Understanding Windows Azure Pack and your service
offerings
From time to time, I meet with customers (and also other
system integrators) that is not fully aware of the definition of cloud
computing.
I never expect people to know this to the very nasty
details, but have an overview of the following:
·
Deployment models
·
Service models
·
Essential characteristics
What’s particular interesting when discussing Windows
Azure Pack, is that the deployment model that’s relevant, is the private cloud. Yes, we are touching your
own datacenter with these bits – the one you are in charge of.
For the service models, we are embracing Infrastructure
as a Service (IaaS – using the VM Cloud Resource Provider), and Platform as a
Service (PaaS – Using the Web Site Cloud Resource Provider).
The essential characteristics are also very important, as
we’ll find elasticity, billing/chargeback, self-service, resource pooling and
broad network access.
If you combine just self-service and IaaS, this tells us
that we empower our users to deploy virtual machines on their own. Right?
So having the flexibility to provide such service, we
also rely on the underlying architecture to support this. Due to scalability
(elasticity), we need to ensure that these users constantly have access to the
solution – no matter what device they are using (broad network access), we need
to find out who is consuming what (billing/chargeback), and last but
not least – be able to produce these services in an efficient way that makes it
cost effective and profitable (resource pooling).
So, it starting to make sense.
There is a reason for what we are seeing and we are
providing these services by abstracting the underlying resources into clouds,
plans and subscriptions with the Cloud OS.
Implementing a complete IaaS solutions may bring some
obstacles to the table.
Organizations tends to think that IaaS is something they
have provided for years. Perhaps they have provided virtual machines, but not a
complete IaaS solution.
The reason for that is that IaaS is relying on abstraction at every layer. This is not
only about virtual compute (memory, CPU), but also about virtual storage and
virtual networking.
This is when it gets interesting, using network
virtualization.
Remember that self-service is an essential characteristic
of the cloud, right?
So delivering IaaS would also mean that the user is able
to do stuff with the networking aspect as well, with no interaction from the
service provider/cloud administrator.
This is why Software-Defined Networking (NVGRE) is so
essential to this service model, and hence we run into the following obstacles.
·
The customer (most often service provider) wants
to continue to provide managed services, such as:
o Backup
(both crash consistent and app consistent)
o Monitoring
(above the operating system level, covering the application stack)
This is what they are doing today, with their
infrastructure. But this also has a high cost to operate due to all the manual
operations needed and involved to get the wheels moving.
Luckily, Windows Azure Pack is able to cover both scenarios, providing a consistent
experience to users/tenants no matter if they are running resources in a “legacy”
infrastructure, or a new modern IaaS infrastructure.
The following architecture shows that we are using two
Virtual Machine Management Stamps.
Both of these are located behind the SPF endpoint – which
present the capabilities, capacity and much more to the service management API
in Azure Pack.
A cloud administrator then creates a Hosting Plan in the
Admin Portal of Azure Pack, which is associated with the legacy cloud in the
legacy VMM server. This plan is available for the users/tenants who are
subscribing to managed services.
A new plan is created, associated with the IaaS cloud and
the IaaS VMM server, available for the users/tenants that need IaaS, without
the requirement of managed services. They are dealing with these themselves.
Hopefully this blog post gave you an overview of what’s
possible to achieve using Azure Pack and combine both kind of services using a
single solution.
(Want more info? – please join my TechEd session in
Barcelona next week).
