Wireless is an increasingly dominant communication medium.
The continued quest for wireless connectivity in a multitude of mobile devices,
along with the emerging bandwidth hungry applications, has resulted in a huge growth of the wireless traffic.
Multiple-Input Multiple-Output (MIMO) is considered the dominant technology to provide
gigabit wireless links, and to accommodate the increasing demand of speed over wireless.
By using multiple transmit and receive antennas, MIMO can support more reliable and faster
communication. But how efficient are the current MIMO systems?
Our experiments with commodity MIMO 802.11n devices reveal that, the current MIMO wireless is low speed
and energy hungry. The fundamental reason for MIMO devices' poor performance is the use of legacy 802.11a/b/g, single antenna designs
over the multiple antenna, MIMO 802.11n setting. Specifically, the existing designs used over the new MIMO 802.11n devices,
are oblivious to MIMO unique communication characteristics. They do not also consider that, MIMO speed comes
at the cost of increased power consumption, proportional to the number of antennas.
In order to investigate solutions to these problems, this dissertation first experimentally studies the unique features
of MIMO wireless and their impact on existing designs' performance. Then, it revises the key mechanisms that control speed
and energy over MIMO wireless, named Rate Adaptation, and MIMO Energy Save, and develops three systems.
History-Aware Robust Rate Adaptation (HA-RRAA) is our first step towards gigabit wireless.
It opportunistically selects the best goodput PHY transmission rate for legacy 802.11a/b/g networks
by introducing novel mechanisms to capture short-term channel dynamics.
Different from HA-RRAA, our MIMO Rate Adaptation (MiRA) proposal, seeks to identify the best goodput PHY
transmission rate in MIMO 802.11n networks by considering the unique features of MIMO.
Finally, MIMO Energy Save seeks to select the optimal antenna setting at runtime to minimize energy consumption.
Our proposals depart from existing designs in three fundamental ways.
They manage the unique MIMO communication modes in a distinct manner.
They consider new metrics, to capture the tradeoffs between speed and power consumption.
Our proposals also apply novel learning mechanisms to capture the wireless channel dynamics.
There are three main contributions in this dissertation. First, it builds a strong connection between wireless communication
theory and wireless system design. Specifically, this dissertation provides the first experimental study of fundamental
MIMO wireless communication tradeoffs (i.e. diversity vs. spatial multiplexing MIMO modes, speed vs. number of antennas)
using 802.11n standard-compliant commodity testbeds. Then, it uncovers their impact on existing designs' performance.
Second, it proposes novel and practical rate adaptation and energy save designs that consider MIMO unique characteristics, and are able deliver
high performance gains. Third, this dissertation provides the first implementation and evaluation of MIMO rate adaptation and energy save
using 802.11n standard-compliant commodity devices. The high performance gains in real world settings make our proposals
a significant step towards gigabit and green wireless networks.