Thepparit Banditwattanawong

Economical and efficient big data sharing with i-Cloud

Big data can be hosted on cloud and being shared distributedly through cloud services in an unprecedented volume, variety and velocity. This causes not only cloud network congestions and delayed cloud services but also increases in public cloud data-out charges. Client-side cloud cache alleviates these problems. Furthermore, cloud cache must be aware of nonuniform data-out costs when big data is stored in hybrid clouds built with different public cloud providers. Deploying i-Cloud approach as the core mechanism of cloud cache could save data-out cost up to 14.78% or 4,425 USD saved per annum based on our representative scenario, and delivered 17.24% byte-hit, 17.96% delay-saving and 29.33% cache hit outperforming LRU, GDSF and LFU-DA approaches. A main finding is that i-Cloud, learning uniform cost patterns, could perform well against nonuniform cost environment.

An Intelligent Cloud Cache Replacement Scheme

Cloud computing services heavily relies on data networks. The continuous and rapid growth of data in external private clouds accelerates downstream network-bandwidth saturation and public cloud data-out overspends. Client-side cloud caching is a solution. This paper presents the core mechanism of the cloud caching, called i-Cloud cache replacement policy. Simulation results showed that 1) i-Cloud could deliver stable performances and outperformed three well-known cache replacement policies in all standard performance metrics against almost all workloads, 2) i-Cloud could attain optimal hit and byte-hit ratios without sacrificing one to the other, 3) i-Cloud did not give performance minima if properly trained, 4) i-Cloud could perform well for longer runs than its training periods, and 5) in terms of scalability and economy, i-Cloud is suitable for small cache sizes whereas nonintelligent-mode i-Cloud suffices larger cache sizes and the realization of responsive cloud services.

Multi-provider cloud computing network infrastructure optimization

Cloud-adopting enterprises have been increasingly employing multiple cloud providers concurrently, for example, to consume unique services and to mitigate data lock-in risk. As a consequence, the enterprises must be able to address contrasting quality-of-service degrees offered by the different providers. This paper presents an intelligent cloud cache eviction approach, namely i-Cloud, as the core component of a client-side cloud cache. i-Cloud is capable of reducing public cloud data-out expenses, improving cloud network scalability and lowering cloud service access latencies specifically in multi-provider cloud environments. Trace-driven simulations have shown that i-Cloud outperformed well-known approaches in all performance metrics. In addition, i-Cloud is not only able to achieve optimal performances in all metrics simultaneously but also delivered relatively stable performances across all performance metrics. The results have also indicated that taking the nonuniformity of data-out charge rates into cache eviction decisions improved caching performances in all metrics. Byte-hit rate and hit rate could be optimal simultaneously in a nonuniform cost model.i-Cloud outperformed popular LRU, GDSF and LFUDA schemes in a nonuniform cost environment.i-Clouds performances were stable and close to those of infinite cache size.Window size had small performance effect when relative cache sizes were big.Accounting data-out charge rates improved all performance aspects at small cache sizes.

An economic model for client-side cloud caching service

Cache-as-a-Service (CaaS) benefits cloud computing users from personal to corporate levels. One of the important factors contributing to the success of CaaS at the corporation level is much lower expense compared to performance received. However, careful consideration must be given when choosing CaaS to achieve cost-effectiveness. Thus, the service model of CaaS allowing custom-made SLA will be extremely useful. Currently, almost all available CaaSes are server-side. This paper presents the flexible economic model of client-side CaaS as an early attempt in the field. The model considers both CapEx and OpEx as well as fixed costs and variable costs of the client-side CaaS.

A Cost Model for Client-Side CaaS

Deploying cache-as-a-service (CaaS) at the corporation level reduces network bandwidth expense and improves performance. Careful consideration must be given when choosing CaaS to achieve cost-effectiveness. This requires the service model of CaaS allowing custom-made SLA. This paper presents the flexible economic model of client-side CaaS as an early attempt in the field. The model has been evaluated to be promising based on a realistic scenario.

Egress Cloud Computing with Big Data Attribution

To share big data stored in cloud in a distributed manner can cause so huge data transfer out of cloud that affects cloud service responsiveness, public cloud data-out monetary costs and network bandwidth consumption. Our previous works showed that deploying cloud cache with i-Cloud eviction scheme is an effective solution. Through trace-driven simulations, this paper aims to analyze algorithmic factors influencing the performance of i-Cloud. A main finding is that TTL attribute creates the greatest contribution in i-Cloud efficiency in distributed big data sharing. In addition, this paper presents a guideline for content providers to allow the distributed sharing of their big data in an effective manner.