Bringing ILM inside the box
Information lifecycle management
IT budgets are shrinking even as the reliance on data and application availability grows. IT managers must spend carefully to keep from negatively impacting the bottom line and one area where IT departments can realise significant savings is storage. Adrian Groeneveld of Pillar Data Systems provides a breakdown of the actual costs paid for storage when real-world utilisation and performance numbers, in addition to costs, are revealed.
As IT budgets tighten, the growth of corporate information and data mining (the utilisation of information for business intelligence) means that the need for storage is increasing faster than ever. Companies have realised efficiencies with server virtualisation, but their array of utilisation is still likely to be very poor – on average less than 40 percent. And, while there are ways to improve utilisation rates such as to use more existing capacity, the resulting performance impact on applications tends to make those approaches unrealistic because using the capacity will create contention with the existing applications. Thin provisioning can help, but as virtual capacity becomes utilised, companies need to add physical capacity to improve application performance: The vicious cycle has no end.
Achieving efficiency through ILM and storage resource management
Technical efficiency initiatives are common throughout the IT world today. However, most companies see efficiency simply as a function of equipment power consumption and data centre cooling. This makes sense, given that power consumption now accounts for a growing share of operational costs every year. It will soon cost more to power a server than to purchase one. This is only half the story though. An efficient system must also address the total resource constraints of today’s data centre, not just power consumption.
Companies are starting to recognise that even the lowest consumption technologies may not deliver the performance and utilisation essential to ongoing operations. Also, because there is no standardised method for measuring storage efficiency, customers are left to evaluate how significant a manufacturer’s claims really are. Storage efficiency does not begin and end with a single implementation of storage. Rather, finding a vendor that understands that multiple elements must be seamlessly integrated in order to maximise storage efficiency is essential to addressing the inherent issues with data storage through both software and hardware technology.
One way to improve storage costs and utilisation is through hierarchical storage management (HSM). Multiple applications and tiers of storage are consolidated into one single platform, which drops costs and increases utilisation rates to 80 percent or higher, while maintaining the most effective quality of storage for each.
Storage resource management (SRM) and information lifecycle management (ILM) promise to ease management and reduce costs, but there always seems to be a catch. In order to save money you have to buy the technology your particular vendor is selling, which typically involves several storage systems with different price/performance/capacity dynamics as well as the vendor’s associated software, services and maintenance costs.
The thinking behind buying several different arrays is that it is cheaper than buying one big box with the same capacity of the combined four. While this is probably true for the initial capital outlay, it probably isn’t true when factoring in the operating expenses of managing and training for four individual boxes, not to mention power and cooling costs and the data centre space taken up. Additionally, it may be necessary to purchase software that is capable of migrating the data to these arrays based on some policies.
Some vendors have ILM functionality integrated in their database product, but the value of ILM still relies heavily on the customer buying various tiers of storage to dedicate to ILM – usually new storage, as existing arrays are already being utilised. It seems that when implementing an HSM, SRM or ILM solution, saving money can be an expensive proposition.
Vendors need to think ‘inside the box’ and work to solve the storage problems associated with ILM. They must create a solution where all of the ILM functionality is integrated into a single product, making it simple to deploy, implement and manage, which saves both initial capital costs as well as operating and management costs.
Series of storage arrays, designed to economically and operationally deliver the value of ILM with no duplicative hardware or software costs is essential to this process. Users need the ability to configure the price/performance/capacity dynamics of ILM tiers within the array. These ILM tiers map to various price/performance bands when implementing ILM partitions. Once a logical unit number or file system is placed in a price/performance band, it will be placed in a virtualised storage pool associated with that band and given a certain priority when contention for similar resources exists within the array.
Implementing ILM inside your existing database can provide not only cost savings but also a significant improvement in database performance and simplified compliance with regulatory requirements.
To implement ILM, the database administrator must follow four steps:
- Define the data classes,
- Create storage tiers for the data classes,
- Create data access and migration policies and
- Define and enforce compliance policies.
Once ILM is employed, both deployment and management are simple and cost-effective because the functionality is native to the individual products and does not require a third-party application to mediate between the database and storage has been required with previous ILM implementations.
Many useful and novel technologies have come and gone because they were too expensive, too complex or both. ILM is a good idea that has been poorly implemented in the past. The original purpose of ILM was to save costs and reduce complexity; this idea can now become a reality by creating an integrated ILM solution which offers the following capital and operational cost savings:
Buy fewer storage arrays. There is no need to buy multiple tiers of arrays when a single system can support both performance and reliability of the highest tier of storage as well as the cost benefits of near-line disk.
Stop overspending on capacity. By pushing aged data to a less expensive disk and making it accessible from the database, the need for a single, large space of high-performance and expensive drives is diminished.
Simplify management for end-to-end ILM. Provisioning storage from a familiar interface takes the complexity and cost out of storage management.
ILM functionality within the database is both easy to implement and manage. When implemented in conjunction with a single array with integrated ILM functionality, the benefits of end-to-end ILM come to fruition.
Adrian Groeneveld, Pillar Data Systems
(ITadviser, Issue 62, Summer 2010)