Automated tiering: Evaluating the criteria — InfoStor Article by Russ Fellows

Monday, February 14th 2011

There are several technology and marketing messages competing for headlines.  The concepts that resonate are those that make life easier for IT users, lower costs, or do both.  Storage tiering is gaining traction because it provides real value to users by lowering the administrative burden, improving performance, and reducing costs.

This is especially true as IT is pushed to optimize their data centers.  Data center optimization is focused on making existing equipment and processes operate more efficiently.  In order to improve efficiency, either new processes or new technologies must be utilized. For mature, well-run IT organizations, operational efficiency improvements can be difficult to implement and deliver only modest gains.

Automated tiering is a concept that has quickly caught on with IT users and business application owners alike, precisely because it helps run applications faster, and often at lower cost. The age-old ideas of HSM, ILM and other manual tiering and data migration processes are now finally viable due to technology advances.

Two recent trends are driving the demand for better use of storage.  The first is that data centers are literally running out of power, space and budget.  It is no longer feasible to add more capacity as a way to hide the fact that storage is not being used optimally.

The second trend is the use of flash and other emerging solid-state storage technologies to address the rising I/O needs, particularly in virtualized environments.  Due to the relatively high cost and limited capacity of these technologies, it is imperative that data placed on these devices is the most critical to business users.

The rise of automated storage tiering

For years, the ideas behind tiered storage have been well known, but seldom implemented.  In discussing “why not?” with many IT departments, the answers come down to two very good points.

·         It is too time consuming and costly to find data to migrate (cheaper to add more storage)

·         If things are working, don’t change anything (you are likely to break something)

These two reasons are well known and logical as to why manual tiering and migration efforts were unsuccessful, despite their potential benefits.  Rather than risk downtime and spend valuable IT admin resources, it has been cheaper to buy more storage.  From the CEO and CIO to the IT administrator, it is now clear that simply adding more storage is no longer acceptable.

The innovation of high-speed flash storage, combined with automation and transparent migration, enable a new tool to improve data center operations.  Rather than being a manual process, the new generation of tools provides automatic analysis and movement of data.  The key part of the equation is the tools to analyze data usage, or so- called “hot spots” of data access.  This enables storage systems to automatically move the right data to the appropriate storage media, at the right time.

To many, this solution sounds reminiscent of caching technologies that have been used within servers and storage for many years.  Automated tiering is similar to caching in many respects.  So it goes the same attributes are relevant and they include the following:

Feature Optimal Implementation Typical Implementation
Frequency of Data Analysis Continuous Daily
Data Movement Continuous Hourly – Weekly
Level of Automation Depends on environment Varies
Size of blocks migrated < 128MB Varies
Are all tiers writable Yes Depends

Table 1: Tiering Evaluation Criteria

The features listed above are important, but many high-end environments will desire more controls.  Automation is a feature storage administrators desire only with control over when, where and why data movement will occur.

For larger environments, automated tiering is desired, but only with fine-grained controls.  These environments rely upon dedicated storage administrators who want control over when, why, where and how migrations occur.  In this case the value is in the analysis and the process of automating manual steps, with the ability for administrators to review and approve each step.

For systems designed for departmental or remote offices, fewer controls are desired.  Instead, the ideal is that IT generalists who manage many types of IT equipment want automated operations.  Equipment should be setup up initially, and then run itself without intervention whenever possible.

Final Thoughts

Storage tiering is an important tool for data center optimization.  The technologies have been refined over the years and vendors are now rolling out a variety of products that address this need.  Automated tiering solves a real need, and is a topic Evaluator Group’s end-user customers have expressed a great deal of interest in learning more about.

Evaluator Group’s review of storage vendors’ implementation of automated tiering shows a wide variety of approaches.  It is important to review your requirements and understand each vendor’s approach, and the tools they provide.  Vendors are now shipping first- and second-generation automated tiering, but it is imperative to understand the differences in implementation, and which offerings best match your requirements.

Russ Fellows is a senior analyst with the Evaluator Group research and consulting firm.

Connect with us on Twitter and LinkedIn

Forgot your password? Reset it here.