Transcript
Published April 2014
VMM (Fabric) Management and Resource Pooling AUTHOR: MICROSOFT MVP THOMAS MAURER
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
Management Summary In today’s rapidly changing world, service providers and enterprise IT administrators must be able to react quickly to the evolving needs of their customers. Administrators must be prepared to spin up new applications and services on demand, provide additional capacity as loads increase, and contain costs by making the most efficient use of hardware. In this white paper, Savision and Microsoft MVP Thomas Maurer outline how fabric resources like compute, storage and networking can be managed efficiently and explain how to use System Center Virtual Machine Manager to build a datacenter abstraction layer.
The Shift to Cloud Computing Cloud computing allows companies to move beyond a static infrastructure model to one where they can nimbly scale up and down new services as needed without having to make changes to the underlying physical hardware. In setting out their cloud vision, Microsoft has defined the following four core attributes: Elastic and Scalable –A cloud must deliver a flexible way to allocate new resources depending on the customer and business needs. This allows customers to quickly react to higher demand - for example peaks such as holiday sales - and it also allows customers to go back to smaller offerings when the demand decreases. Resources that need to scale include compute power, the number of virtual machines, or even the number of mailboxes. Pooled Resources – Resources for the basic infrastructure such as compute, storage and networking should be treated as homogenous collections which can be drawn upon as required. If, for example, the storage demand increases, the service provider should be able to add new storage resources to the pool which will automatically be made available to customers and running services. Self-Service Provisioning – Customers should be able to deploy new resources such as services, virtual machines, mailboxes, and user accounts without requiring IT intervention. Usage-Based Metering– Resource consumption should be recorded. Metrics such as the number of minutes a virtual machine has spent running, or how much disk space was consumed over the course of a month should be available for reporting purposes. Service providers can use this data to charge customers for consumed resources (known as chargeback), or simply generate “showback” reports. In addition to these core attributes, cloud solutions should also offer automation, management, multi-tenancy and security.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
2
The Philosophy of Datacenter Abstraction Today we live in an application or service centric world where application and services are the most important thing in the datacenter. Without services there wouldn’t even be a need to build a datacenter, so everything we do in the datacenter is to build and support services. In a modern datacenter we have to allow customers to deploy services without considering the underlying physical resources such compute, storage or networking. The problem with traditional datacenters and environments is that they were not designed to work in such ways. Pre-cloud applications and services have many dependences on the physical infrastructure and network topology such as static IP addresses and storage requirements. These dependences keep existing services from being elastic and scalable, and able to be deployed anywhere at any time. The elimination of dependencies on the underlying physical servers by server virtualization was one of the key concepts that allowed datacenters to transform from a traditional deployment model. However, server virtualization was only the beginning of the move from a static to a dynamic infrastructure. Other issues still remain such as networking and storage. Today’s networking and storage environments are designed to be very robust and reliable but this also makes them very static. Rather than changing the concepts and static nature of the underlying fabric, software-defined datacenters can put an extra layer between the fabric resources and the services or virtual machines to create an environment that fulfills the core attributes of cloud computing. Microsoft System Center Virtual Machine Manager and Windows Server can help you build that extra layer of abstraction.
Fabric Management with System Center Virtual Machine Manager Microsoft System Center Virtual Machine Manager supplies additional layers on top of your existing fabric to transform your traditional datacenter into a dynamic, flexible and cloud-oriented service deployment platform. System Center Virtual Machine Manager lets you manage fabric components and pool them together in clouds. The resources in these clouds can then be associated to different customers or tenants who can make use of them in the deployment of new virtual machines and services via self-services portals.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
3
Compute Today, most enterprise companies and service providers run multiple hypervisors in their datacenters. System Center Virtual Machine Manager is a multi-hypervisor management solution which allows you to manage not only Microsoft’s own Windows Server Hyper-V but also other hypervisors such as Citrix Xen Server and VMware ESX/ESXi from a single pane of glass. Virtual Machine Manager builds an abstraction layer on top of the different hypervisors. For example, when a customer or tenant deploys a new virtual machine, Virtual Machine Manager decides where the VM will be placed depending on the requirements of the virtual machine and the resources available across the various hypervisors. The ability to manage multiple hypervisors from a single management solution has several benefits:
Simplicity: abstracting the hypervisor layer reduces complexity and makes it easier to manage your heterogeneous environments. The common API allows you to build solutions, such as self-service and reporting portals independently from your fabric. Consistency: applications, services and virtual machines can be deployed and managed in a consistent way across all hypervisors. Choice: you can utilize a mix of different hypervisors if you have a multi-hypervisor strategy for your business critical apps. Protection: You can protect your existing investment in a hypervisor vendor without losing the flexibility to later change it.
