At a recent community event, I met “Scott”, a savvy IT Pro who was in the process of planning the architecture for a new datacenter location at his company. Scott is an experienced datacenter engineer who manages several other virtualized datacenters. Although he is currently running VMware for the hypervisor platform in his existing datacenters, he was intrigued by Windows Server 2012 and the free Hyper-V Server 2012 enterprise-grade bare-metal hypervisor, primarily because he had been hearing good things about Hyper-V from peers who had been evaluating Microsoft Private Cloud in their organizations.
So, Scott decided to go on a “shopping trip” to explore what would be involved with implementing the Private Cloud solution he was looking for in his next datacenter. What Scott found really surprised him in terms of capabilities and dramatic cost differences when leveraging Microsoft Private Cloud vs the related VMware offerings. Read on to learn about Scott’s findings … you may be shocked too!
This article provides a summary of Scott’s interesting journey and reports on his findings from his research on implementing Private Clouds using Microsoft Private Cloud vs VMware vSphere, vCenter and vCloud offerings.
NOTE: All example costs provided in this article are based on current ( as of this article’s date ) retail pricing for each product. Although Scott’s organization qualifies for discounted pricing levels of both Microsoft and VMware products, comparing retail prices provides the best “apples-to-apples” comparison on a broader basis.
For current retail prices, Scott used the following publicly available resources online:
You may wish to leverage these tools for your own comparisons to determine if your results are the same as what Scott found.
Microsoft and VMware virtualization solutions include a huge list of capabilities and features. However, for Scott’s evaluation, he chose to forego the “feature hype” and identify the specific Private Cloud requirements that he would need for his new datacenter.
Enterprise-grade virtualization. Private Cloud solutions certainly begin with enterprise-grade virtualization as a foundation, and Scott’s scenario was certainly true to this. Before exploring the more advanced Private Cloud capabilities, validating that a hypervisor platform had the capabilities for high-density scalability, highly available clustering, easy multi-server management, VM mobility, dynamic resource leveling and centralized virtual infrastructure provisioning were all core criteria for Scott’s new datacenter.
Pooled resources. But, in his new datacenter, Scott was also looking to move beyond mere server virtualization into a full Private Cloud solution. His company had grown to the point of complexity where the ability to manage a virtualized infrastructure as pooled resources would be valuable – particularly because of the elastic nature of his company, which had grown and transformed through a variety of mergers and acquisitions. The ability to abstract his compute, storage and network infrastructure into resource pools that could grow and shrink in alignment to his business needs was really appealing.
Self-service and Delegation. But, Scott was looking for more than just pooled resources out of his next Private Cloud solution … his company also had several large development teams, and those teams regularly needed new virtualized dev/test environments to be spun-up and torn-down – sometimes they seemed to have new requests on an almost daily basis. Much of Scott’s current time these days was spent catering to the infrastructure needs of one development team or another in his organization, and other projects were piling up on his desk as a result. For his new datacenter, Scott was looking to carve out one or more Private Clouds that could be delegated to these development teams so that they could manage their own virtualization needs within the confines of the resources and templates that Scott allocated to them.
Application Workload Monitoring and Reporting. Scott saw monitoring and reporting needs as becoming increasingly more granular. In his existing datacenters, he primarily was focused on monitoring the infrastructure, or virtualization “fabric”, components. While this level of monitoring certainly provided great value to help ensure overall datacenter health, he was also finding that, more and more, he needed to have deeper visibility into the applications running across each set of VMs – to assist his application engineering and development teams with application hygiene, troubleshooting and problem resolution.
Automation. Scott was also looking to implement a specific level of Private Cloud automation. His team had previously spent a fair bit of time defining a number of standard “enterprise architectures” for common application models that they deployed internally – single-tier apps as well as multi-tier apps. Although these standards were well-documented as “standard operating procedures (SOPs)”, he still spent a ton of time following the hundreds of manual steps spelled out in these SOP documents when provisioning some of the more complex architectures for new application workloads.
Charge-back / Show-back. Finally, Scott was finding himself being increasingly asked to breakdown how various departments in his organization were using datacenter resources by his IT Director. IT in his company was looking to poll those departments for budget dollars to manage and expand the infrastructure that they depended upon, and they needed this information to show justification. They were also planning to use this information to report back to the Senior Execs on how IT was empowering those departments to drive value in their business.
