropping andretrieving content, as well as the version control functions.Alternatively, visitors with Android devices are also able toparticipate in the demonstration using their own devices, ifthey wish to install our prototype app. This demonstration hasno particularly special technical requirements. We will provideour own devices with uDrop installed. Access to power forcharging the phones would be preferable to ensure that thedevices are charged during the entire demonstration.

V. RELATED WORK

There have been a number of proposals for proximitybased content sharing services taking advantage of the recentdevelopments of ad-hoc networking and opportunistic commu-nication technologies [2], [3], [4], [5], [6]. BlueTorrent [2] is aBluetooth-based mobile peer-to-peer application, which allowsusers to cooperatively share contents locally. Proximiter [3]enables the sharing of various types of contents on mobiledevices, including text messages, profiles, pictures and etc.Miyamoto et al. [4] proposed an aerial-based survivor loca-tor framework, where the broadcast capabilities of existingmobile devices is leveraged to work together with the high-mobility drones for efficient message ferrying during natural

disasters. MoodChat [6] is a context-aware local sharingframework, while Disseminate [5] provides an implementationof a publish-subscribe based protocol for sharing contents viaWiFi-Direct. All these proposals only provides sharing in prox-imity whereas uDrop provides the convenience of seamlessconnectivity when users are both online and offline. Moreover,uDrop, in contrast to ad-hoc sharing, offers simultaneouscontent sharing among multiple users.

In addition to research proposals, Bump [7], being one ofthe most popular app in iOS and Android, achieved more than125 million downloads. Bump utilizes the device’s location,network and sensor data to connect peer users. However, itrequires an active Internet connection to do the matching andcontent transmission does not work in a peer-to-peer manner.As a result, users consume their own cellular data for contentexchange. Moreover, numerous groups have investigated onlocation-based content storage for mobile users [8], [9]. Again,both WhereStore [8] and Crowd-Cache [9] require a dedicatedstorage unit to provide the service.

In this context, the key innovation behind uDrop is that itextends the content sharing and collaborating features of cloudstorage into the smart mobile device without the requirementof Internet connection whilst still providing the convenienceof centralized cloud storage service. Put simply, uDrop istrying to put in practice efforts of the research community thatsuccessfully addressed, from a system design and performanceanalysis point of view, the proximity-based and opportunisticcommunication issues.

REFERENCES

[1] F. Jiang, K. Thilakarathna, M. A. Kaafar, F. Rosenbaum, and A. Senevi-ratne, “A Spatio-Temporal Analysis of Mobile Internet Traffic in PublicTransportation Systems: A View of Web Browsing from The Bus,”in Proceedings of the 10th ACM MobiCom Workshop on ChallengedNetworks. ACM, 2015, pp. 37–42.

[2] S. Jung, U. Lee, A. Chang, D.-K. Cho, and M. Gerla, “Bluetorrent:Cooperative content sharing for bluetooth users,” Pervasive and MobileComputing, vol. 3, no. 6, pp. 609–634, 2007.

[3] B. Xing, K. Seada, and N. Venkatasubramanian, “Proximiter: Enablingmobile proximity-based content sharing on portable devices,” in Interna-tional Conference on Pervasive Computing and Communications, 2009.IEEE, 2009, pp. 1–3.

[4] A. Miyamoto, D. J. Dubois, Y. Bando, K. Watanabe, and V. M. Bove,“Demo abstract: A proximity-based aerial survivor locator based onconnectionless broadcast,” in Pervasive Computing and CommunicationWorkshops (PerCom Workshops), 2015 IEEE International Conferenceon, mar 2015, pp. 184–186.

[5] V. Srinivasan, T. Kalbarczyk, and C. Julien, “Demo abstract disseminate:A demonstration of device-to-device media dissemination,” in PervasiveComputing and Communication Workshops (PerCom Workshops), 2015IEEE International Conference on, mar 2015, pp. 196–198.

[6] T. Diaz and C. Julien, “Demo abstract MoodChat: Using context-awareness to connect likeminded co-located individuals,” in PervasiveComputing and Communication Workshops (PerCom Workshops), 2015IEEE International Conference on, mar 2015, pp. 202–204.

[7] S. SEGAN, “Review: Bump, the One Billionth iPhone App,” Tech. Rep.,2009.

[8] W. Zeng, Y. Zhao, K. Ou, and W. Song, “Research on cloud storage ar-chitecture and key technologies,” in Proceedings of the 2nd InternationalConference on Interaction Sciences: Information Technology, Culture andHuman. ACM, 2009, pp. 1044–1048.

[9] K. Thilakarathna, F. Jiang, S. Mrabet, M. A. Kaafar, A. Seneviratne,and P. Mohapatra, “Demo: Crowd-cache–popular content for free,” inProceedings of the 12th annual international conference on Mobilesystems, applications, and services. ACM, 2014, pp. 358–359.