Cross Inference of Throughput Profiles Using Micro Kernel Network Method

Rao, Nageswara S. V.; Al-Najjar, Anees; Imam, Neena; Liu, Zhengchun; Kettimuthu, Rajkumar; Foster, Ian

doi:10.1007/978-3-030-98978-1_4

Cited by 4 publications

(5 citation statements)

References 11 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…We utilize the micro-Kernel Network (mKN) method proposed in [18] to account for the inaccurate VIT link throughput measurements, as shown in the Fig. 10.…”

Section: ) Cross-inference Of Throughput Profilesmentioning

confidence: 99%

“…The differences in measurements between VIT and physical ecosystems have a significant impact on the performance assessment, and in extremely cases could lead to inaccurate conclusions, such as I/O limit in previous section. To overcome this limitation, the mKN method [18] maps VIT measurements to ecosystem measurements by leveraging testbed measurements and strategically collected measurements on the VIT host, as described in Section IV-B3. We apply two ML methods to estimate the map f , namely, the smooth Gaussian Process Regression (GPR) and nonsmooth Ensemble of Trees (EOT) methods that represent two different approaches.…”

Section: B Network Throughput Cross-inferencementioning

confidence: 99%

“…Different mechanisms are used to specify the network latency of VIT and testbed connections. The Traffic Control Linux tool (TC) [39] and the TCLink mininet The mkN method has been tested using measurements collected on multiple laptops and desktops which are described in detail in [18]. Both estimates demonstrate the overall concave and smooth profiles t hat a re s imilar t o testbed profiles b oth i n t erms o f t hroughput values a nd shape.…”

Section: B Network Throughput Cross-inferencementioning

confidence: 99%

“…A VIT is implemented using mininet [17], and is executed either within a Virtual Machine (VM) or natively under Linux Ubuntu OS. Mininet is limited in emulating high bandwidth, long distance connections and in preserving TCP dynamics, and consequently its throughput measurements are often inaccurate [14], [18]. We implement a service for an accurate throughput profile estimation using network and input/output (I/O) measurements collected over VIT by leveraging measurements from a physical testbed with hardware network emulators and the mKN method developed in [18].…”

Section: Introductionmentioning

confidence: 99%

See 3 more Smart Citations

Virtual Infrastructure Twin for Computing-Instrument Ecosystems: Software and Measurements

Al-Najjar

Rao

2023

IEEE Access

View full text Add to dashboard Cite

This manuscript has been co-authored by UT-Battelle, LLC, under contract DE-AC05-00OR22725 with the US Department of Energy (DOE). The US government retains and the publisher, by accepting the article for publication, acknowledges that the US government retains a nonexclusive, paid-up, irrevocable, worldwide license to publish or reproduce the published form of this manuscript, or allow others to do so, for US government purposes. DOE will provide public access to these results of federally sponsored research in accordance with the DOE Public Access Plan (http://energy.gov/downloads/doe-public-access-plan).

show abstract

“…We utilize the micro-Kernel Network (mKN) method proposed in [18] to account for the inaccurate VIT link throughput measurements, as shown in the Fig. 10.…”

Section: ) Cross-inference Of Throughput Profilesmentioning

confidence: 99%

Section: B Network Throughput Cross-inferencementioning

confidence: 99%

Section: B Network Throughput Cross-inferencementioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

Virtual Infrastructure Twin for Computing-Instrument Ecosystems: Software and Measurements

Al-Najjar

Rao

2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…On each connection, TCP dynamics and throughput on host systems and switches closely match those on an equivalent physical connection, since Ethernet packets are received and processed as per connection RTT and loss rate, and delivered by the hardware IXIA emulator. This setup more accurately reflects the physical TCP flows in practice than simulators such as ns-3 and riverbed (previously OPNET), and software emulators such as mininet that are subject to host system limitations [56].…”

Section: A Network Testbed Measurementsmentioning

confidence: 99%

Game Strategies for Data Transfer Infrastructures Against ML-Profile Exploits

Rao¹,

Chris²,

He³

2022

Preprint

View full text Add to dashboard Cite

<p>Data transfer infrastructures composed of Data Transfer Nodes (DTN) are critical to supporting distributed computing and storage capabilities for clouds, data repositories, and complexes of supercomputers and instruments. The infrastructure's throughput profile, estimated as a function of the connection round trip time using Machine Learning (ML) methods, is an indicator of its operational state, and has been utilized for monitoring, diagnosis and optimization purposes. We show that the inherent statistical variations and precision of the throughput profiles estimated by ML methods can be exploited for unauthorized use of DTNs' computing and network capacity. We present a game theoretic formulation that captures the cost-benefit trade-offs between an attacker that attempts to hide under the profile's statistical variations and a provider that counters by using additional measurements at an added cost.</p> <p>The Nash equilibrium conditions of this game provide qualitative insights and bounds for the success probabilities of the attacker and provider, based on the generalization equation of the ML-estimate. We present experimental results that illustrate scenarios wherein a significant portion of DTN computing capacity is compromised without being detected by an attacker that exploits the ML estimate properties.</p>

show abstract