Overview:

The project develops novel solutions that enhances throughput in wireless networks by aggregating multiple radios that coexist at a wireless device. It consists of two independent and complementary approaches: Hydra-Glia [1,2] for homogeneous radios and Hydra-Super-Aggregation [3,4] for heterogeneous radios. Both of them are software-only solutions that can be deployed with offthe-shelf hardware and are compatible with legacy systems.

Impact:

Aggregating homogeneous off-the-shelf radios that operate on orthogonal channels doesn't yield proportional aggregate throughput. Our experimental results show that aggregating 12 Wi-Fi a/g radios yield 70Mbps as opposed to 480Mbps (12 x 40Mbps each). Hydra-Glia is able to effectively exercise an array of Wi-Fi radios and achieve close to ideal aggregate performance (6.5x throughput).

Most wireless devices today are equipped with multiple heterogeneous radios, such as Wi-Fi and 3G/4G. Aggregating heterogeneous radios only give marginal throughput improvement (merely 3%) due to the disparity in the data rate. Hydra-Super-Aggregation is able to intelligently exercise heterogeneous radios to provide aggregate throughput higher than the sum of the individual parts, and it demonstrates up to 3x throughput in the test-bed.

Concept:

Hydra-Glia is a layer-2.5 software solution that consists of two design principles in aggregating homogeneous radios. Act-as-One coarsely bonds individual radios and creates a single logical radio that uses all available channels simultaneously. Exploit-the-Many allows the right radio-channel association for both the transmitting and receiving radios such that the maximum aggregate throughput can be achieved.

Hydra-Super-Aggregation is a layer-3.5 software solution that consists of three design principles in aggregating heterogeneous radios. Each design principle explicitly leverages heterogeneous characteristics of Wi-Fi and 3G/4G to improve TCP throughput. Selective Offloading eliminates self-contention in the Wi-Fi link, Proxying uses 3G to act on behalf of the Wi-Fi link under blackouts, and Mirroring prevents the adverse effects of random wireless loss by retrieving lost data via 3G.

Results:

Hydra-Glia is prototyped with madwifi driver and Linux OS. The testbed consists of two desktops that are equipped with 12 Atheros-based Microtik R52 802.11a/b/g miniPCI cards (Figure 1). The solution is able to achieve 465 Mbps, which is very close to the ideal 480 Mbps (Figure 2).

Figure 1: Hydra-Glia prototype

Figure 2: Hydra-Glia performance

Hydra-Super-Aggregation is prototyped with a user-space TCP implementation and Linux OS. The testbed include the laptop and the smartphone platforms, and both are equipeed with a Wi-Fi radio and a 3G radio (Figure 3). Hydra-Super-aggregation principles improves TCP throughput in WiFi network by 30% to 200% (Figure 4).

Figure 3: Hydar-Super-Aggregation prototype

Figure 4: Hydra-Super-Aggregation performance

Patents and Publications:

1. R. Sivakumar and S. Kakumanu, "HYDRA: Achieving High Data-Rate Wireless Communication Using Multiple 802.11 Radios," patent filed in 2010.
2. S. Kakumanu and R. Sivakumar, "Glia: A Practical Solution for Effective High Datarate Wifi-Arrays," ACM International Conference on Mobile Computing and Networking (MOBICOM), Beijing, China, Sept. 20-25, 2009.
3. R. Sivakumar and C.-L. Tsao, "Systems and Methods for Improved Wireless Interface Aggregation," patent filed in 2010.
4. C.-L. Tsao and R. Sivakumar, "On Effectively Exploiting Multiple Wireless Interfaces in Mobile Hosts," ACM International Conference on Emerging Networking Experiments and Technologies (CoNEXT), Rome, Italy, December 1-4, 2009.

People:

• Sandeep Kakumanu (Student)
• Cheng-Lin Tsao (Student)
• Raghupathy Sivakumar (Professor)