This week I was reading some blog posts in InfoWorld, Cloud Computing Blog by David Linthicum. David is a frequent blogger offering advice on both the private and public cloud.
One recent blog post is of particular interest – ‘Shadow IT’ can be the cloud’s best friend – the blog focuses on how employees within an organisation, responsible for profit centres will acquire and manage IT resources outside the control of corporate IT. David comments:
… IT needs to face reality: For the past three decades or so, corporate IT has been slow on the uptake around the use of productive new technologies.
GigaOM released this month a 7 step guide to cloud success. The research by Jo Maitland starts by contrasting the intention of cloud versus what’s actually happening. Typically many organisations start with a few servers in a room, but this can quickly turn into “multiple clouds on different platforms that all use different providers and management tools. It’s this kind of stovepipe approach to IT that creates inefficiencies, high costs, and labour-intensive management.”
The introduction highlights the need for IT departments to manage as much of their infrastructure as possible with a single solution. This echoes sentiments by Tony Lucas in his blog series on Managing Virtual Machine Sprawl where he highlights a number of problems many organisations face when they start down the cloud route.
In Jo’s seven steps, the one that resonates most is “Step 3. Create an orchestration layer that builds and places services based on policies.” Jo says:
Orchestration software is not typically seen in traditional IT environments, because those don’t rely on so much automation. But when it comes to providing cloud attributes, orchestration software is critical, because that layer binds multiple products, technologies, and processes to enable IT-process automation throughout the stack. It helps create workflows or run books that can automate complicated tasks like cluster deployment, host patching, and virtual machine (VM) provisioning.
As we reach the end of the series on ‘Managing Virtual Machine Sprawl,’ we come across a key issue workload centric vs. VM centric management.
Most IT Departments and especially users don’t really want to manage virtual machines, networks, storage, etc. Instead they want to manage applications. They want everything needed to power that application abstracted or hidden away unless they need to access it.
Imagine you have 10 projects, each with 10 servers. All you have to manage them is a portal which shows you 100 servers, with no clear way to identify which ones are which apart from their name. What if that became 100 projects, with 1000 servers total, or more?
By the very fact that provisioning virtual machines has become increasingly easier, the growth and sprawl that is potentially caused can also result in significant new issues for managing the machines.
In the first of my six part series on managing virtual machine sprawl, I highlighted how end user provisioning can lead to wasted resources and money, unmaintained applications, potential security issues and more.
I focused on the requirement for organisations to have cloud orchestration software that can provide detailed access and authentication controls.
Today I’m writing about three important factors for the placement of virtual machines:
One of the key use cases in cloud is being able to easily provision horizontal scaling capabilities, by launching additional virtual machines as and when required. Unfortunately, without the ability to guarantee that each of these virtual machines is located on a different compute node than the others, means that you may not be as resilient as you first thought.
End user self-service is one of the greatest benefits of cloud services, but also one of its greatest weaknesses.
On the one hand, end users can enjoy the benefits of provisioning the resources they require, on demand, without the need to go through layers of requests.
On the other hand, by enabling end users to provision and manage their own resources, virtual machine sprawl can quickly become a problem. Service providers can end up with wasted resources and money, unmaintained applications, potential security issues and more.
1. Authorisation and access controls
2. Placement of virtual machines
3. BSS/OSS integration for validation and compliance
4. Customised interfaces for different usage profiles
5. Granular billing and reporting
6. Workload centric rather than VM centric management
In the first of the series, I’m focusing on the extremely important topic of authorisation and access controls.