The implementation cost can also vary from less than $100,000 to several million dollars.
Initial migration costs are just one component , and typically not the most expensive component, of the overall cost to migrate to the cloud. A rough estimate is often the first step in a cloud migration project. The initial costs in terms of money and time are largely consumed by unit, system, and performance testing.
The cost of VM migration is higher than the memory state transfer. The migration cost is reduced by replicating the VM images in the background before live migration, and then synchronizing the disk states later [ 71]. However, comparison of current and previously transferred pages involving big data will take more time due to the increased number of comparisons. The incremental effect in comparisons will also bring a heavy load on CPU effecting the performance and making the proposed approach unsuitable for big data applications.
There are typically two types of workloads: delay-sensitive nonflexible interactive jobs and delay-tolerant flexible batch jobs. As in Ref. [ 70 ], each DC processes its batch jobs locally. The total arrival rate of interactive jobs to a DSP at time t is denoted by Λ ( t ). The DCs’ front-end server splits the total incoming workload Λ ( t) into separate workloads of geo-dispersed DCs, denoted by { λ i ( t) } i ∈ I.
Application migration diagrams identify application migrations, from the current situation right through the application components of the target application. This migration is broken down into steps in order to describe the migrations between each intermediate situation. This type of diagram enables a more precise estimation of migration costs by presenting exactly which applications and interfaces must be mapped between each migration step.
This correlation could arise because highly educated workers may be more efficient at learning about employment opportunities in alternative labor markets, thus reducing migration costs. It is also possible that the geographic region that makes up the relevant labor market is larger for more educated workers. Consider, for example, the labor market faced by college professors. There are few firms in any given city and professors' skills are highly portable across colleges and universities. In effect, college professors sell their skills in a national—and perhaps even an international—labor market.
Migrating the workload from front-end server to geo-distributed DCs can be very costly . Therefore, the migration cost to DC i can be modeled as [ 22]