With the continued evolution of the ‘always on’ age and data-driven business strategy, the need to reengineer the data center around new principles has become evident. In a recent Executive Brief on Cloud-Scale Data Centers, Microsoft noted the strategic differences between the way enterprise IT has typically planned data center services, and the unique perspectives that make the strategy ‘cloud-scale’. (See Figure 1)
| Infrastructure | Enterprise IT | Cloud Scale |
|---|---|---|
| Operability Metrics | Mean Time Between Failure (MTBF) | Mean Time To Recovery (MTTR) |
| Architecture Resiliency | Active/Back-Up | Active/Active/Active |
| Hardware Components | Highly Customized | Commodity |
| Deployments | Highly Customized | Highly Automated; Measured in Minutes |
Figure 1: Cloud-Scale Thinking (Ref: Microsoft Corporation. “Cloud-Scale Data Centers”. Executive Strategy Brief, 2013.)
The deployment of commodity hardware in active/active/active architectures is no longer a pipe-dream, it has become de rigueur for any forward thinking CIO planning for high availability AND true resiliency. But the change in thinking required to truly implement on these concepts means a lot more than cheap hardware and deployment automation.
To really implement cloud-scale thinking, IT teams need to move beyond simple converged infrastructure, where pooled infrastructure is shared amongst multiple business units to improve asset utilization. In its place comes the converged IT stack which aligns software, hardware and infrastructure around service workloads. Workload segmentation has been gaining proponents at a rapid rate, with IT teams acknowledging that the divergent demands of many application environments are best addressed by workload-specific architectures. It turns out that for developing cloud-scale thinking, this same segmentation strategy is equally important.
However, moving from concept to reality is not easy. For small businesses with just a few critical applications running, the task is obvious and simple. But for the average enterprise running myriad applications across diverse hardware platforms located in geographically-dispersed data centers, just getting an accurate inventory of what you are running can be a herculean task.
As a result, moving to a converged IT stack should be considered a process path as opposed to a project. Workload segmentation has become a critical consideration in virtually all aspects of data center strategy, including performance optimization, security and resilience. So, the basic concepts of segmentation cannot be ignored.
To begin, consider elements like:
- Programming architecture
- Workload use characteristics, especially market specific patterns
- Data load and operational performance considerations (multi-threading, parallel processing, etc.)
- Logical to physical mapping of current architecture
It is typically rather easy to define common workload systems requirements for many applications. However, the last bullet…mapping logical to physical devices is often the most complex in large enterprises. Deployment diffusion and presumptive failover strategies typically lead to complex maps that take time to understand. The required discovery phase work can require serious investment, but cannot be bypassed. Thoroughly understanding how individual workloads must perform and how they need to be deployed/accessed is the essential foundation for beginning the converged IT stack process. Building out the mapping of that deployment strategy is the subsequent start point for creating an actual converged IT stack program.
BRUNS-PAK’s End-to-End Support Program
From assisting in workload segmentation strategy through construction and deployment of state-of-the art cloud-scale infrastructures using either traditional or modular/pod construction strategies, BRUNS-PAK offers IT clients a unique, comprehensive services support. BRUNS-PAK has been a leader in the planning, design, construction and commissioning of some of the most advanced data centers in the world. Our proven “Three (3) Step” Process:
- PHASE I – Consulting/Options-Budgets
- PHASE II – Design/Plan
- PHASE III – Build/Deploy
results in a vendor neutral data center plan that includes options/alternatives/cost estimates and schedules and that accounts for both short-range and long-term capacity considerations.
To learn more about how BRUNS-PAK can help you build and implement a comprehensive data center plan for reaching ‘cloud-scale’, contact Jackie Porr at 732-248-4455 or via e-mail at jporr@stage-14xghe2.dream.press, or by using the Information Request Form.
