Wednesday, October 22, 2008

Scalability – a multidimensional challenge for storage managers

Traditionally scalability has been myopically viewed in terms of capacity and growth demands and addressed by simply adding more storage. However, an extreme rate of data growth, greater demands for online data availability and accessibility and a growing demand for operational efficiencies (power, space, management), are all trends that question the viability of this lazy mans approach to managing storage growth and increasingly demanding service levels. Unfortunately however, this tradition still lingers and is SOP in many data rich enterprises.

Scalability is a multidimensional challenge and while just one of many selection criteria, its importance requires deeper thinking. Data growth, increasing application demand, service level expectations and vendor product roadmaps should be understood and balanced when considering the complex question of scalability.

Data Classification: While not strictly a scalability issue understanding your data should be the first question to answer. Why? Different data has different characteristics that drive different storage requirements which in turn drive key cost factors (CapEx and Opex) and ongoing service delivery growth options.
• Data Growth: Does capacity expand easily and transparently to match projected data growth over an acceptable time horizon? Suggested planning horizon would be 3 to 5 yrs.
• Time: Does the solution scale through time meaning can it be upgraded/enhanced easily with newer technology as it becomes available over the life of the product. Such scalability increases the useful life of the product and drives greater return on your initial investment? Examples would be higher capacity drives, new storage technologies such as SSD, compute resources and connectivity, Understand the vendors roadmap and match it to your future requirements and beware the smoke.
• Performance: Is application growth understood? Will they drive an increase in supported users? Will they increase the time sensitivity of the data? Will they increase the complexity of the queries and workloads? All are business driven requirements which drive greater demand for I/O, faster response time expectations or at least the same response time while servicing greater workloads and increases in ingress/egress data rate demands? If the answer is yes to any or all of the above, how easily and transparently can performance be enhanced?

Budget dollars are scarce and are likely to get scarcer. So when committing to a storage solution make sure it can grow with you without major turmoil or service interuption. Select a planning horizon of 3 to 5 years and plan for the future. Know the limits of the proposed solution and plan accordingly.

No comments: