Datacenter & Infrastructure Modernization
Transforming a Predominantly Physical Environment Into a Modern Two-Site Infrastructure Platform
VMware • Compellent SAN • Dell • Force10 • 40Gb Networking • Multipath iSCSI • Cisco • HA/DR
When I joined a previous employer in 2011, much of the infrastructure still ran on physical servers. The environment included an aging IBM Fibre Channel SAN, mixed server and network equipment, and systems that had accumulated over time as the business grew.
The goal was not simply to virtualize servers. The larger objective was to build an infrastructure platform that was easier to operate, more resilient, more efficient, and capable of supporting the company's future growth.
Virtualization Wasn't New to Me
I had already worked with virtualization technologies in previous roles, including Oracle VM, Virtual Iron and later Citrix XenServer.
The important concept was not the particular hypervisor. It was separating operating systems and applications from individual pieces of hardware.
That changes how infrastructure can be maintained, expanded and recovered.
At this company, VMware became the foundation for taking that approach much further.
Building the New Platform
The new platform was built around VMware virtualization, Dell compute, Compellent shared storage and dedicated datacenter networking, with MPIO iSCSI.
Migrating the Workloads
Moving from a predominantly physical environment to virtualization required more than simply converting every server into a VM.
Some systems were good candidates for physical-to-virtual migration. Others were better handled through native application or data migration onto newly built virtual machines.
Legacy systems required particular care because application dependencies, operating-system support and hardware assumptions could complicate migration.
The objective was always to move each workload into the most appropriate supported architecture while minimizing operational risk.
Modernizing the Datacenter Network
Virtualization and shared storage changed the requirements of the network.
Server-to-server traffic, storage traffic and VM mobility placed very different demands on the infrastructure than the predominantly physical environment that preceded it.
Cisco was considered, but the cost was difficult to justify for this particular environment. Force10 had recently become part of Dell and provided an attractive combination of datacenter switching capability, performance and cost.
The resulting Force10 architecture used a 40Gbps QSFP backbone, with 10Gbps connectivity for the modernized server and storage environment. Cisco remained part of the broader network infrastructure.
This was an example of choosing technology based on the requirements rather than standardizing on a vendor simply because it was already present elsewhere.
Storage and Compute
The platform combined Dell servers with Compellent shared storage and VMware. The legacy IBM SAN had used Fibre Channel. The replacement Compellent architecture instead used multipath iSCSI, initially over four 1Gbps connections and later over redundant 10Gbps links following the network upgrade. This provided multiple paths between the compute and storage infrastructure while significantly increasing available bandwidth as the environment evolved. Separating compute and storage also provided greater flexibility for maintenance, capacity planning and hardware replacement while removing individual physical servers as dependencies for most workloads.
Reducing the Physical Footprint
As workloads moved onto the VMware platform, the number of physical servers dropped by 75%. That reduction had benefits beyond rack space. Fewer physical systems meant less power consumption, less cooling demand, less cabling, fewer hardware components to maintain and fewer individual hardware failures that could directly affect an application. At the same time, the remaining infrastructure became more capable and much more resilient.
Power and Cooling
Consolidation increases the importance of the infrastructure supporting the remaining equipment.
When many workloads depend on a smaller number of virtualization hosts, storage systems and network switches, power and cooling failures can have a much larger impact.
The datacenter therefore included protected power and redundant dedicated cooling as part of the overall infrastructure design.
High availability is not just a server or storage feature. It depends on the entire environment supporting those systems.
Building a Second Site
A second facility extended the architecture beyond a single datacenter.
The secondary site supported both active workloads and disaster-recovery capabilities, using a similar architectural approach adapted to the requirements of that location.
This introduced another level of infrastructure planning around workload recovery, data replication, network connectivity and site-level redundancy.
The objective was no longer simply protecting against the failure of a server or storage component.
The architecture also had to consider the loss of an entire facility.
Vendor Selection Is an Engineering Decision
One lesson from the project was that enterprise architecture does not require buying every component from the same vendor.
Cisco was already an important part of the broader network environment, but that did not automatically make it the right choice for every datacenter layer.
Force10 provided the performance, topology and economics required for the datacenter network. Dell servers provided the compute platform. Compellent provided shared storage. VMware provided the virtualization layer.
Together they formed a coherent architecture even though no single vendor supplied the entire stack.
The technology decision should follow the requirements, not the logo.
The Result
The modernization transformed a predominantly physical environment into a shared, resilient infrastructure platform incorporating:
- VMware virtualization across shared compute resources
- Compellent shared storage using multipath iSCSI
- 10Gbps server and storage connectivity
- 40Gbps Force10 QSFP datacenter backbone
- Redundant power and dedicated cooling
- Two-site infrastructure supporting production and disaster recovery
- 75% fewer physical servers, and reduced space utilization to one quarter of original size.
The result was not simply fewer servers.
It was a more resilient and manageable architecture in which compute, storage and networking could be maintained and expanded independently.
The project also reinforced an approach that has remained important throughout my career:
Understand the requirements, identify the real constraints, and design the architecture around the problem.