A Computational Approach to Packet Classification
نویسندگان
چکیده
Multi-field packet classification is a crucial component in modern software-defined data center networks. To achieve high throughput and low latency, state-of-the-art algorithms strive to fit the rule lookup structures into on-die caches; however, they do not scale well with number of rules. We present novel approach, NuevoMatch , which improves memory scaling existing methods. A new structure, xmlns:xlink="http://www.w3.org/1999/xlink">Range Query Recursive Model Index (RQ-RMI), key that enables NuevoMatch replace most accesses main model inference computations. describe an efficient training algorithm guarantees correctness RQ-RMI-based classification. The use RQ-RMI allows rules be compressed neural networks hardware cache. Further, it takes advantage growing support for fast network processing CPUs, such as wide vector instructions, achieving latency tens nanoseconds per lookup. Our evaluation using 500K multi-field from standard ClassBench benchmark shows geometric mean compression factor $4.9\times $ notation="LaTeX">$8\times notation="LaTeX">$82\times average performance improvement notation="LaTeX">$2.4\times notation="LaTeX">$2.6\times notation="LaTeX">$1.6\times compared CutSplit, NeuroCuts, TupleMerge, all algorithms.
منابع مشابه
A Set-based Approach to Packet Classification
Firewalls, and packet classification in general, are becoming more and more significant as data rates soar and hackers become increasingly sophisticated and more forceful. In this paper, we present a new packetclassification approach that uses set theory to classify packets. This approach has significant theoretical advantages over current approaches. We demonstrate its practicality by implemen...
متن کاملA Modular Approach to Packet Classification: Algorithms and Results
The ability to classify packets according to pre-defined rules is critical to providing many sophisticated value-added services, such as security, QoS, load balancing, traffic accounting, etc. Various approaches to packet classification have been studied in the literature with accompanying theoretical bounds. Practical studies with results applying to large number of filters (from 8K to 1 milli...
متن کاملApplying Two Computational Classification Methods to Predict the Risk of Breast Cancer: A Comparative Study
Introduction: Lack of a proper method for early detection and diagnostic errors in medicine are some fundamental problems in treating cancer. Data analysis techniques may significantly help early diagnosis. The current study aimed at applying and evaluating neural networks and decision tree algorithm on breast cancer patients’ data for early cancer prediction. Methods: In the current stu...
متن کاملApplying Two Computational Classification Methods to Predict the Risk of Breast Cancer: A Comparative Study
Introduction: Lack of a proper method for early detection and diagnostic errors in medicine are some fundamental problems in treating cancer. Data analysis techniques may significantly help early diagnosis. The current study aimed at applying and evaluating neural networks and decision tree algorithm on breast cancer patients’ data for early cancer prediction. Methods: In the current stu...
متن کاملذخیره در منابع من
با ذخیره ی این منبع در منابع من، دسترسی به آن را برای استفاده های بعدی آسان تر کنید
ژورنال
عنوان ژورنال: IEEE ACM Transactions on Networking
سال: 2022
ISSN: ['1063-6692', '1558-2566']
DOI: https://doi.org/10.1109/tnet.2021.3131879