Key Techniques Of Data Disaster Preparedness In Cloud Auto Scaling

With the rapid development of cloud computing technology, more and more Internet companies choose to rent computing resources provided by Cloud Service Providers’(CSP) infrastructure on demand. Enterprises expect reducing the hardware cost while improving the computational efficiency and providing better experience for users through the service provided by CSP. Therefore, how to dynamically adjust the capacity of virtual computing resources with user’s request flow as well as meet customer Service Level Agreement (SLA) are becoming hot spots of cloud computing capacity adaption research. However, as it’s not long since the cloud computing adaptive expansion had come out, the storage safety factor has not been taken into consideration yet. Actually, such loopholes in the system model will provide an opportunity for hackers to attack the user data in the cloud expansion phase, and it can seriously affect the expansion effect and cause significant losses to user.This thesis gains the following results:1. We proposed a safety auto scaling model.Through researching the current mainstream of cloud computing self adaption strategy, we extract the key components of cloud computing capacity expansion strategy. Then we designed the feedback safety automatic expansion program model combined with a mature lightweight data security authentication algorithm.2. We proposed two typical application simulation scenarios. We collect actual user traffic in real business company and made two representative simulation scenes based on Internet mail service:The receiving mail scene at working day and big strike traffic scene.3. We proposed several test cases based on LoadRunner simulation software. The model in this paper contains follow modules:load’banlancer module, data security detection module and the feedback1module. We simulated with the two mentioned scenes and analyze the performance of the proposed model.4. We proposed different traffic prediction algorithm options under different load scenarios.We combined auto scaling strategy with application scenes organically and compared the performance under different combinations. We can choose reactive and history prediction strategy for load with historical character and choose reactive strategy for load with unpredictable character. Meanwhile we should adjust simulation spacing according to the actual situation.