Web Proxy Cache Replacement Scheme Based on Backpropagation Neural Network

Part of the SpringerBriefs in Computer Science book series (BRIEFSCOMPUTER)


Web proxy caches are frequently used to reduce the strain that contemporary web traffic causes for web servers and network bandwidth providers. In this chapter, a novel approach to web proxy cache replacement is developed and analyzed. Unlike previous approaches, this approach utilizes neural networks for replacement decisions. The neural networks are trained to classify cacheable objects from real-world data sets using information known to be important in web proxy caching, such as frequency and recency. The networks are able to obtain correct classification ratios (CCRs) of between 0.85 and 0.88 both for data used for training and for data not used for training. We use the two most important metrics, hit rate and byte-hit rate, to evaluate the performance of our cache replacement technique. We provide extensive simulations using various neural network structures and cache conditions. We compare our approach to the least recently used (LRU) and least frequently used (LFU) cache replacement techniques, in addition to the optimal case which always rates an object with the number of future requests. In simulation, the final neural networks achieve hit rates that are 86.60 % of the optimal in the worst case and 100 % of the optimal in the best case. Byte-hit rates are 93.36 % of the optimal in the worst case and 99.92 % of the optimal in the best case. We provide insight into the variation of performance of the neural networks under various structures or cache parameters. We deeply examine the input-to-output mappings of the individual neural networks to determine causes for higher or lower performance under various cache conditions.


Neural network proxy cache replacement (NNPCR) Web proxies Proxy caching Neural networks Replacement algorithm Performance evaluation Training and simulation Hit rate Byte-hit rate Trace file IRCache  


  1. 1.
    R. D. Reed, R. J. Marks II, Neural smithing: supervised learning in feedforward artificial neural networks (The MIT Press, Cambridge, 1999)Google Scholar
  2. 2.
    J. de Villiers, E. Barnard, Backpropagation neural nets with one and two hidden layers. IEEE Trans. Neural Networks 4(1), 136–141 (1993)CrossRefGoogle Scholar
  3. 3.
    S. Podlipnig, L. Böszörmenyi, A survey of web cache replacement strategies. ACM Comput. Surv. 35(4), 374–398 (2003)CrossRefGoogle Scholar
  4. 4.
    G. Ganger, B. Worthington, Y. Patt and J. Bucy, The DiskSim Simulation Environment (2005), [Online document] Available Cited 16 Sept 2005
  5. 5.
    IRCache Home, [Online document] Available Cited 5 Sept 2005
  6. 6.
    S. C. Rhea, K. Liang, E. Brewer, Value-based Web caching, in Proceedings of the 12th International Conference on World Wide Web, ACM Press, pp. 619–628 (2003)Google Scholar
  7. 7.
    M. Arlitt, L. Cherkasova, J. Dilley, R. Friedrich, T. Jin, Evaluating content management techniques for Web proxy caches. ACM SIGMETRICS Perform. Eval. Rev. 27(4), 3–11 (2000)CrossRefGoogle Scholar
  8. 8.
    R. Caceres, F. Douglis, A. Feldmann, G. Glass, M. Rabinovich, Web proxy caching: the devil is in the details. ACM SIGMETRICS Perform. Eval. Rev. 26(3), 11–15 (1998)CrossRefGoogle Scholar
  9. 9.
    L. Rizzo, L. Vicisano, Replacement policies for a proxy cache. IEEE/ACM Trans. Networking 8(2), 158–170 (2000)CrossRefGoogle Scholar

Copyright information

© Hala ElAarag 2013

Authors and Affiliations

  1. 1.Department of Math and Computer ScienceStetson UniversityDeLandUSA

Personalised recommendations