Cloud data storage: a queueing model with thresholds
- PDF / 1,051,238 Bytes
- 21 Pages / 439.37 x 666.142 pts Page_size
- 39 Downloads / 216 Views
Cloud data storage: a queueing model with thresholds Apoorv Saxena1
· Dieter Claeys2,3 · Bo Zhang4 · Joris Walraevens1
© Springer Science+Business Media, LLC, part of Springer Nature 2019
Abstract In the past decade, cloud platforms have become a standard across the industry for data storage and operations. Such platforms offer high quality of service in terms of reliability and ease of setup at an effective cost. With exponentially high rates of increase of data storage requirements, data is now increasingly stored in clouds. However, there are limited studies which analyze the processes performing the storage operations. Queueing models offer a very natural way of modeling these storage processes. The data packets waiting for storage form a queue which is served by a storage server. Since data packets are transmitted to the cloud in batches for efficiency, this storage server is modelled as a batch server. The storage server goes into sleep mode in between data transmission periods which are, in turn, modelled as vacations. The storage service is resumed after a vacation if there are enough packets in backlog or enough time has elapsed since last storage. This is modelled as restarting thresholds in our model. Analyzing this model helps us evaluate the quality of service (QoS) of storage processes in terms of measures such as backlog size and probability of a new connection to cloud server. These measures are then used to define a user cost function and QoS constraints, and compute optimal storage parameters. Keywords Cloud data storage · Stochastic modeling · Queueing theory · Batch service
B
Apoorv Saxena [email protected] Dieter Claeys [email protected] http://www.FlandersMake.be Bo Zhang [email protected] Joris Walraevens [email protected]
1
SMACS Research Group, Department TELIN, Ghent University, Sint-Pietersnieuwstraat 41, 9000 Ghent, Belgium
2
Department of Industrial Systems Engineering and Product Design, Ghent University, Technologiepark 46, 9052 Zwijnaarde, Belgium
3
Industrial Systems Engineering (ISyE), Flanders Make, Lommel, Belgium
4
LYFT, New York, USA
123
Annals of Operations Research
1 Introduction In the past few years, organizations have moved from local storage systems to cloud storage systems. Moreover, new businesses are building their infrastructure using the services of public cloud system providers. Big corporations such as Amazon (2018a), Microsoft (2018), IBM (2018) etc have built their own data infrastructures and offer their platform to users on a lease, also known as IAAS in cloud computing (see Kumar et al. 2014). Such systems not only reduce the investment in the setup and maintenance of storage systems locally, they also promise remarkable benefits such as high reliability, availability, ease of set up etc (see Chang and Wills 2016). Recent studies have shown that the amount of data generated is going to increase exponentially. The International data corporation, in Reinsel et al. (2017), have estimated that the amount of data generated would re
Data Loading...
