Published:
October 26 2010, 11:33 AM
|
no comments
by
Shirief Nosseir
Yesterday, I mentioned four types of sprawl that organizations commonly face when moving to virtual and cloud environments. Today let's look at the first type: virtual machine (VM) sprawl.
VM sprawl is the most commonly cited type of sprawl. What's important, but often not mentioned, is that we need to distinguish between two primary problems when tackling VM sprawl.
The first, and more well known, boils down to concerns over the weed-like growth in the number of VMs that get introduced (in articles it usually goes something like this "as users realize how easy and quick it is to create virtual machines..."). This issue is compounded by the fact that often these VMs never get decommissioned, even after having no business reason for them to continue running. From an IT management perspective, this results in a catch-22 situation: as VM sprawl leads to unnecessary overuse of infrastructure resources (since they use up memory, CPU cycles, energy, etc), it defies virtualization's promise to maximize resource utilization, lower costs and reduce energy consumption. Security wise, as long as all the VMs remain configured according to a standard build, the problem here stays limited to an increase in the surface attack area as well as the burden on our resources (human and otherwise) to secure superfluous server workloads. It's true that from an administrative perspective the appeal of virtualization includes how easy it is to provision, clone and move VMs and how this can be done at the touch of a button or even automatically (which is a key capability for reaching higher levels of IT agility). However, if the full lifecycle (e.g., configuring, provisioning, monitoring, cost accounting, deprovisioning, etc) of VMs is not well thought of and managed from the very start, our virtualization efforts will backfire at us.
The real security issue, though, lies in the second problem related to virtual sprawl. The concern here is more about the proliferation of rogue VMs due to inadequate IT controls. It's about ensuring that VMs remain consistently configured and hardened and that they are properly managed and secured across all our various virtualization platforms (since similar to how we typically rely on more than one operating system, we will also rely on more than just one hypervisor platform. And as far as security is concerned, all platforms need to be equally secure for purpose). This is not exceptionally difficult to achieve, especially in organizations with more mature and automated policies and processes. However, simply applying the same change and configuration management practices currently in place for our physical servers is not a viable approach. Many of the processes currently in place are not optimized to address the dynamic nature of virtual and cloud environments and applying them without a revisit would severely impact our agility. Another notable challenge here is around managing and securing our dormant VMs, and ensuring that when they wake up they are compliant with our corporate security policies and configured according to the relevant standard build (somehow similar to how we apply changes to physical server machines that are switched off).
Embarking on virtualization and cloud computing initiatives without addressing change and configuration management implications is a sure recipe to introduce significant operational problems as our environment grows far more quickly than our ability to control it. Make no mistake, virtualization and cloud computing are disruptive innovations and require us to revisit how we can deliver business value through IT services. We need to reconsider our organizational, operational, functional, architectural and technological requirements and capabilities, to name just a few aspects. What's exciting though is that there's a genuine desire for change and the IT industry is willing to make the leap forward. Granted, it's not an easy ride, but there is no reason why it cannot actually enable us to reach new levels of automation that power the agility and efficiencies we're desperately striving to reach.
In tomorrow's blog post, I'll talk about entitlement sprawl.
Other blog posts in this series:
1. Securing your road to virtualization and cloud: Welcome to the sprawl galore
3. Entitlement sprawl: Are you managing privileged users in your virtual environment?
4. Data sprawl? Content-aware identity and access management to the rescue
5. Taming the wild west of cloud service sprawl