Cloud fabric infrastructure planning

Planning a cloud fabric infrastructure involves practice and involvement of business and technical decision makers. Cloud fabric has to be planned in such a manner that ensures that services are available for the business in a reliable and efficient manner. Current infrastructure assessment, forecasting and future resource planning, cloud service catalogues, along with strategic vision and financials are the key facts to consider while planning for cloud infrastructure. Let's have a look at the foundation and basic principles for planning a WAP-based cloud fabric infrastructure for delivering IaaS services.

Choosing the hardware

While choosing hardware for your cloud infrastructure, there is a long list of precautions to be taken care of. The following are some of these precautions:

• Follow the Microsoft certified hardware and software supportability and compatibility list. It is not a mandatory requirement as a WAP-based solution can be installed on almost any x86 based virtualization supported hardware, however, it is recommended to go with Microsoft certified hardware to avoid any glitches at a later stage in production. Getting support could be challenging sometimes while using non-supported hardware.
• While going for commodity hardware, check for compatibility and features of individual pieces of server such as NIC cards and so on.
• Use latest hardware supporting new features such as RDMA and other hardware offloading capabilities to take advantage of all capabilities of your hypervisor and cloud layer.
• Check for the scalability flexibilities, sufficient number of processors, memory, PCI slots can be helpful in increasing resources as and when required.
• Hardware failure happens, it's a good idea to check for hardware vendor support and SLAs in case of hardware failure part replacement and more.
• While comparing multiple hardware options for your cloud, include data center physical requirements (space, power, cooling, and so on) as one of the parameters, physical requirements usually become constraints in the real world.

While all preceding points are valid, today's data center technologies trends include software defined strategy. Using commodity hardware with best of breed software runs the data centers. It is fairly possible and good to use commodity hardware in your WAP cloud solution, though building the commodity hardware has to be planned very carefully.

For example, using non-supported disk drives leads to IOPS related performance issues, and the same goes for NIC speed and features. Follow the industry best practice guideline along with Microsoft Hyper-V compatibility and features support while designing commodity hardware.

Choosing a combination of both proprietary and commodity hardware is also not uncommon; organizations can choose to run critical workload on proprietary hardware to ensure performance and support, whereas use commodity hardware for less critical or test and development workloads.

Enabling commodity hardware options is helping small and midsized businesses (SMBs) to utilize enterprise class virtualization and cloud technologies for their IT needs in a limited budget. The same goes for the cloud service providers, small vendors with budget constraints can also get into the cloud services market using commodity hardware.

Tip

The Microsoft supportability and compatibility list can be accessed at https://www.windowsservercatalog.com/.

Sizing the hardware

After choosing the hardware, the next thing which comes is sizing. Sizing for a cloud fabric for a private cloud and service provider cloud is an altogether different science with a common foundation of data center infrastructures sizing practice. Projecting future needs is the most important aspect while sizing any cloud hardware that is cloud should be able to handle hardware needs for the next X years, depending upon businesses strategy. Microsoft provides MAP toolkit (Microsoft Assessment and Planning) which can be utilized by organizations and service providers for accessing current infra and planning for target infrastructure.

Let's have a look at few major sizing best practices and procedures for both private clouds and service providers' cloud.

Sizing the cloud hardware for private cloud:

• Assess current workload requirements and plan hardware as per Hyper-V Server Virtualization sizing guidelines pertaining to virtual CPUs, memory, and more.
• Utilize MAP toolkit to assess current running workload in traditional computing.
• Perform server consolidation planning if P2V or hardware refresh is also planned.
• List down the expected new workload requirements (like new application planned in the near future and so on.)
• List down the expected resources increment in the coming X years in terms of compute, network, and storage.
• Analyze isolated infrastructure requirements if any (a few apps may need to be isolated from each other in terms of network/server due to compliance and more.)
• Factor for Business Continuity and Disaster Recovery (BCDR) planning.
• Bring all collected details together and size the hardware as per Hyper-V virtualization and IT infrastructure sizing guidelines.
• Add hardware requirements for cloud management infrastructure depending upon workload to be managed.

Sizing the cloud hardware for service provider cloud:

• Access customer needs and technological requirements in service providers' technology space
• Define technology offering and flexibilities in terms of infra and apps with their fabric requirements
• Set business goals like having X number of resources/customers in the coming X years as per service provider's strategy
• Bring all collected details and strategy decisions together and size the hardware as per Hyper-V virtualization and IT infrastructure sizing guidelines
• Add hardware requirements for cloud management infrastructure depending upon the workload to be managed as planned here.

Note

Sizing the hardware for cloud or any traditional data center is an altogether different detailed study and practice. The above mentioned guidelines are to help readers in understanding the base foundation for cloud fabric hardware planning. Refer to product and practice specific technical and economical guidelines for a production sizing.

