Overview:

Applications on mobiles (Ex: smartphones), when compared to their counterparts on a desktop, are constrained primarily because of two reasons: a) poor wireless connectivity and b) poor device resources in terms of real-estate for keyboard/display, processing and memory. In this context, we present an application-aware solution suite to overcome the typical behavioral problems of applications on mobiles. The solution suite comprises of design principles that interact/modify the transport and application layers of the network stack.

Impact:

We argue that the behavior of applications all but negates any performance improvements achieved by optimized application-unaware transport protocols. We present an application-aware yet application transparent middleware solution, known as application-aware acceleration (A³) [1, 2] that offsets the typical behavioral problems of real-life applications through an effective set of principle and design elements. Delving into the problem of poor device resources in mobiles, we identify the effects of limited memory on TCP behavior and present an adaptive flow control for TCP [3]. Delving even further, we focus on the burden a user faces while interacting with applications on a mobile. This burden leads to user frustration and some applications not being used at all on mobiles. We present Mobiscope, as an application layer solution to reduce user burden.

Concept:

In A³, we identify various behavioral problems of applications, such as thin session control messages, batched data fetches, non-prioritization of data, and non-use of data reduction techniques. We present design principles for handling each of the identified behavioral problems namely transaction prediction, redundant and aggressive retransmissions, prioritized fetching and applicationaware compression. Traditional flow control in TCP is very simple because it assumes no major constraints in memory which is a big issue in mobiles. We handle this problem with an adaptive flow control algorithm for TCP that relies not just on the available buffer space, but also on the application read-rate at the receiver. For reducing user burden in accessing applications on a mobile, we present operator aggregation as the core mechanism in Mobiscope. Using operator aggregation a user can replace multiple operators involved in executing a task with a single operator on the mobile.

Results:

Figure 1 shows the A³ testbed, and Figure 2 compares A3 performance with wireless optimizations of TCP for simple mail transfer protocol (SMTP). Similar performance improvements can be observed for other specific applications protocols such as CIFS, HTTP etc using A³.

Figure 1: A3 Deployment Model

Figure 2: Comparing A3 performance with TCP flavors

Figure 3: Throughput performance with Adaptive Flow Control

Figure 4: Mean opinion score using Mobiscope vs Default

Publications:

1. Z. Zhuang, T.-Y. Chang, R. Sivakumar, and A. Velayutham, "Application-Aware Acceleration for Wireless Data Networks: Design Elements and Prototype Implementation," IEEE Transactions on Mobile Computing, vol. 8, no. 9, September 2009.
2. Z. Zhuang, T.-Y. Chang, R. Sivakumar, and A. Velayutham, "A3: Application-Aware Acceleration for Wireless Data Networks," ACM International Conference on Mobile Computing and Networking (MOBICOM), Los Angeles, CA, USA, September 2006.
3. S. Sanadhya and R. Sivakumar, "Adaptive Flow Control for TCP on Mobile Phones," IEEE International Conference on Computer Communications (INFOCOM), Shanghai, China, April, 2011.

People:

• Sandeep Kakumanu (Student)
• Cheng-Lin Tsao (Student)
• Shruti Sanadhya (Studen)
• Zhenyun Zhuang (Alumni)
• Raghupathy Sivakumar (Professor)