WWW2003 Poster Template

Storage Provisioning for Mobile Devices to Enable Faster Internet Access

Harish Kammanahalli
Applied Research Group, Satyam Computer Services Ltd.
14 Langford Avenue, Lalbagh Road
Bangalore, India 560025
+91 80 223 1696
Harish_kammanahalli@satyam.com
Srividya Gopalan
Applied Research Group, Satyam Computer Services Ltd.
14 Langford Avenue, Lalbagh Road
Bangalore, India 560025
+91 80 223 1696
Srividya_gopalan@satyam.com
Sridhar V
Applied Research Group, Satyam Computer Services Ltd.
14 Langford Avenue, Lalbagh Road
Bangalore, India 560025
+91 80 223 1696
Sridhar@satyam.com

ABSTRACT

Utility of network connectivity to provide web based services to mobile devices could be greatly enhanced by providing a personalized network attached storage for mobile users. Dependency on the best-effort Internet to access storage poses well-known constraints on timely availability of data. In this paper, we propose storage overlay network, a network of storage devices that are inter-connected by high speed connectivity, to provide a fast, reliable, and secure access to network storage for any mobile device anytime anywhere for any information.

Keywords

Storage Overlay Network, wireless 2.5G, wireless 3G, extended storage, mobility management

1. INTRODUCTION

Accessing the World Wide Web data by using mobile devices is increasing due to the deployment of 2.5G and 3G services. A mobile users web access is largely determined by the user specific preferences and the presentation of data is constrained by the capabilities of the device used. User profile is a good representation of the users preferences, likes and dislikes, and lifestyle related information. Constrained presentation is achieved by using device specific features, user specific preferences, and context specific characteristics in transcoding process. As transcoding is compute-intensive, it is useful to store multiple fidelity versions of an object in a store to provide faster access. Reachability to such stored transcoded objects is dependent on the prevailing network and load conditions. In order to reduce the delay in accessing an object in the store, context based pre-fetching is useful. The proposed dedicated Storage Overlay Network (SON) provides an environment for reliably addressing the above mentioned aspects. SON helps in overcoming the problems due to intermittent connectivity and also reduces severe load conditions on the wireless network infrastructure due to data networking in mobile systems.

Performance enhancing proxy servers that perform various functions like content transformation, caching and pre-fetching to decrease latency are being considered. Jiang et al [3] study various design considerations for placing proxy servers near the base station, near the access router, or near the gateway router inside wireless networks. Knowledge of a user's preferences is used by Cherniack et al [2] to pre-fetch the data into the cache in a proxy. Even if data is pre-fetched, the mobility of devices pose challenge for ensuring availability of coherent data. Kulkarni et al [5] propose techniques for ensuring coherency of data in wireless networks using application level handoff where the data is moved from one proxy to another across Internet based on users movement. But the movement of the data during handoff is still dependent on Internet that exposes users to uncertainties of the best-effort nature of Internet.

In a world that is always connected, the information should be made accessible always and many a times, the information could be manipulated by using more than one device. In this ubiquitous network, there is a need for networked data storage that is accessible from multiple devices. Villate et al [7] propose an agent based data locker facility where the data is located nearer to a mobile device. The data lockers are present at gateway support nodes of the wireless network, and an agent dedicated to a mobile unit interacts with locker agents to maintain the data storage even during user mobility.

Providing data near to the physical location of a mobile unit through Internet is prone to delays due to the dependence on the best-effort Internet. Performance could be increased by the usage of a dedicated network of storage systems that provides a fast and reliable movement of the data store. We propose to use a storage overlay network, a network of storage devices that are inter-connected by high speed connectivity, to provide a fast, reliable and secure access to network storage for any device at any location and at any time.

2. STORAGE OVERLAY NETWORK

RON (Fig. 1 provides a table of acronyms) is an architecture that allows end-to-end communication across the wide-area Internet to detect and recover from path outages and periods of degraded performance within several seconds [1]. Using the overlay concept, we have proposed storage overlay network on top of a wireless network. The overlay nodes in storage network provide extended storage to mobile users and also manage the users stores and storage-related requests. Further, SON nodes manage the mobility of a users stores to an easy access store when the user is roaming and provides extended storage to the user anytime anywhere with minimum delay. The storage nodes in the SON are connected by means of high-speed connectivity to aid in faster access to stored data. SON can be IP based or SAN based and can also be geographically distributed.