Storage The same challenge we have with hypervisors also applies to storage. Many companies use multiple storage devices which are all managed separately. The challenge for building cloud solutions is to bring the different storage arrays under a common abstraction layer that can be made visible to services.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
4
System Center Virtual Machine Manager ‘s storage integration allows you to manage several different storage solutions such as block based SAN, file based 3rd Party NAS devices, or the new Scale-Out Fileserver (SOFS) solution built into Windows Server. Virtual Machine Manager uses the Windows Server integrated Windows Storage Management API (SMAPI). SMAPI allows the management of directly attached storage and external storage arrays. SMAPI is combined with a Storage Management Provider (SMP), or the Microsoft StandardsBased Storage Management Service and an SMI-S (Storage Management Initiative Specification) provider. To make use of this features Storage Vendors have to integrate SMI-S providers into their storage devices. Virtual Machine Manager can automatically discover local and remote storage – such as storage arrays, pools, logical units like volumes and LUNs, disks, volumes, and virtual disks - allowing fabric administrators to provision and decommission storage resources directly from the Virtual Machine Manager console. For example, you can create new logical units from available capacity on the storage array and provision them directly to Hyper-V servers or clusters. If you are using file based storage for Hyper-V over SMB, you can directly deploy new file shares on the Windows-based file servers or NAS devices and Virtual Machine Manager will automatically set the right permissions on the file shares for the Hyper-V hosts. Virtual Machine Manager also integrates with Storage Spaces and allows you to create new virtual disks on a Microsoft Storage Spaces solution. To build an abstraction layer between storage and virtual machines, Virtual Machine Manager uses the concept of classifications. This allows LUNs or file shares to be classified into groups such as gold or silver according to their performance and other characteristics.These groups or classifications can then be used by services and virtual machines without regard to the underlying storage fabric.
Networking One of the most difficult challenges in terms of abstraction is networking. Traditional datacenter networks are architected as large, robust, and static IP networks. This design was perfect when applications did not move from one place to another while they were in use. However, in today’s modern datacenter, customers need to be able to communicate with an application or service no matter where it is. One day the service may be running in location A and several days later in location B, but it’s externally visible IP address cannot change. Customers may also wish to create short-lived ad-hoc networks that are torn down only hours after being set up.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
5
To allow application and services to be deployed, moved, and scaled on demand, a dynamic overlay network is required atop the static physical network. Software-defined networking (SDN) solutions build an overlay and service abstraction on top of the underlying network and move data across the datacenter quickly and efficiently without requiring changes to the applications, servers or storage.
Microsoft’s software-defined networking solution is based on an open standard called NVGRE, and is integrated into Windows Server. This is the same technology that Microsoft uses to power Windows Azure. Virtual Machine Manager partitions the network into two abstraction layers. The underlying network is called the provider address space (PA) and is a logical network with an IP address pool. Virtual Machine Manager will automatically assign IP addresses from the provider address space to the virtual switches running in the HyperV hosts.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
6
Overlay networks known as the customer address space (CA) can then be created using the underlying provider address space. Administrators or customers can simply create new virtual machine networks on top of the logical network and Virtual Machine Manager will handle the IP address bookkeeping as virtual machines are created or moved from one host to another. To connect to the physical network from a virtual network using Hyper-V network virtualization a gateway is needed. Microsoft provides a network virtualization gateway in Windows Server 2012 R2 but there are also third party solutions and appliances. Network virtualization gateways are just one of the many network Services that can be connected to Virtual Machine Manager. There are other network services such as Windows Server IP Address Management, network load balancing and third party services which can be integrated in Virtual Machine Manager. For example, Cisco provides their own software defined network that may be leveraged.
Clouds and Tenants One of the biggest challenges in the shift to cloud computing is the move from a single tenant to a multi-tenant infrastructure solution. In this new paradigm, multiple tenants may be making use of the same underlying fabric resources at the same time. This adds new complexities to management, configuration, and security.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
7
System Center Virtual Machine Manager aggregates fabric resources - such as compute, networking and storage -into clouds or resource pools. These pooled resources or clouds may be assigned to different user roles such as tenants and application administrators. Tenant administrators and application administrators are user roles to which permissions may be assigned. Permissions include the ability to deploy or remove virtual machines, to configure virtual machines through either Virtual Machine Manager or a self-service portal such as Windows Azure Pack, and to deploy new services into the cloud.
VM and Service Templates Virtual Machine Manager provides the service designer to build and architect new services which are distributed across many virtual machines. Consider, for example, a simple online store. The store is built with a 3-tier design - a web tier, an application tier and a database tier - each of which may require multiple virtual machines for redundancy and load balancing reasons.
