臺灣博碩士論文加值系統

最近的研究結合機器學習與軟體定義網絡(SDN)以支持智慧流量工程。但是,大
多數框架僅支持機器學習在遠端的控制器中,會產生明顯的信號開銷和數據轉發成本。在這項研究中,我們提出了一個新的架構,稱為網絡內推斷(INI)透過神經計算棒(NCS)實現本地學習,NCS 為 Intel 最近發布的便攜式設備,其可通過 USB 端口連接到可編程交換機。NCS 可以靈活地擴展交換機的計算能力,然而其有限的容量無法承受大量流量的實時推論需要。為了開發實用的本地學習框架,我們設計一個兩階段的學習架構,通過知識蒸餾進行本地學習共同實現低成本和準確的流量分類。我們進一步設計推論模型的部署和調整演算法,以利用配備不同功能的多個 NCS 設備切換以共享網絡的推論工作量。我們的實驗表明,兩階段架構降低了推論拒絕率達到 46.5% 並保持推論準確度為 98.10%。模擬實驗中驗證了提出的自適應模型放置方案考慮了負載平衡,從而更好地利用 NCS 服務於動態推論請求。

Recent research has incorporated machine learning with software­defined networking to support intelligent traffic engineering. However, most frameworks only enable machine learning in remote controllers, which introduce significant signaling overhead and data forwarding costs. In this work, we present a new architecture called in­network inference (INI) to realize local learning in Neural Compute Stick (NCS), a portable device that can be connected to a programmable switch via a USB port. While NCS can flexibly extend the computing power of a switch, its limited capacity however cannot afford real­time inference for enormous traffic demands. To develop a practical local learning architecture, we design a two­phase learning framework that combines local learning with knowledge distillation and remote learning to achieve lightweight but accurate traffic classification. We further design an inference model deployment and adaptation algorithm to utilize multiple NCS devices equipped with different switches to share the inference workload of a network. Our testbed experiments show that the two­phase learning framework reduces the inference rejection rate by 46.5% and maintains the inference accuracy of 98.10%. The trace­driven simulations verify that the proposed adaptive model placement scheme considers load balancing and, hence, better utilizes the computing resources of NCS to serve dynamic inference requests.

摘要 i
Abstract ii
Acknowledgement iii
Table of Contents iv
List of Figures v
List of Tables vi
1 Introduction 1
2 Related Work 3
3 INI Architecture 5
3.1 Architecture of In­Network Inference 5
4 Motivation 10
5 Adaptive Two­Phase Learning 14
5.1 Two­Phase Inference with Knowledge Distillation 14
5.2 Adaptive Model Deployment and Assignment 20
6 Testbed Experiments 25
6.1 Complexity of In­Network Inference 25
6.2 Performance of Knowledge Distillation 29
7 Large­Scale Simulation 35
8 Conclusion 45
References 46

[1] Z. A. Qazi, J. Lee, T. Jin, G. Bellala, M. Arndt, and G. Noubir, “Application­awareness in
SDN,” in ACM SIGCOMM, 2013.
[2] X. Yun, Y. Wang, Y. Zhang, and Y. Zhou, “A semantics­aware approach to the automated
network protocol identification,” IEEE/ACM Transactions on Networking, vol. 24, no. 1,
p. 583–595, Feb. 2016.
[3] J. Zhang, X. Chen, Y. Xiang, W. Zhou, and J. Wu, “Robust network traffic classification,”
IEEE/ACM Transactions on Networking, vol. 23, no. 4, pp. 1257–1270, 2015.
[4] P. Poupart, Z. Chen, P. Jaini, F. Fung, H. Susanto, Yanhui Geng, Li Chen, K. Chen, and
Hao Jin, “Online flow size prediction for improved network routing,” in IEEE ICNP, 2016.
[5] Y. Huang, W. Shih, and J. Huang, “A classification­based elephant flow detection method
using application round on SDN environments,” in Asia­Pacific Network Operations and
Management Symposium (APNOMS), 2017.
[6] Peng Xiao, Wenyu Qu, Heng Qi, Yujie Xu, and Zhiyang Li, “An efficient elephant flow
detection with cost­sensitive in sdn,” in International Conference on Industrial Networks
and Intelligent Systems, 2015.
[7] S. Chao, K. C. Lin, and M. Chen, “Flow classification for software­defined data centers
using stream mining,” IEEE Transactions on Services Computing, vol. 12, no. 1, pp. 105–
116, 2019.
[8] J. Ashraf and S. Latif, “Handling intrusion and ddos attacks in software defined networks
using machine learning techniques,” in National Software Engineering Conference, 2014.

