IT Transformation and Cloud Migration
From the perspective of IT, migration to the cloud allows the transformation of IT as a cost center into IT as a service. IT as a service is generally understood to mean the delivery of a standardized catalogue of applications and platforms to end users within one organization. The goal is to create a standardized service catalogue, rather than maintain an infrastructure customized for each application.
The IT organization has to put the data center onto a platform suitable for automation, tools that automate and monitor performance. The main characteristics that are distinct form the traditional data center are:
• Dynamic scalability: The amount of available resources should be able to dynamically increase or decrease based on demand.
• High availability: The cloud infrastructure should have as much uptime as possible with minimal unplanned outages.
• Chargeback model: Cloud resource consumption should be monitored and measured to calculate user chargeback.
• Self-service: Users themselves should be able to request and provision cloud resources as needed.
• Automation: Processes within the cloud infrastructure should be as automated as possible to ensure timely deliverables to users. Private cloud is well suited for applications with workload that are input output intensive, latency sensitive and require special treatment for security compliance.
Problem & Approach
For approaching a migration to a cloud environment, the following few fundamental questions need to be addressed:
• What kind of cloud deployment model is best suited for each of my applications?
• Which applications can and should I move to the cloud?
• How do I transition my legacy applications to the cloud?
• How do I maintain security and policy compliance in the cloud?
• How do I transition my organization to best take advantage of cloud?
• Commercial cloud or Open source cloud architecture?
The cloud migration should be considered as much a methodology as it is a technology. It goes beyond virtualization and includes automated workload management, self-service interfaces and some form of usage metering or chargeback. It also requires the ability to share resources to maximize utilization – among functional groups that may not share resources today.
Enterprise Cloud enablement is an evolutionary journey involving technology, process and people. Each step towards Cloud enablement results in reduced cost and increase in agility. Support and participation of key stakeholders including Business, Application teams and Infrastructure operations tea, is key to the success of cloud enablement.
By introducing virtual machines, attention shall be paid to provisioning and administration, as the processes the entire life cycle of virtual machines and their supporting resources are defined and managed. A cloud management platform shall also meet future requirements. It must address a broad range of capabilities: automation and orchestration, self-service provisioning, chargebacks, security and compliance audits, and governance, ability to support a multi-cloud environment (multiple private’s cloud, hybrid clouds).
The other important element is orchestration. Automation is associated with a single task while orchestration is linked with a process that involves workflow around multiple automated tasks.Orchestration deals with all the infrastructure elements. It is the automated coordination and management of servers, storage, security and networks to deliver services to users and is based on policies that define the relationships between the users, computing servers, data storage element, security and network devices.
Virtualization has added a layer of abstraction to traditional monitoring; which is an abstraction of the various physical devices. New monitoring tools are required to have insight into both physical and virtual elements. Monitoring ensures that applications meet specific performance targets and also help to have insights into the performance of different elements such as response time of storage devices, computation and storage utilization of the application. This information is needed for efficient tuning of the cloud elements.
The answer to the above questions will lead to the best choice of cloud layer service type (IaaS, PaaS, SaaS), cloud model type (private, public, hybrid), also will help to define the timeline for moving applications in the cloud infrastructure and the type of changes that might be implemented in the organization to get the maximum benefits of the cloud computing.
The choice of which cloud model (public, private – hosted internally or externally, hybrid) to be used is based on the assessment of the following aspects:
• Application portability
• Application interfaces (within application and to other external applications)
• Data architecture (access, security)
• Licensing agreement
First is the server virtualization. This will allow the IT organizationto avoid maintaining multiple vendors’ server virtualization software, since that typically leads to an increased complexity of managing and maintaining the server virtualization infrastructure.
Should they decide to introduce in the future a new hypervisor (for reasons such as cost, merger & acquisition, or turnkey application vendor requirement) besides what is being used today (and specifically moving towards Open source virtualization, for cost reductions), attention shall be given to the automation and orchestration tool that can handle multiple hypervisors. This is not the case for some of today’s management and performance-monitoring tools.
This is followed by the Storage virtualization. Main centralized data storage is in existence today, based on system provided by existing storage and NAS vendors. The array based virtualization route is the best route for storage virtualization. To achieve the benefits of storage virtualization, evolution is towards dynamic tiering and thin provisioning with the virtualization of storage systems.
And finally comes network virtualization, where the network (routers, switches, firewalls, etc.) is seen as a continuation of the virtualized computing and storage infrastructures. The network controller sitting on top would understand the semantics of the virtualized entities and automate network provisioning accordingly.
The approach put in place in our deployment case studies resulted in the following:
• Setting up a Cloud project with cross-departmental members plus a few dedicated resources.
• Defining measurable objective of servers virtualization such as % of resources to be achieved within the next 3-5 years
• Proof of concept and/or implementation of critical mission application virtualization to understand the impact of virtualization in Service Level
• Agreements (availability, scalability, security, QoS, etc.) and in organizational aspects.
• Proof of concept for automation/orchestration and monitoring tools by using trials version offered by vendors.
• Defining revising quarterly or at minimum half year intervals, sets of current, intermediate and longer term requirements for virtualization and cloud computing, based on the following lessons learned during the past sever virtualization, coming mission critical application virtualization, on best practices and trends, as well as, adherence to standards.
• This included defining and measuring KPI along the lines of:
(•) Performance measurement
(•) Configuration management including SW patch, load management
(•) Servers, Data storage, network elements, security
(•) New trends such as virtual desktop, network virtualization, application virtualization
• Identifying and implementing the adequate tools to cater for the established requirements.