With his Private Cloud requirements in-hand, Scott set out to explore what would be involved with implementing his new Private Cloud, comparing relevant solutions from VMware, his incumbent virtualization platform, to the Microsoft Private Cloud offerings.
Hardware “Scale Units”
Of course, virtualization applications, management tools, and the hypervisors themselves all ultimately run on-top of hardware. For his new datacenter, Scott started with developing a modular hardware design, where each rack within his new datacenter would be populated with storage, top-of-rack (ToR) network switches and server hardware as a datacenter “scale-unit” - sometimes called a “stamp” in the high-density datacenter world. Typically, “scale-units” are sized to match the average new capacity expected to be required in a datacenter on a quarterly or semi-annual basis, and as such the configuration of a scale-unit can vary on an organization-by-organization basis from a few servers to several racks of equipment. Scott’s concept behind his scale-unit-in-a-rack approach was to drive down hardware costs by leveraging a standardized commodity hardware configuration for each rack that would make growing overall datacenter capacity easier by adding new racks of equipment with common hardware configurations. In Scott’s design, each new rack would start with 20 scaled-up dual-processor multi-core hypervisor host servers, an iSCSI SAN and LAN/SAN network switches.
In this article, we’ll focus on comparing the necessary software components to meet the Private Cloud requirements that were important to Scott for a single scale-unit, or rack, of 20 hypervisor host servers.
Guest Operating System support
Before diving too deeply into the virtualization and management components, Scott wanted to confirm that both VMware and Hyper-V could meet his Guest Operating System (OS) needs in the new datacenter. Like many organizations, Scott’s company primarily runs virtualized Windows Server workloads, but he also needs to support virtualized Linux & Oracle workloads. While VMware supports all of Scott’s required Guest Operating Systems, Scott was pleased to find that Oracle has certified Windows Server 2012 Hyper-V as a directly supported hypervisor, something that they have not done for VMware, and that a variety of common Linux distributions are directly supported by Hyper-V, including CentOS, Debian, Oracle, Red Hat, SUSE and Ubuntu.
With Guest OS compatibility in check on both platforms, Scott wanted to make sure that he accounted for the required Windows Server Guest OS licenses in each scale unit by including the cost of Windows Server 2012 Datacenter Edition, which provides unlimited Windows Server OS licenses for VMs on each licensed host.
Bare-metal Enterprise-grade Hypervisor
Next, Scott turned his attention to the hypervisor. Scott wanted to ensure that the core hypervisor would scale to high-levels of density and also support a highly-available clustered hypervisor host scenario so that VMs could failover to remaining hosts in the event of an unplanned host server outage. In addition, being able to move running VM’s and VM storage between hosts to support planned host maintenance needs without application downtime was a core requirement. Finally, Scott was looking for built-in VM replication between scale-units ( racks ) either at the same datacenter or across datacenters for additional disaster recovery protection.
In terms of the Microsoft solution, Scott found that his Windows Server 2012 Datacenter Edition licenses, which he already planned to purchase to license the Windows Server OS inside his VMs, included the Hyper-V bare-metal hypervisor at no additional cost and could meet all of these core hypervisor needs. In the VMware solution, Scott concluded that he would need to purchase at least VMware vSphere 5.x Standard Edition to support these requirements.
Multi-Server VM Management Tools
The hypervisor itself doesn’t provide much value unless we can easily manage the hypervisor hosts on a centralized basis. In the Microsoft solution, Scott found that Windows Server 2012 already includes multi-server hypervisor & VM management tools. In the VMware solution, multi-server management is provided via the separately licensed VMware vCenter Server. So, let’s add in Scott’s management tools cost …
Dynamic VM Workload Balancing and Power Management
For each scale-unit ( rack ), Scott wanted to make sure that the hypervisor resources were being leveraged in a manner that would contribute to the best utilization of hardware at the lowest possible power consumption costs. To support these needs, Scott discovered that Microsoft System Center 2012 SP1 supports Dynamic Optimization (DO) and Power Optimization (PO) for Hyper-V, VMware vSphere and Citrix XenServer hosts. This multi-hypervisor support appealed to Scott because although he may initially implement each scale-unit as a homogenous hypervisor solution, he didn’t want to risk being locked in with a single hypervisor vendor for future plans.
