Presented By Sneha Arvind Mani

1. The Method and Tool of Cost Analysis for Cloud ComputingXinhui LiYing LiTiancheng LiuJieQiuFengchun Wang Presented By SnehaArvind Mani

2. OUTLINE • Introduction • Calculation of Cloud TCO • Calculation of Cloud Utilization Cost • Calculation Tool • Analysis on Balance Factor • Related Work • Conclusion

3. Introduction • This paper aims at formatting suits of metrics and formulas for the calculation of Cloud Total Cost of Ownership (TCO) and Utilization Cost , considering the elastic feature of Cloud Infrastructure and Virtualization in Cloud • This provides a foundation for evaluating economic efficiency of Cloud and provides indications for cost optimization of Cloud. • Attainment of economic efficiency in Cloud is a two-step process: • Resources must be applied in an area in which its benefits outweighs its cost. • For maximum efficiency to be achieved, resources must be applied in a way that costs are minimal.

4. Introduction(2) • Total Cost of Ownership (TCO) is the cost attributed to owning and managing IT infrastructure in a business. • For Cloud, it’s not only sufficient to calculate and analyze the TCO but also the cost associated with real resources committed to a particular user or application should also be analyzed, which is the Utilization Cost. • Elastic resource utilization: • Cloud adopts the architecture which can continuously adapt to the users’ changing requirements automatically. • Current cost analysis technologies do not consider the elastic characteristics of Cloud. It mainly depends on each cost item and summary of all cash outlays.

5. Wide adopted virtualization • Virtualization refers to hosting multiple applications on a common set of servers. This allows for consolidation of application workloads on smaller number of servers. • Traditional cost calculation mainly takes physical servers as the unit to calculate server cost. With virtualization, the usage of physical server is consolidated. • Even the software usage requires calculating the needed software license by VM rather than physical server.

6. Calculation of cloud TCO • The items composing the cost of cloud are: Server, Software, facilities, support and maintenance, network , power, cooling and real-estate. • With Cloud Cost Amortization model, the monthly depreciation cost is calculated for each metrics item from the initial purchase expense based on the duration over which investment is amortized. • Amortizable Rate Paramater is given by • Arp = (1+0.05) * time / (30*24*Ap) Ap – Amortization period unit ( default is one month) time – Hours consumed.

7. Calculation Model • Server Cost: In Cloud, servers are mounted in racks and constructed into a resource pool. Calculation of TCO takes all these servers into consideration. • Software Cost : According to how software is licensed, it is divided into 3 types: • Type I : Operating Systems – the license number of this software equals to the number of virtual images applied • Type II: Application Server – the license number is calculated as the number of processors where these soft wares run.

8. Calculation Model(2) • Type III: Management software – their license number is calculated by the processors they manage. • Network Cost: The cost related to networking is caused by switches, NIC and cables which are used to attach physical servers to the network.

9. Calculation Model (3) • Network cost: • Support and Maintenance Cost: This metric includes software distribution and upgrading, asset management, troubleshooting, traffic management, servers’ configuration, virus protection, disk management and performance.

10. Calculation Model(4) • Power Cost: Power consumption in Cloud is primarily caused by Computing Infrastructure including servers, network switches and Network Critical Physical Infrastructure, transformers, fans, ACs, pumps and humidifiers.

11. Calculation Model(5) • Cooling Cost: The power consumed at the datacenter is converted to heat. A Cooling Load Factor (L) is defined which represents how much power is consumed by the cooling equipment for every 1W of heat dissipation in the datacenter.

12. Calculation Model(6) • Facilities Cost: Facilities are not equipment but necessary for the normal running of equipments , like PDU, KVM cables. They are all wrapped into the rack.

13. Calculation Model(7) • Real- Estate Cost : Due to the special infrastructure required like cooling , power, the space required by a Cloud is more expensive. • Where Sspace is space taken up by all the racks under utilization. • The value of pressure confronted by unit floor cannot be beyond

14. Calculation of Cloud utilization Cost • TCO consists of all the servers in resource pool, all facilities to support these servers. However, Cloud uses part of these servers and resources to satisfy users’ workload. • Cost of the used part of resources, changing sensitively with various work loads, is important to know the usage efficiency of given resources. This is referred to as Utilization cost. • Virtualization – Cloud providers package arbitrary user applications as sets of virtual machines. The basic unit of Cloud consumed is not the physical servers but VMs the users are applied to. • A derivation-based approach is required which calculated the Utilization Cost from input VMs without dependency on monitoring or accounting.

15. Calculation Model: • The three-layer model to calculate utilization-cost. • Adding up all the eight costs gives the utilization cost.

16. Calculation Model(2) • VM Density and Number of Physical Servers: VM density is the number of VMs that can be hosted on one physical server. The number of physical servers is given by Nvm – Number of applied VMs. • Number of Racks: If we know the space a server needs and the server number Ns, we can calculate the number of racks acquired for providing services

17. Utilization Cost • Cloud providers can input derivation rules specific to their infrastructure. There are 2 ways: • Assign relationship between two different metrics item • Assign relationship between cost of two different metrics items. • All aspects of utilization cost is determined by the physical server numbers. So if server cost increase directly, other cost will increase simultaneously. • CALCULATION TOOL: • The tool presents the users with a tree of metrics parameter. Each branch represents a cost metrics item. Each child label on the branch represents a parameter under the given cost metrics.

18. Calculation Tool • Metrics Tree for TCO calculation: • Calculation of Utilization Cost

19. Analysis on Balance Factor • Balance Factor is used to describe the balance extent of all cost items’ utilization and to indicate where potential bottlenecks are being developed. • When a bottleneck is discovered, we need to consider the connected utilization changes among various items to balance the bottleneck.

20. Related Work • Currently, there are some industrial tools for TCO: • Alinean’s tool evaluates their Data Backup solution with focus on Data Transferring Cost. • PMTC focus on Network Deployment cost • Solarrcom focus on evaluation of desktop devices. • APC give series calculators for Data Center on the energy cost and carbon footprint and power related capital.

21. Conclusion • In this paper, cloud TCO is quantified with a suite of metrics and formulas considering the characters of virtualization and elasticity • Utilization Cost leverages the metrics items defined in Cloud TCO and takes workload. • These calculation models are developed into a tool which can execute cost calculation without knowing exact data and provides a flexible way to analyze the effect of different metrics on the cost.

22. THANK YOU!