Choosing hypervisor's deployment options and editions

Microsoft added a variety of all enterprise class virtualization features to Windows Server 2012. With Windows Server 2012 / R2 Hyper-V becomes the enterprise class proven virtualization platform supporting greater flexibility and scalability.

Microsoft Windows Server 2012 R2 Hyper-V provides an enterprise class virtualization platform for WAP-based cloud solution. Hyper-V can be deployed in three ways:

• Windows Server 2012 R2 Server with GUI with Hyper-V role
• Windows Server 2012 R2 Server Core with Hyper-V role
• Hyper-V Server 2012 R2 (free)

The Server with GUI option installs and configures all user interface options, such as server manager, desktop experience, and so on. It installs Windows Server also in the same traditional manner with all features and functionality available.

The Server Core version is a stripped down version of full installation, in this no GUI components gets installed. It has to be managed via a command line, PowerShell or remote server management options. Benefits of deploying the Server Core version includes but is not limited to less resource consumption, less disk footprint, better security by reducing attack surface of eliminated components in comparison to the Server with GUI option, and so on.

It is recommended to deploy Server Core in a cloud infrastructure for hypervisor's as these servers shall be used for running VMs only. This can take advantage of server core deployment benefits in large infrastructures.

With Windows Server 2012 and R2, it is possible to switch between full editions and core edition after deployment. Administrators can choose to install the Server with GUI initially and switch to the core version post configuration for easy deployment.

Windows Server editions

There are multiple editions of Windows Server such as Essentials, Foundation, and so on, but majorly for any virtualization needs below two editions are used:

• Windows Server Standard edition
• Windows Server Data center edition

Both Standard and Data center editions of Windows Server is licensed on processor basis, one license per two processors. Both Standard and Data center editions provide the same set of features, the difference lies in the virtualization rights. Standard edition allows you to run two virtual machines with Windows Server OS, in the case of Data center edition the number goes unlimited. Taking a scenario Windows Server 2012 R2 Standard edition licensed hypervisor hosts allows you to create two virtual machines with Windows Server 2012/R2 Standard Guest OS without purchasing any additional licensing. Licensing downgrade rights apply for running VM with Windows Server 2008 R2 OS.

Choosing the right edition for your cloud

Choosing the right edition depends upon the type and number of operating systems in the cloud infrastructure. It's recommended to go with the Data center edition if virtualized workload includes Windows as Guest OS majorly. Standard edition can be helpful and budgetary while the number of virtual machines with Windows OS isn't high.

As a best possible solution, in a large infrastructure combination of both Standard and Data center editions can be utilized depending on the OS of virtualized workload.

While for organizations it will depend on workload nature if it is Windows or non-Windows to choose the right edition, server providers can use a combination of both Data center/Standard edition to service customers' requests for both Windows and non-Windows infra requirements.

Tip

Refer to the Windows Server 2012 product and edition comparison guide on http://www.microsoft.com/ for detailed comparison between various Windows editions.

Free Microsoft Hyper-V Server 2012 R2

Microsoft Hyper-V Server 2012 R2 (also known as Standalone Hyper-V) is a free (yes, completely free) edition of Microsoft's proprietary hypervisor. It is a stripped down version of Windows Server operating system comprising only hypervisor and virtualization capabilities, Windows Server drivers, and supporting components such as cluster capabilities. Free Hyper-V version doesn't include any additional Windows feature apart from hypervisor, making it available with less footprint, management burden, and attack surface.

Free Hyper-V Server 2012 R2 provides free access to Hyper-V virtualization capabilities only, virtual machine guest OS licensing has to be separately considered. If you are running Windows Server OS inside a virtual machine, it's a good idea to use Standard or Data center edition depending upon the count of Windows VMs. This edition can be helpful in scenarios where Guest OS inside VMs wouldn't be Windows.

A major use case of free Hyper-V Server is to be utilized by organizations or service providers for hosting non-Windows workloads.

Note

Ensure to check for the supportability of the free Hyper-V version while extending WAP cloud solution with custom offerings or partner provided solutions.

Microsoft Hyper-V is the best choice for a WAP-based cloud solution, though there are integration possibilities with other hypervisor's vendors such as VMware and Citrix. See product specific guidelines to choose the edition and deployment model along with the supportability matrix of features while choosing VMware or Citrix.

Choosing the system center deployment options and edition

Microsoft System Center 2012 R2 is the core management component for delivering cloud services in a WAP-based cloud solution. System Center products can be deployed in multiple ways such as:

• Physical or virtual machines
• Standalone or high availability mode (product dependent)

System Center components requires Windows Server as an operating system to run. It can be in a virtual machine or a physical server. Since we virtualize every piece of our cloud management and tenant workload, it doesn't make sense to keep system center components on physical servers unless there is some specific requirement or exception.