In the VMware solution, Scott reported that he’d need to “step-up” from vSphere 5.x Standard Edition to Enterprise Edition to support dynamic load balancing and power management across vSphere hosts via the VMware DRS and DPM feature sets.
Let’s add these costs into our shopping list …
Profile-driven Storage, Distributed Switches and Bare-metal Host Deployment
Provisioning storage, virtual switches and hypervisor hosts quickly and consistently in a large datacenter environment can be a real chore. To reduce the provisioning time for these resources, Scott added centralized provisioning and management of these resources to his Private Cloud shopping list. In the Microsoft solution, Scott found that System Center 2012 SP1 already includes profile-driven storage, distributed logical switches and bare-metal host deployment.
In the VMware solution Scott found that he’d need to “step-up” again … this time to the top-of-the-line vSphere 5.x Enterprise Plus edition of their hypervisor.
But Wait … We still don’t have a Private Cloud yet!
At this point, Scott stepped back to review his progress and he realized that so far he’d only assembled the components for a highly virtualized datacenter environment. He still had his Private Cloud needs to address around pooled resources, delegation, self-service, deep application insight, automation and charge-back/show-back. While reviewing these Private Cloud requirements, Scott realized that Microsoft System Center 2012 SP1 actually does provide all of these Private Cloud capabilities, too. Scott was really pleased to learn that he could start with a single purchase of System Center 2012 SP1, leverage it right away for managing a highly virtualized datacenter, and then easily progress down the path of full Private Cloud management without additional licenses and management tools being needed. Just like the Dynamic Optimization and Power Optimization capabilities discussed earlier, System Center 2012 SP1 can also manage entire Private Clouds that span multiple hypervisors – Hyper-V, VMware vSphere, or Citrix XenServer – thus avoiding being locked-in to a single hypervisor for Scott’s Private Cloud down-the-road.
In the VMware solution, Scott learned that he’d need to “step-up” one more time to vCloud Suite Enterprise Edition to gain access to all of his requirements in the context of Private Cloud. In particular, Scott’s Private Cloud needs around providing deep application insight for monitoring, automation and configuration management drove him towards the Enterprise Edition, rather than Advanced Edition, of vCloud Suite which adds these application-centric features to VMware’s vCloud Suite family of Private Cloud offerings.
Scott surmised that Microsoft has embraced Private Cloud as a core component of their datacenter management strategy revolving around Microsoft System Center 2012 SP1, whereas VMware charges a premium cost for organizations wishing to extend their datacenter management practices beyond mere server virtualization into a complete Private Cloud solution that can be aligned to evolving business needs.
When beginning his Private Cloud shopping trip, Scott realized that both solutions offered a long, seemingly never-ending, list of “features and capabilities”, but he decided to “cut through the hype” and narrowed his comparison down to the specific Private Cloud requirements that he knew he’d need.
Scott found that both solutions could certainly deliver on his Private Cloud requirements, albeit at very different cost structures as summarized below.
Based on the large cost differences between each solution and the ability for heterogeneous hypervisor management via Private Clouds powered by Microsoft System Center 2012 SP1, Scott chose to pursue the Microsoft Private Cloud solution for his new datacenter. He’s currently in the process of formally building a Microsoft Private Cloud pilot lab environment that will serve as the basis for the production design in his new datacenter.
Be sure to check out these additional resources:
Keith Mayer is a Senior Technical Architect at Microsoft, focused on helping ISV partners leverage the Azure cloud platform. Keith has over 20 years of experience as a technical leader of complex IT projects, in diverse roles, such as Network Engineer, IT Manager, Technical Instructor and Consultant. He has consulted and trained thousands of customers and partners worldwide on design of enterprise technology solutions.
Keith is currently certified on several Microsoft technologies, including Azure, Private Cloud, System Center, Hyper-V, Windows, Windows Server, SharePoint, SQL Server and Exchange. He also holds other industry certifications from VMware, IBM, Cisco, Citrix, HP, CheckPoint, CompTIA and Interwoven.
You can contact Keith online at http://aka.ms/AskKeith.