The ultimate goal of virtual reality (VR) is to create an experience indistinguishable from actual reality. To provide such a "life-like" experience, (i) the VR headset (VRH) should be wireless so that the user can move around freely, and (ii) the wireless link, connecting the VRH to a high-performance renderer, should support high data rates (tens to hundreds of Gbps). Industry is already pushing towards such wireless VRHs; however, these wireless links can only support a few Gbps rates. In general, current radio-frequency (RF) links (including mmWave) are not able to provide desired data rates. In this paper, we build a system, we call Cyclops, which uses free-space optical (FSO) technology to create a high-bandwidth VR wireless link. FSO links are capable of very high data rates (up to Tbps) due to the high frequencies of light waves and narrow beams. The main challenges in developing an effective FSO link are: (i) designing a link with sufficient movement tolerance, and (ii) developing a viable tracking and pointing (TP) mechanism which maintains the link while the VRH moves. As traditional TP approaches seem infeasible in our context, we develop a novel TP approach based on learning techniques, leveraging the VRH's inbuilt tracking system. We build robust 10 Gbps and 25Gbps link prototypes from commodity components, demonstrate their viability for expected movement speeds of a VRH, and show that, with certain custom-built components, we can support much higher movement speeds and bandwidths.

Circuit switched networks with high-bandwidth links are essential to handling ever increasing traffic demands in today's data centers. As these networks incur a non-trivial reconfiguration delay, they are mainly suited for bursty traffic or large flows. To address the reconfiguration delay vs. high-bandwidth tradeoff in circuit networks, an essential traffic scheduling problem is to determine a sequence of network configurations to optimally serve a given traffic. Recent works have addressed this scheduling problem for one-hop traffic in fully-connected circuit networks.

The virtual network embedding (VNE) problem of mapping virtual network (VN) requests to a substrate network is a key component of network virtualization in datacenters. In a bid to improve datacenter network's performance and cost, there has been recent interest in "reconfigurable" network architectures, wherein the network topology can be changed at runtime to better handle current traffic patterns. Such reconfigurable networks seem naturally well-suited for efficient network virtualization- as networks can be "tailored" to accommodate the incoming VN requests. Motivated by the above, in this paper, we address the problem of virtual network embedding in reconfigurable networks; to the best of our knowledge, this has not been addressed before. In particular, we address the VNE problem in reconfigurable networks under two different models of VN link demands: fixed-bandwidth and stochastic-bandwidth demands. The former is the traditional model, while we propose the the latter to improve network utilization and leverage the runtime reconfiguration capability of reconfigurable networks. For the stochastic demand model, we employ a novel concept of embedding with "runtime-binding", wherein the embedding of a VN link is "configured" at runtime (via network reconfiguration) depending on the prevailing network state and traffic. We evaluate the efficiency of our proposed models and techniques via simulation using real VN requests and traffic statistics from large datacenters, and show that our proposed models and techniques offer significant performance advantages (up to 30-40%) over traditional models.

Expected increase in cellular demand has pushed recent interest in picocell networks which have reduced cell sizes (100-200m or less). For ease of deployment of such networks, a wireless backhaul network is highly desired. Since RF-based technologies are unlikely to provide the desired multi-gigabit data rates, we motivate and explore use of free space optics (FSO) for picocell backhaul. In particular, we present a novel network architecture based on steerable links and sufficiently many robust short-range links, to help circumvent the key challenge of outdoor effects in reliable operation of outdoor FSO links. Our architecture is motivated by the fact that, due to the high density of picocells, many short-range links will occur naturally in a picocell backhaul. Moreover, use of steerable FSO links facilitates networks with sufficient redundancy while using only a small number of interfaces per node. We address the key problems that arise in the context of such a backhaul architecture, viz., an FSO link design with desired characteristics, and related network design and management problems. We develop and evaluate a robust 100m FSO link prototype, and simulate the proposed architecture in many metro US cities while show its viability via evaluation of key performance metrics.

To overcome the shortcomings of traditional static (wired) data center architectures, we recently proposed FireFly, a fully-flexible and fully-wireless data center network fabric based on free space optical (FSO) links. To facilitate a clear line-of-sight between the FSO devices placed on racks, FireFly uses a full-ceiling mirror for beam redirection. Use of a full-ceiling mirror imposes significant operational and infrastructural challenge and expense. The focus of our paper is to propose and evaluate alternative schemes to provide line-of-sight for FSO links in FireFly. In particular, we propose two schemes: (i) strategically placed multiple but small overhead mirrors, and (2) “towers” on top of racks, on which FSOs can be placed at regular heights. We develop comprehensive techniques for the above suggested schemes, and demonstrate the viability of these schemes by evaluating their performance using simulations for various performance metrics of interest.