1. Research Purpose
– In the environment that heterogeneous Managers providing cloud services exist, the purpose of this research is to develop the technique and the cloud collaboration computing architecture which provides the QoS-recognitive cloud services and cloud resources to the customers with sharing the contents to process.
2. Cloud Collaboration Content
– In Cloud computing, multiple CSPs provide their own services with various performance and cost. Cloud service user should integrate a number of services from different CSPs and utilize proper services considering cost, performance, and target application’s characteristics. Also, active cloud collaboration is needed. Therefore, integration technique for integrating cloud-based resources and cloud collaboration technique is essential.
Figure 1. Cloud Collaboration Environment
– We consider cloud broker platform prior to development such techniques. Cloud broker plays a role as a middleware between multiple heterogeneous cloud environment and cloud service user. It determines and provides proper cloud services which ensures user SLA and minimize the cost instead of cloud user.
Howerver, cloud broker for federated cloud which have heterogeneous OS, heterogeneous cloud platform, and various cloud service resources only performs integration of multiple CSPs and requests VM instance. Recent cloud broker also provides only the function of VM instance specification and resource find operation.
Figure 2. Traditional Cloud Broker Model
Figure 3. Traditional Cloud Broker Model
– In cloud collaboration environment, VMs with the same flavor type may have different performance because of heterogeneity of physical nodes and VM interference. However, traditional cloud broker cannot cope with that situation so that SLA cannot be ensured.
– Also, in the case of guaranteeing network performance between VM instances are important, traditional cloud broker do not consider this situation so that they can’t assure SLA. Also, traditional cloud broker only considers to provide on-demand VM instances so that they do not consider cost reduction using reserved VM instance.
– We propose a cloud broker platform named Cloud Collaboration Platform(CCP). CCP considers computing, network performance using cloud resource profiling function. Also, CCP reduces VM management cost by adapting reserved VM instances and VM provisioning operation so that it dynamically provides optimized service while meeting user SLA and QoS with minimized cost. Moreover, in cloud environment, many requests are in a form of running scientific applications or workflows. Therefore, CCP provides workflow scheduling function to optimized use of resources. For contents management, CCP manages multi-teanant and provides interface which is suitable for active collaboration.
– The entire framework for CCP is divided into Active Contents Collaboration Platform (ACCP) and Adaptive Resource Collaboration Framework (ARCF).
Figure 4. Traditional Cloud Broker Model
– ARCP manages heterogeneous multi-clouds in a integrated form and provides proper cloud resource in cost adaptive way while meeting user QoS. ARCF provides policy based resource management mechanism considering multiple cloud services’ performance and cost. Moreover, ARCP provides application service for various scientific applications such as genetic analytics, scientific experiments.
– ACCP provides collaboration environment for sharing services and contents in a virtual workspace based on cloud. Also, it automatically performs active work according to the event occurred by contents. ACCP provides scalable service through interconnection to cloud infrastructure, and development environment for collaborative works. Also, ACCP provides virtual workspace named content space so that users can rapidly and easily collaborate with co-workers with it.
– It provides Content Space, which operates as the interface for task flow processing service to offer a convenient environment for various collaborative works.
– Cloud collaboration technique provides contents collaboration environments required for next-generation mobile services such as business and social media. Also, it integrates different services from diverse cloud vendors then provides a unified interface so that users can easily develop and quickly utilize an application. In this document, we concentrate upon taking care of Adaptive Resource Collaboration Framework.
First functionality is the heterogeneous resource monitoring. The ARCF maintains the resource information (CPU, Memory, Storage and so on) of VM instances and the required time to complete workload. This information are collected in real time and processed to stay in specific form the heterogeneous which can be utilized by resource profiler. Moreover, the scalability should be guaranteed not to exceed the threshold value related with performance degradation even if the number of VM instances increase.
Second essential functionality of ARCF is the Heterogeneous Resource Profiling taking care of VM placement problem. It equips with VM placement technique with two QoS constraints that client cannot handle. One of them is guaranteeing network performance among multiple VM instance. We provide network-aware VM placement scheme to overcome data transmission delay. The other is computing aware VM placement scheme. It enables for a client to use VMs of same computing power with cheaper price by performing resource profiling.
Third functionality is resource provisioning. We considered reserved VM to reduce resource leasing cost. We achieve this goal by employing two techniques; one of them is Adaptive Resource Reservation Scheme (ARRS). It decides optimized number of newly leasing reserved VM by leveraging the concept of marginal cost. The other is Adaptive Resource Allocation scheme. It consists of recycling reusable on-demand VM (OVM), replacing targeted VM type, and repositioning the task in execution from OVM to RVM.
Fourth functionality is workflow scheduling. It is about processing active works while satisfying user SLA (e.g. deadline). When multiple workflow processing request is given, we partition each workflow request into fragment based on its unique critical path and schedule each fragment. Also, as cloud resource performance is not static, we suggest a scheduling scheme which divides a task into multiple subtasks then process them in parallel manner to overcome the unforeseen performance degradation.
– Integrating Eucalyptus and Openstack with resource management API library development and Linking ACCP
– Development of Collect-d and SIGAR API-based cloud node VM instance real-time monitoring module
– Resource profiling using the prediction method by comparing the application similarity, guarantee the task deadline, and achievement the 17 to 20% error rate with real-time application processing time prediction
– performance improvement by managing and allocating the RVM-based instance s and OVM-based instances. cost reduction by 70% and resource utilization 20% advancement