Dual-Resource TCP/AQM for Processing-Constrained Networks

Abstract

Abstract-This paper examines congestion control issues for TCP flows that require in-network processing on the fly in network elements such as gateways, proxies, firewalls and even routers. Applications of these flows are increasingly abundant in the future as the Internet evolves. Since these flows require use of CPUs in network elements, both bandwidth and CPU resources can be a bottleneck and thus congestion control must deal with "congestion" on both of these resources. In this paper, we show that conventional TCP/AQM schemes can significantly lose throughput and suffer harmful unfairness in this environment, particularly when CPU cycles become more scarce (which is likely the trend given the recent explosive growth rate of bandwidth). As a solution to this problem, we establish a notion of dual-resource proportional fairness and propose an AQM scheme, called DualResource Queue (DRQ), that can closely approximate proportional fairness for TCP Reno sources with in-network processing requirements. DRQ is scalable because it does not maintain perflow states while minimizing communication among different resource queues, and is also incrementally deployable because of no required change in TCP stacks. The simulation study shows that DRQ approximates proportional fairness without much implementation cost and even an incremental deployment of DRQ at the edge of the Internet improves the fairness and throughput of these TCP flows. Our work is at its early stage and might lead to an interesting development in congestion control research.

Processor-Network Speed Scaling for Energy–Delay Tradeoff in Smartphone Applications

“…Workloads of each application demand different CPU processing resources. We call this notion as processing density (in cycles/bit) for NA, for NNA which are defined as the average number of CPU cycles required per bit when the application is processed by the CPU [31]. Smartphones with DVFS capability adjust CPU speed (in cycles/ ) 2 every time slot depending on the system policy [6].…”

“…Then, intuitively the time average of the process can be expressed as a sum over the steady-state probabilities as follows: (31) where is the steady-state probability of . Since the probability vector can be represented by sequence by a law of large number, there exist randomized control policies which support any arrival rate within capacity region.…”

SLA-based resource allocation in cluster computing systems

“…It is non-decreasing and can be inverted, i.e., ψ −1 exists. For instance, suppose that a cluster represents a group of CPUs in the SLA-based service model as discussed in [25]. Then, ψ(n) can be seen as a CPU scaling factor for the number of CPUs from 1 to n. According to [7], ψ(n) = ξ log 2 n , where ξ is a basic scaling factor from 1 CPU to 2.…”

Optimal control of DDoS defense with multi-resource max-min fairness

“…Because above methods are only for bandwidth resource, they may fail in case of CPU-consuming attacks, e.g., ping flood and udp flood. And current QoS researches suggest taking both bandwidth and processing time into consideration [14,15]. To improve defense effectiveness, we propose FFDRF.…”