For both organizations and service providers, it is recommended to deploy system center components inside virtual machines to take the advantage of hypervisor level HA and other capabilities.

Along with hypervisor HA, deploying servers in Guest OS or app level HA adds additional benefits such as application level failure detection and load balancing.

System Center products can be deployed in OS/app level cluster or redundant mode.

Organizations or service providers can choose a combination of standalone or HA mode products depending upon their availability business and technical requirements.

System Center editions

System Center is available as a suite in two editions. When you buy a System Center suite you are entitled to use all components (except SCCM) depending upon licensing.

The two editions of the System Center suites are as follows:

• System Center Standard edition
• System Center Data center edition

Licensing is required only for a number of endpoints being managed by system center components. There is no licensing required for running the system center components. Licensing is available in a similar fashion with Windows Server that is a processor-based licensing (two products per license). There is no difference in terms of features with Standard and Data center editions; the difference lies in the number of virtual instances you can manage with Standard and Data center editions. With Standard edition, two OSEs (Operating System Environments) can be managed by the System Center and is unlimited in the case of Data center edition. Microsoft licensing policy may change from time to time and depending upon agreements such as EA/SA and so on. It is recommended to touch base with Microsoft for planning licenses and editions.

Tip

Refer the How to buy System Center 2012 R2 page on Microsoft's official website to know more on licensing and comparisons.

Choosing the right edition for System Center for organizations and service providers directly corresponds to the edition chosen for Windows Server. It's good to go with System Center Data center edition to manage Windows Server Data center edition and vice versa to enable a consistent feature set and flexibilities.

Note

The licensing details given are up to the date of writing this chapter; licensing policies may change in future, and it is advised to consult while planning for licensing and editions.

Cloud management infra and tenant workload infrastructure

A cloud infrastructure is usually segregated in two categories, cloud management and tenant workload. Let's have a look at these two:

• Cloud Management Infrastructure: Management infrastructure or management cluster is used to host virtual machines for cloud management components such as Windows Azure Pack server and system center components. It is recommended to have a separate hypervisor cluster dedicatedly for hosting cloud management workload. The sizing of the management cluster would depend upon the number of services offered and the size of tenant workload to be managed.
• Tenant Workload Infrastructure: Tenant workload infrastructure is used for hosting customers' cloud resources such as virtual machines and databases. Tenant workload infrastructure has to be designed keeping available cloud services, QoS and SLAs in mind.

Core MS infrastructure services planning

Microsoft's WAP-based cloud solution uses core Microsoft core infrastructure services such as Active Directory, DNS, DHCP, and so on for foundation operations such as authentication and more. These core infrastructure services have to be deployed with product specific guidelines and industry best practices to ensure availability and efficiency . Let's have a look at the core services and the deployment best practices:

• Active Directory: Microsoft AD is the most critical component of most of MS infrastructure solutions including WAP cloud solution. AD is used for authentication, authorization along with other capabilities in WAP cloud. AD deployment can be planned with redundancy (deploying additional domain controllers) and security best practices. Regular backup and disaster recovery should be planned for Active Directory infrastructure.
• DNS: Usually co-located with AD, DNS provides name resolution for the entire cloud. DNS failure may cause entire cloud failure, hence it needs to be planned to ensure every time availability and security from attacks. Regular backup and DR should be planned.
• NTP: Another crucial component of any cloud or any IT solution, NTP ensures time synchronization among all cloud resources. Time lag may cause the entire cloud solution to fail as well as incorrect billing. The NTP device must be highly available as well as accurate. All components should be configured to use the same NTP device. Domain joined Windows operating systems utilize the domain controller (PDC role) for time synchronization.

  Tip

  Follow Microsoft's guidelines while virtualizing domain controllers holding NTP role. See https://technet.microsoft.com/en-us/library/virtual_active_directory_domain_controller_virtualization_hyperv%28v=ws.10%29.aspx to know more.

• DHCP: Dynamic Host Configuration Protocol can be utilized for Dynamic IP address allocation and management. The deployment of DHCP should be highly available and efficient for serving cloud IP address requirements.
• Other infrastructure supporting services (PXE Boot server, IPAM and so on): Depending upon deployment and use cases other MS services such as PXE boot and deployment server for bare metal provisioning or network boot, WSUS or SCCM for patch management, IPAM for IP address management, jump servers for centralized access, security products should be deployed in an efficient and highly available manner as applicable.

Tip

Planning and deployment options for cloud network and storage shall be discussed in the coming topics in this chapter.