SON comprises of nodes consisting of multiple home and visitor stores. A node in SON can be connected to one or more BTSs, BSCs, SGSNs, and GGSNs in a GPRS network. Different architectures have been proposed in the literature where BTS functionality has been extended by way of providing proxy servers [4]. Figure 2 depicts the proposed architecture of SON on a GPRS network.

3. SON CLIENT AND SERVER

SC is an agent that is present in all the components of a mobile network, such as BTS, BSC, SGSN, and GGSN, communicates with a storage node of SON and is a part of SMM. SS is present in all nodes of SON. A typical layered architecture is shown in Fig. 3. The main functions of SC and SS are to collectively provide an efficient access to user store.

During power on, a store attach request is communicated via a BTS to the associated SN. SN obtains the home store location of the user, creates a visitor store (if necessary), retrieves the relevant objects, such as calendar and inbox, into the visitor store, and communicates back a unique SID. The SM of MS uses this SID while communicating the storage service requests initiated by a user and handled by SMM. The user can view the store directory, apply filters, and issue queries onto the store. These service requests reach SN via BTS and SA initiates the appropriate actions. During mobility, if there is a handoff, SC communicates <SID, SN> to the SC of the target BTS and the SC communicates <SID, SN> to SN1 associated with the target BTS. SN1 in turn interacts with SN to move the visitor store contents (assuming that the users home store is different from SN) from SN to the newly created visitor store. This sort of store mobility provides an efficient access to the users data. During power off, a store detach request is initiated and in response, SA updates the visitor store contents onto the home store and the session is closed. In order to avoid too much of communication to home store during a session, the objects in home store are arranged based on a Utility Value (UV) [6] and the objects with UV greater than a pre-specified threshold are brought to the visitor store during store attach requests. Also, in case during the session, if the requested object is not found in the visitor store, the same is retrieved from the home store. Many a times, too much mobility causes store thrashing. With respect to Fig. 4, if the user is highly mobile and moves from BTS11 to BTS21 to BTS12 to BTS22 and so on, then the visitor store is shifted to SN (Node 2 in Fig. 4) associated with the BSC to avoid frequent handoffs.

SON is extended to have a direct connectivity with the Web as shown in Fig. 2. This connectivity is used to pre-fetch context relevant personalized web objects and stored either in the visitor store (if the session is on) or in the home store. This direct pre-fetching of data from the Web reduces data related traffic on wireless networks and hence reduces the congestion.In a GPRS network, at various access points of hierarchy, the aggregated traffic could become a bottleneck for providing data to a user. In the proposed model, the data is made accessible through a high-speed storage network that is connected to all the access points of wireless network. Since the data is accessible at one click by a user, the user experience is greatly enhanced.

4. CONCLUSION AND FUTURE WORK

In this paper, we have provided a brief overview of SON based extended storage for mobile devices that enables faster Web access. Ongoing work is in the direction of elaborating the SON architecture and enhancing the same to support the storage of cached objects.

5. REFERENCES

  1. D. Anderson et al, The case for resilient overlay networks. Proc. of HotOSVIII, Germany, May 2001.
  2. M. Cherniack et al, Profile-Driven Cache Management. ICDE 2003 , Bangalore, India March 2003.
  3. Z. Jiang et al, Incorporating Proxy Services Into Wide Area Cellular IP Networks. Journal of Wireless Communications and Mobile Computing, July 2001.
  4. Z. Jiang et al, Link Condition Based Proxies for QoS Management in Wireless Networks. Proc.of IEEE PIMRC, CA, 2001.
  5. P. Kulkarni et al, Handling Client Mobility and Intermittent Connectivity in Mobile Web Accesses. MDM 2003, Melbourne, Australia, Jan. 2003.
  6. H. Kammanahalli et al, A Time Zone Based Dynamic Cache Replacement Policy. ICEIS 2003 , Angers, France. April 2003.
  7. Y. Villate et al, Agent-Based and Mobile External Storage for Users of Mobile Devices. Workshop on UbiAgents , Bologna, Italy, July 2002.