Partial Review-2 On the paper Above the clouds: A Berkeley View of Cloud Computing By Nikhil Ramteke Sr. No.- 07125
6. Cloud Computing Economics Observation in Cloud Economics mainly concerns with following 3 points. Elasticity Hardware changes Variable spikes and their requirement meet. Shifting the risk Being sure about to meet the elasticity with own datacenter is critical. So being on cloud shifts this risk to cloud providers. Sometimes, there is no up-front requirements, at that time money can be invested in business model. New hardware loading in own datacenter can take weeks to load in, whereas in cloud, acquiring hardware is matter of minutes, If under provisioned, provider may loss the user permanently and if over provisioned then most of the time machine will be idle wasting resources and energy. Shifting of risk to cloud may be chargeable, but assuming the hazards, one can afford it. If own datacenter is provisioned, then load should be 0.6-0.8, thus it requires over-provision, where as in cloud just apy for what you have used. In fact above points enhances the concept of scaling in clouds If hardware changes it reflects automatically for all cloud buyers because cloud provider updates configuration time to time to provide good service. Computing Cost: Should I move to cloud? While moving to cloud following important factors should be considered. o Pay separately for each resource you are using, calculate the estimate and compare with own datacenter cost. o Estimate the cost of power, cooling and physical plant establishment and see how much profitable it is.
o Operation cost of own datacenter and being on cloud should also be compared. 7. Top 10 Obstacles and Opportunities for cloud computing Availability of service Service of could should be available 24x7. Replication of data and elasticity may be helpful Data lock-in Due to non-availability of software stack in could computing, its not guaranteed for free data-lock-in operation. Solution may be to keep data on different clouds so as failure of one can be overcome by the other. Or maybe we can use cloud only when big tasks needs it otherwise run rest of task at own private cloud. Confidentiality and auditability Clouds are public and thus have more opportunities for attackers. Solution can be encrypted storage, firewalls, filters, and virtual LAN, prohibits some illegal action. Auditability provides restriction of data in a circle or region making it secure from outsiders. Data transfer bottlenecks Sending data through network is chargeable. It is not always feasible to send data through network, its good sometime to send hard disks physically saving money and time. But cloud users may diver due to charges applicable or spend during transformation. So cloud provides should find some good attractive reason to attract SaaS providers to keep their data on cloud. WAN cost should be reduced. (2/3 rd of the charge is alone of routers) WAN bandwidth in intra-node processing will be useful.
Performance Unpredictability CPU shearing is managed but I/O shearing in cloud computing is problematic. Solution can be to improve OS architecture for virtualization of interrupts and IO. Try to use flash memory, it decreases the IO interference. In case of high performance computing, scheduling g of VM s for batch processing becomes problem. Solution can be use of gang scheduling. Scalable storage Data storage should support various type of complex data structure. It is problematic for scalability (still open research problem) Bugs in Large Scale Distributed Systems Big challenge to remove bus from big Distributed systems. Solution may be use VM s because it may process valuable information that can t be processed without it. Scaling quickly Cloud provides the scalability with assurance to add resources in minutes. Some cloud providers do it automatically. It needs machine intelligence to diagnosis and scale up or scale down automatically. Idle machine also uses 2/3 rd of the power so if it is idle then shutdown/restart mechanism should be developed to save power and money. Reputation and fate shearing Maintaining reputation and assurance to customer for being with them is one of the most important tasks. Software licensing Application uses software form the cloud. If it is licensed, then users are supposed to pay charges according pay as you use strategy.
8. Conclusion Some cloud providers use free software moreover because others are not compatible with their architecture and support of their clouds. Due to elasticity model of cloud, it become possible to row the business exponentially in low lost. Pay as you use motivates the new software launchers and thereby to contribute in IT world with their own technology. The problem of over-provisioning and under provisioning g has been solved efficiently with additional advantage of saving money and time and energy. Cloud provides location is important while choosing the cloud computing saving network usage. There are many obstacles in cloud computing, but for overcoming them it needs more explosive research in cloud field consistently. Developers should start designing their application which can be deployed on clouds. Application software in future will keep an instance of software and data over a local data-center and will use cloud in case of over-provisioning, local datacenter will be useful when disconnected from the cloud. Infrastructure software needs to be improved to work on VM s and should have capability of built in billing systems. Hardware systems should be improved to work on low power and should incorporate the flash memory s a sandwiched between DRAM and disk.