Wednesday, September 24, 2014

Providing Disaster Recovery for Windows Azure Pack tenants

Providing Disaster Recovery for Windows Azure Pack tenants

In today’s datacenters where service providers are using Hyper-V, System Center and Windows Azure Pack, we will find a wide diversity of technologies and many moving parts that together will deliver one or more advanced sophisticated solutions.

To deliver cloud computing at scale, the most important thing is an understanding of the solutions as well as the preferred design to meet these goals.

In a nutshell, the holy grail of disaster recovery for tenants is something like this:

“A failover should occur with a minimum or none interaction from the tenants”

This one is hard to achieve to be honest, and the underlying design of the management stamp and the rest of the topology need to be designed in order for this to work.

Let us just look at some of the dependencies when a service provider is offering a complete IaaS solution to their tenants, leveraging the VM Cloud Resource Provider in Azure Pack.

·         VMM Management Stamp
o   The management stamp is the backend, the fabric controller of your VM Clouds in WAP
o   VMM manages the scale-units as well as the clouds abstracted in to WAP.
o   Networking is also a critical part of VMM once NVGRE is implemented. Since VMM act as a network controller in this context, we have in essence a boundary for the managed virtual machines
o   Virtualization Gateways is also managed by the stamp, within the single boundary
* Hyper-V Replica, managed through ASR and VMM
·         SPF
o   Service Provider Foundation exposes the IaaS capabilities in VMM to Azure Pack, and is the endpoint used by the VM Cloud Resource Provider in WAP
·         Active Directory
o   A service provider datacenter may contain several domain controllers and forests to manage their solutions, and hence a critical part of the offerings
o   ADFS for federation using multiple services can be considered as a key component
·         SQL
o   Almost everything in the Cloud OS stack on-premises has a dependency and requirement of SQL servers and databases
·         Windows Azure Pack
o   The API’s, the portals and the extensions are all there to deliver the solutions to the tenants, providing Azure similar services based on the service providers own fabric resources, such as IaaS and PaaS.
§  VM Cloud Resource Provider
·         Remote Console
·         VMM library with VM templates and VM Roles
·         Tenants
·         VMs
·         VM Roles
·         Virtual Networks
·         Usage
o   Other resource providers, such as Automation with SMA, Web Site Clouds, SQL Clouds and more, are all part of the solution too
·         Scale-units
o   Compute
o   Storage
o   Networking

In order to solve this, and address many of the challenges that you will be facing, it is really important to be aware of the structure of these components, how they scale, how they can survive and what it takes to bring them back online again.

We have been working on some designs that may fit the majority of the organizations out there, each and one of them with pros and cons.

Within the next weeks, I will publish our findings and this will give you a better understanding of what is possible and what may be the best solution to ensure business continuity for both the service provider and the tenants.

No comments: