Leveraging Web Prefetching Systems with Data Deduplication

Abstract

Abstract-The continued rise in internet users and the ever-greater complexity of Web content degrade userperceived latency in the Web. Web caching, prefetching and data deduplication are techniques that are used to try to mitigate this effect. Regardless of the amount of bandwidth available for network traffic, Web prefetching has the potential to consume free bandwidth to its limits. Deduplication explores data redundancies to reduce the amount of data transferred through the network, thereby freeing occupied bandwidth. To the best of our knowledge no previous work has been done which applies these two techniques combined to Web traffic. The motivation of this work is to ask if, by combining these two techniques, it is possible to significantly reduce the amount of bytes transmitted per user request, thereby improving the user-perceived latency in the Web. In the present work, we developed and implemented a system that combines the use of Web prefetching and deduplication techniques. By testing our system using real-world user navigation traces, we show that deduplication is able to substantially reduce the network costs of prefetching (up to 34% savings in transferred bytes), without degrading the latency gains due to prefetching. By adjusting deduplication parameters it is possible to improve the latency relative to when using only prefetching.

KASLR-MT: Kernel Address Space Layout Randomization for Multi-Tenant cloud systems

“…For example the Linux kernel included Kernel Samepage Merging (KSM) in the version 2.6.32 and Windows introduced Memory Page Combining in Windows 8 [20]. Furthermore, data deduplication is commonly used in other areas such as databases or web contents [3,13,25].…”

Off-line Deduplication Method for Solid-State Disk Based on Hot and Cold Data

“…However, the size of the mapping table increased due to the introduction of the secondary mapping table. A. Gupta et al [13,14] used the value locality inside SSDs to reduce their write load while realizing deduplication. However, evaluating the efficiency of deduplication is impossible because they did not discuss fingerprint generation.…”