[9] J. Bakker, B. Ng, W. K. G. Seah, and A. Pekar, “Traffic classification with machine learn-
ing in a live network,” in IFIP/IEEE Symposium on Integrated Network and Service Man-
agement (IM), 2019.

46

[10] A. Shukla, K. N. Hudemann, A. Hecker, and S. Schmid, “Runtime verification of P4
switches with reinforcement learning,” in ACM Workshop on Network Meets AI & ML,
2019.

[11] L. Freire, M. Neves, A. Schaeffer­Filho, and M. Barcellos, “POSTER: Finding vulnerabil-
ities in P4 programs with assertion­based verification,” in ACM CCS, 2017.

[12] L. Freire, M. Neves, L. Leal, K. Levchenko, A. Schaeffer­Filho, and M. Barcellos, “Un-
covering bugs in P4 programs with assertion­based verification,” in ACM SOSR, 2018.

[13] E. Cidon, S. Choi, S. Katti, and N. McKeown, “AppSwitch: Application­layer load bal-
ancing within a software switch,” in ACM Asia­Pacific Workshop on Networking (APNet),

2017.
[14] N. K. Sharma, M. Liu, K. Atreya, and A. Krishnamurthy, “Approximating fair queueing
on reconfigurable switches,” in USENIX NSDI, 2018.
[15] Y. Li, R. Miao, C. Kim, and M. Yu, “FlowRadar: A better NetFlow for data centers,” in
USENIX NSDI, 2016.
[16] P. Vörös and A. Kiss, “Security middleware programming using P4,” in Human Aspects
of Information Security, Privacy, and Trust, 2016.
[17] R. Datta, S. Choi, A. Chowdhary, and Y. Park, “P4Guard: Designing P4 based firewall,”
in IEEE MILCOM, 2018.
[18] B. Lewis, M. Broadbent, and N. Race, “P4ID: P4 enhanced intrusion detection,” in IEEE

Conference on Network Function Virtualization and Software Defined Networks (NFV-
SDN), 2019.

[19] R. Cohen, L. Lewin­Eytan, J. S. Naor, and D. Raz, “Near optimal placement of virtual
network functions,” in IEEE INFOCOM, 2015.
[20] Y. Sang, B. Ji, G. R. Gupta, X. Du, and L. Ye, “Provably efficient algorithms for joint
placement and allocation of virtual network functions,” in IEEE INFOCOM, 2017.

47

[21] R. Cziva, C. Anagnostopoulos, and D. P. Pezaros, “Dynamic, latency­optimal vNF place-
ment at the network edge,” in IEEE INFOCOM, 2018.

[22] D. Li, P. Hong, K. Xue, and j. Pei, “Virtual network function placement considering re-
source optimization and SFC requests in cloud datacenter,” IEEE Transactions on Parallel

and Distributed Systems, vol. 29, no. 7, pp. 1664–1677, 2018.
[23] B. Addis, D. Belabed, M. Bouet, and S. Secci, “Virtual network functions placement and
routing optimization,” in IEEE CloudNet, 2015.
[24] H. Feng, J. Llorca, A. M. Tulino, D. Raz, and A. F. Molisch, “Approximation algorithms
for the NFV service distribution problem,” in IEEE INFOCOM, 2017.
[25] Wei Wang, Ming Zhu, Xuewen Zeng, Xiaozhou Ye, and Yiqiang Sheng, “Malware traffic

classification using convolutional neural network for representation learning,” in Interna-
tional Conference on Information Networking (ICOIN), 2017.

[26] “DeepTraffic dataset.” [Online]. Available: https://github.com/echowei/DeepTraffic/tree/
master/1.malware_traffic_classification
[27] G. Hinton, O. Vinyals, and J. Dean, “Distilling the knowledge in a neural network,” in
NIPS Deep Learning and Representation Learning Workshop, 2015. [Online]. Available:
http://arxiv.org/abs/1503.02531

[28] S. H. Khan, M. Hayat, M. Bennamoun, F. Sohel, and R. Togneri, “Cost­sensitive learn-
ing of deep feature representations from imbalanced data,” IEEE Transactions on Neural

Networks and Learning Systems, vol. 29, pp. 3573–3587, 2018.
[29] L. Perron and V. Furnon, “Or­tools,” Google. [Online]. Available: https://developers.
google.com/optimization/

[30] M. Mitzenmacher, “The power of two choices in randomized load balancing,” IEEE Trans-
actions on Parallel and Distributed Systems, vol. 12, no. 10, pp. 1094–1104, 2001.