More Like this

1. Minimizing network bandwidth under latency constraints: The single node case

   Jiayi Song, Roch Guérin (September 2021, Proceedings 33rd International Teletraffic Congress (ITC 33))

   null (Ed.)

   Much of today's traffic flows between datacenters over private networks. The operators of those networks have access to detailed traffic profiles with performance goals that need to be met as efficiently as possible, e.g., realizing latency guarantees with minimal network bandwidth. Of particular interest is the extent to which traffic (re)shaping can be of benefit. The paper focuses on the most basic network configuration, namely, a single link network, with extensions to more general, multi-node networks discussed in a companion paper. The main results are in the form of optimal solutions for different types of schedulers of varying complexity. They demonstrate how judicious traffic shaping can help lower complexity schedulers perform nearly as well as more complex ones. 

   more » « less
2. On the Benefits of Traffic “Reprofiling” the Multiple Hops Case—Part I

   https://doi.org/10.1109/TNET.2024.3392030  

   Qiu, Jiaming; Song, Jiayi; Guérin, Roch; Sariowan, Henry (August 2024, IEEE/ACM Transactions on Networking)

   This paper considers networks where user traffic is regulated through deterministic traffic profiles, e.g., token buckets, and requires hard delay bounds. The network's goal is to minimize the resources it needs to meet those bounds. The paper explores how reprofiling, i.e., proactively modifying how user traffic enters the network, can be of benefit. Reprofiling produces "smoother" flows but introduces an up-front access delay that forces tighter network delays. The paper explores this trade-off and demonstrates that, unlike what holds in the single-hop case, reprofiling can be of benefit even when "optimal" schedulers are available at each hop. 

   more » « less
3. Expanding across time to deliver bandwidth efficiency and low latency

   William M. Mellette, Rajdeep Das (January 2020, USENIX Symposium on Networked Systems Design and Implementation (NSDI))

   Datacenters need networks that support both low-latency and high-bandwidth packet delivery to meet the stringent requirements of modern applications. We present Opera, a dynamic network that delivers latency-sensitive traffic quickly by relying on multi-hop forwarding in the same way as expander-graph-based approaches, but provides near-optimal bandwidth for bulk flows through direct forwarding over time-varying source-to-destination circuits. Unlike prior approaches, Opera requires no separate electrical network and no active circuit scheduling. The key to Opera's design is the rapid and deterministic reconfiguration of the network, piece-by-piece, such that at any moment in time the network implements an expander graph, yet, integrated across time, the network provides bandwidth-efficient single-hop paths between all racks. We show that Opera supports low-latency traffic with flow completion times comparable to cost-equivalent static topologies, while delivering up to 4x the bandwidth for all-to-all traffic and supporting up to 60% higher load for published datacenter workloads. 

   more » « less
4. Achieving Virtual Network Function Load-Balanced Flow Migration in Dynamic Cloud Data Centers

   Aguilera, Phillip; Gonzalez, Christopher; Tang, Bin (March 2021, Proceedings of the First Computer Science Conference for CSU Undergraduates (CSCSU 2021).)

   null (Ed.)

   Virtual Network Functions (VNFs) are software implementation of middleboxes (MBs) (e.g., firewalls) that provide performance and security guarantees for virtual machine (VM) cloud applications. In this paper we study a new flow migration problem in VNF-enabled cloud data centers where the traffic rates of VM flows are constantly changing. Our goal is to minimize the total network traffic (therefore optimizing the network resources such as bandwidth and energy) while considering that VNFs have limited processing capability. We formulate the flow migration problem and design two efficient benefit-based greedy algorithms. The simulations show that our algorithms are effective in reducing the network traffic as well as in achieving load balance among VNFs. In particular, our flow migration algorithms can reduce upto 15% network traffic compared to the case without flow migration. 

   more » « less
5. Envisioning a Unified Programmable Dataplane to Monitor Slow Attacks

   https://doi.org/10.1109/ICNP61940.2024.10858535  

   Wei, Cuidi; Tu, Shaoyu; Hasegawa, Toru; Koizumi, Yuki; Ramakrishnan, K K; Takemasa, Junji; Wood, Timothy (October 2024, 2024 IEEE 32nd International Conference on Network Protocols (ICNP) Workshop)

   Recent work shows that programmable switches can effectively detect attack traffic, such as denial-of-service attacks in the midst of high-volume network traffic. However, these techniques primarily rely on sampling or sketch-based data structures, which can only be used to approximate the characteristics of dominant flows in the network. As a result, such techniques are unable to effectively detect low-volume attacks that stealthily add only a few packets to the network. Our work explores how the combination of programmable switches, Smart network interface cards, and hosts can enable fine-grained analysis of every flow in a network, even those with only a small number of packets. We focus on analyzing packets at the start of each flow, as those packets often can help indicate whether a flow is benign or suspicious. We propose a unified architecture that spans the full programmable dataplane to take advantage of the strengths of each type of device. We are developing new filter data structures to efficiently track flows on the switch, dataplane-based communication protocols to quickly coordinate between devices, and caching approaches on the SmartNIC that help minimize the traffic load reaching the host. Our preliminary prototype can handle the full pipe bandwidth of 1.4 Tbps of traffic entering the Tofino switch, forward only 20 Gbps to the SmartNIC, and minimize the traffic load to 5 Gbps reaching the host due to our efficient flow filter, packet batching, and SmartNIC-based cache. 

   more » « less