In enterprise networks, policies (e.g., QoS or security) are often defined based on the categorization of hosts along dimensions, such as the organizational role of the host (faculty versus student) and department (engineering versus sales). While current best practices (virtual local area networks) help when hosts are categorized along a single dimension, policy may often need to be expressed along multiple orthogonal dimensions. In this paper, we make three contributions. First, we argue for attribute-encoded IPs (ACIPs), where the IP address allocation process in enterprises considers attributes of a host along all policy dimensions. ACIPs enable flexible policy specification in a manner that may not otherwise be feasible owing to the limited size of switch rule-tables. Second, we present Alpaca, algorithms for realizing ACIPs under practical constraints of limited-length IP addresses. Our algorithms can be applied to different switch architectures, and we provide bounds on their performance. Third, we demonstrate the importance and viability of ACIPs on data collected from real campus networks.

Traffic often needs to be split over multiple equivalent backend servers, links, paths, or middleboxes. For example, in a load-balancing system, switches distribute requests of online services to backend servers. Hash-based approaches like Equal-Cost Multi-Path (ECMP) have low accuracy due to hash collision and incur significant churn during update. In a Software-Defined Network (SDN) the accuracy of traffic splits can be improved by crafting a set of wildcard rules for switches that better match the actual traffic distribution. The drawback of existing SDN-based traffic-splitting solutions is poor scalability as they generate too many rules for small rule-tables on switches. In this paper, we propose Niagara, an SDN-based traffic-splitting scheme that achieves accurate traffic splits while being extremely efficient in the use of rule-table space available on commodity switches. Niagara uses an incremental update strategy to minimize the traffic churn given an update. Experiments demonstrate that Niagara (1) achieves nearly optimal accuracy using only 1.2%--37% of the rule space of the current state-of-art, (2) scales to tens of thousands of services with the constrained rule-table capacity and (3) offers nearly minimum churn.

In enterprise networks, policies (e.g., QoS or security) are often defined based on the categorization of hosts along dimensions such as the organizational role of the host (faculty vs. student), and department (engineering vs. sales). While current best practices (VLANs) help when hosts are categorized along a single dimension, policy may often need to be expressed along multiple orthogonal dimensions. In this paper, we make three contributions. First, we argue for Attribute-Carrying IPs (ACIPs), where the IP address allocation process in enterprises considers attributes of a host along all policy dimensions. ACIPs enable flexible policy specification in a manner that may not otherwise be feasible owing to the limited size of switch rule-tables. Second, we present Alpaca, algorithms for realizing ACIPs under practical constraints of limited-length IP addresses. Our algorithms can be applied to different switch architectures, and we provide bounds on their performance. Third, we demonstrate the importance and viability of ACIPs on data collected from real campus networks.

Debugging faults in complex networks often requires capturing and analyzing traffic at the packet level. In this task, datacenter networks (DCNs) present unique challenges with their scale, traffic volume, and diversity of faults. To troubleshoot faults in a timely manner, DCN administrators must a) identify affected packets inside large volume of traffic; b) track them across multiple network components; c) analyze traffic traces for fault patterns; and d) test or confirm potential causes. To our knowledge, no tool today can achieve both the specificity and scale required for this task.

Software Defined Networks (SDNs) support diverse network policies by offering direct, network-wide control over how switches handle traffic. Unfortunately, many controller platforms force applications to grapple simultaneously with end-to-end connectivity constraints, routing policy, switch memory limits, and the hop-by-hop interactions between forwarding rules. We believe solutions to this complex problem should be factored in to three distinct parts: (1) high-level SDN applications should define their end-point connectivity policy on top of a "one big switch" abstraction; (2) a mid-level SDN infrastructure layer should decide on the hop-by-hop routing policy; and (3) a compiler should synthesize an effective set of forwarding rules that obey the user-defined policies and adhere to the resource constraints of the underlying hardware. In this paper, we define and implement our proposed architecture, present efficient rule-placement algorithms that distribute forwarding policies across general SDN networks while managing rule-space constraints, and show how to support dynamic, incremental update of policies. We evaluate the effectiveness of our algorithms analytically by providing complexity bounds on their running time and rule space, as well as empirically, using both synthetic benchmarks, and real-world firewall and routing policies.

The computation core of many data-intensive applications can be best expressed as matrix computations. The MadLINQ project addresses the following two important research problems: the need for a highly scalable, efficient and fault-tolerant matrix computation system that is also easy to program, and the seamless integration of such specialized execution engines in a general purpose data-parallel computing system.

A Software Defined Network (SDN) enforces network-wide policies by installing packet-handling rules across a distributed collection of switches. Today's SDN platforms force programmers to decide how to decompose a high-level policy into the low-level rules in each switch. We argue that future SDN platforms should support automatic transformation of policies by moving, merging, or splitting rules across multiple switches. This would simplify programming by allowing programs written on one abstract switch to run over a more complex network topology, and simplify analysis by consolidating a policy spread over multiple switches into a single list of rules. This poster presents our ongoing work on a sound and complete set of axioms for policy transformation, to enable rewriting of rules across multiple switches while preserving the forwarding policy. These axioms are invaluable for creating and analyzing algorithms for optimizing the rewriting of rules.