A Virtual Machine Manager service template includes information about the virtual machines that are deployed as part of the service, which applications and features need to be installed on the virtual machines, and the configuration of the network including load balancers. By making use of the versioning features one may have multiple versions of the same service and may easily switch existing services to a newer version or back to an older version if needed. Virtual Machine Manager will take care of all maintenance and scaling required by the service including updating applications and deploying new virtual machines.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
8
Self-Service and Reporting With System Center App Controller and Windows Azure Pack, Microsoft offers two web-based self-service solutions for managing services and virtual machines.
With the 2012 release of System Center, System Center App Controller replaced the Virtual Machine Manager Self-Service Portal and allows users to connect to multiple Virtual Machine Manager instances, Microsoft Azure subscriptions, and other hosting providers. As such, users can manage all their virtual machines and services from a single console, whether they are running in a private or public cloud.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
9
Windows Azure Pack was released with System Center 2012 R2 and is not only a self-service portal but a collection of Microsoft Azure technologies, available to Microsoft customers at no additional cost for installation into their own data center. Windows Azure Pack runs on top of Windows Server 2012 R2 and System Center 2012 R2 and enables providers to offer a rich, self-service, multi-tenant cloud, consistent with the public Microsoft Azure experience.
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
10
To ensure that the underlying fabric is properly configured and tuned, Savision has built a Virtual Machine Manager Add-In that compares your cloud configuration against best practices recommended by Microsoft and Microsoft MVPs. Savision’s Cloud Advisor also includes capacity predictions based on Virtual Machine Manager data collection. Savision’s Cloud Advisor looks for problems and makes recommendations like: “Virtual Machine Appears to be Unused” “Prediction: All Available Memory Will Be Consumed By…” “Virtual Guest Services Are Not Installed” “Starting Memory Is Too High” “Low Disk Space On Cluster Shared Volume” “Dynamic Memory is not enabled” Cloud Advisor is a great, free tool for fabric administrators to tune their environment in an easy and simple way. For administrators requiring deeper knowledge of their fabric and the ability to right-size resource allocations, Savision also offers their Cloud Reporter solution based on Virtual Machine Manager and System Center Operations Manager.
Summary Microsoft System Center Virtual Machine Manager provides an abstraction layer over existing fabric resources to transform a traditional datacenter into a dynamic and flexible cloud services infrastructure. With the extensibility provided by add-ins, it is possible for vendors like Savision to deliver high-class extensions that can
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
11
help fabric administrators tune and configure their cloud infrastructure and focus on delivering a platform for the deployment of elastic and scalable services.
About Thomas Maurer Thomas Maurer works as a Cloud Architect at itnetx gmbh, a consulting and engineering company located in Bern/Switzerland. Thomas is focused on Microsoft technologies, especially Microsoft Cloud Solutions based on Microsoft System Center, Microsoft Virtualization and Microsoft Azure. This includes Microsoft Hyper-V, Windows Server, Storage, Networking and Windows Azure Pack as well as Service Management Automation. Before Thomas joined itnetx he worked as Head of Microsoft Engineering for a service provider where he was responsible for planning and implementing public cloud solutions based on Microsoft technology. In 2012 Thomas Maurer was awarded the first time by Microsoft with the Microsoft MVP (Most Valuable Professional) Award for his expertise in Hyper-V. Thomas works closely with Microsoft and partners to promote Microsoft Technology as speaker for Microsoft and other Technical events such as Microsoft TechDays or System Center Universe. More on www.thomasmaurer.ch
About itnetx gmbh itnetx is located in Switzerland/Europe and is focused on IT infrastructure solutions. As one of the premier Cloud & Datacenter partners in the country we help customers to plan, implement and operate solutions based on Windows operating systems, Microsoft System Center, Microsoft Hyper-V, Windows Azure Pack and Windows Azure. We are a Microsoft Gold Partner and three-times in-a-row Microsoft Switzerland partner of the year for the "Datacenter" area. Details can be found on www.itnetx.ch.
About Savision Savision is the market leader in business service and cloud management solutions for Microsoft System Center. The company’s monitoring and visualizing capabilities bridge the gap between IT and business, by transforming IT data into predictive, actionable and relevant information about the entire cloud and datacenter infrastructure. Savision's intuitive and customizable dashboards provide context for each business service, increasing organizational efficiency, reducing IT operational costs up to 20%, and preventing IT-related problems and business downtime. Savision’s solutions scale from small and medium businesses and government bodies to Fortune 500 companies operating in different fields and have been adopted by over 650 organizations worldwide. Savision is headquartered in Amsterdam, The Netherlands, and has offices in Dallas and Ottawa. For more information, visit www.savision.com
Amsterdam - New York – Ottawa - Dallas www.savision.com
[email protected]
12
