Storage Peer Incite: Notes from Wikibon’s March 13, 2007 Research Meeting
Dave Vellante presents Storage services architectures could manage inventories of resources. The thesis is a new model of storage services is emerging that can change the way businesses value storage assets.
The historical norm in the storage world was to deploy storage in response to application requirements. The challenge for storage administrators, procurers and ultimately business people was that different applications feature different benefits, costs and performance characteristics and as a consequence it was difficult to compare and contrast one set of storage investments relative to another. Storage administrative groups, who enjoyed increasing budgets and a greater sense of control over how monies were ultimately allocated, often reduced the complexity and cost of managing storage by introducing standardized "one-size fits all" storage infrastructures. This was at the expense of user departments who are over or under provisioned.
Recently, we've seen the emergence of a storage services architecture model that can do a better job of directly aligning storage resources, and the quality of those resources, to business benefits. The basic idea is applications or business processes that require higher performance, more secure and perhaps more stable storage resources could pay for those resources whereas application users requiring less stringent storage capabilities could pay lesser dollars. This also could permeate into other types of storage services such as backup and restore, migration, data movement, copy services, etc. The architecture presumes significant processing power at the controller level capable of operating at very rapid speeds even as they are running software that borrows heavily from application domains like supply chain management.
The result of this is that we expect companies like EMC, Hitachi, HP, IBM and others to promote high level storage administration interfaces reminiscent of some of the early efforts at deploying supply chain management technologies in the storage world. There are still gaps that have to be filled between the current state of the technology and this future state where virtualization makes it possible to attach a dollar return to a dollar spent on storage. Users however should begin the process of categorizing their storage investments in terms of business returns such that they can facilitate the migration to a more business-centric view of storage assets provisioned through a services orientation as the technologies come on line.
Action item: Storage administrative personnel will have to get down and dirty with the ugly reality of chargeback to start the process of conceptualizing how storage dollars spent generate business value return. This will permeate not just the buying and allocation of disk but also other critical storage assets from hardware to controller to software and operations.
Storage chargebacks are a pain. The business naturally wants to understand what they're getting for their dollar but at the same time, the information available on actual consumption of resource is anemic. Maybe storage hardware is roughly tracked but software and services, and increasingly automated services are not given much attention.
The key issue with chargebacks is IT has no good way to meter, monitor and charge accordingly for the consumption of service to the business. A storage services architecture has the potential to change this by allowing application owners to purchase only those services they consume. This necessitates a change in the way chargebacks are fundamentally viewed and requires new business models for storage provisioning.
Action Item: Storage administrators need to begin constructing new, pay-as-you-go models of storage that reflect the consumption of storage and related services versus some vague concept of chargeback. Importantly, IT must provide transparency using clear metrics that tie business value to the delivery of storage services.
Some storage administrators will resist any move towards storage services architectures that threaten their ability to dictate storage policy. They are missing the fact that user departments know that they can purchase a 300 gigabyte drive from Fry’s for $100, and they do. When they have to, they come to IT, and when they don’t they jury-rig solutions.
Huge amounts of unstructured corporate data lie distributed and unmanaged within corporations. Storage administrators can extend their role to provide relevant storage services to all departments, and reduce the risks and costs to their organization of storage anarchy.
Action item: Storage administrators and IT should push for the implementation of storage services architectures as a way of demonstrating the value that they bring to the organization, and to start the process of bringing unstructured data into the data center and under control.
Storage services architectures will allow much greater control of the usage of storage resources and functionality. User departments will be happy to pay for storage function which directly improves their business efficiency, such as improved performance. But what happens when a user department wants to cut the cost of (say) remote replication, which impacts the recovery of the organization as a whole? Who should pay if legal are asking for archiving services to protect the company as a whole from being sued? The good news is that storage service architectures can provide transparency to the cost of those services; the current storage architectures often hide significant costs and pass them arbitrarily to user departments. However, the business and IT will have to develop processes that allow key stakeholders to be able to understand the business case for the usage of advanced services, and negotiate and agree funding for them.
Action item: Storage services architectures mean that organizations need to put in place cross-functional processes that will allow discussion and agreement of advanced storage services that affect the organization as a whole.
As storage services architectures emerge, the flaws in storage suppliers' chargeback strategies will be increasingly exposed. Chargebacks are problematic for users. On the one hand, busines lines want to understand exactly what they're paying for and be charged only for those services they consume. On the other hand, the hassle of constructing a chargeback system and auditing its accuracy is often not worth the effort. So most organizations simply perform a cost allocation exercise and charge the business based on some pre-conceived notions of usage. What's more, most vendors charge between $1,000 - $7,500 per server for chargeback licenses, prompting many users to run their chargeback systems on Excel spreadsheets.
A storage services architecture enables the granular provisioning of storage services based on application or business requirements. Chargebacks should be viewed as analytical tools to understand usage patterns, consumption of resources, service levels and importantly, contribution to the business. For example, products should not just track downtime, but the cost of that downtime as measured in lost revenue or productivity-- directly tied to the storage infrastructure. Specific goals such as time to provision, RPO and RTO, cost per GB per tier and other key performance indicators should be readily available and customizable.
Action Item: Leading storage vendors should provide chargeback software for free as they develop storage services architectures. Vendors should look at chargebacks as a demand generator, providing software to enhance understanding of usage patterns and the concomitant purchase of services to the business. This ultimately will drive demand for gigabytes, software and services.
Transformation to a storage services architecture (SSA) is comprehensive, touching virtually every aspect of storage. Different suppliers of storage hardware, software, and services will invest heavily to position their goods as "apex offerings": the investment around which all other project investment, implementation, and management offerings should revolve. However, while the adoption of new technologies, creation of new sourcing relationships, and formation of new management structures will be important, instituting a system of storage metrics that are both semantically usable, as well as syntactically administrable, in business terms will be lynchpin of success. Ultimately, user, developer, legal, procurement, and finance communities inside the business must adopt an SSA metrics system for IT to successfully administer storage resources as a set of services, otherwise parochial issues will inevitably tear the SSA apart.
Action Item: A program for implementing a common, meaningful, and manageable set of SSA-related metrics must be a top priority in any serious SSA implementation effort, otherwise the IT organization will always be in transformation mode, and never quite reach delivery mode.