Backup storage tiers offer business units choices regarding how their data will be protected and how much that protection will cost.
Defining, ranking and implementing storage tiers for backup services is tough. Part of the challenge stems from the fact that there are no tools that make it easy to carve storage technologies into tiers of service. The problem is compounded by storage staffs that don't think in terms of delivering, monitoring and charging for different levels of backup service.
During the past decade, storage managers have been aggressively upgrading their backup capabilities by incorporating advancements in hard disk, tape drive, virtualization and software management technologies. Those efforts have improved backup performance by accommodating shrinking backup windows and the need for better system recovery times.
Storage admins are being asked to rebundle backup capabilities in terms of gold, silver and bronze levels with corresponding prices. For example, within months of taking the helm at MoneyGram International, senior capacity planner Tom Becchetti built a five-tiered service availability model. The main attribute defining each service level at the St. Louis Park, MN, financial services firm was recovery time objective (RTO). Tier one promised no data loss and an aggressive backup-to-disk configuration, while tier five made no promises with regard to data loss. Tiers two through four offered gradually more expansive RTOs and backup functionality.
|Attributes of backup service levels|
|Click here for a comprehensive list of attributes of backup service levels (PDF).|
Because this was MoneyGram's first foray into internal OS and storage service-level agreements (SLAs), Becchetti decided to take an integrated and basic approach that combined backup with other storage-related technologies, allowing every application to be lumped into one of five different tiers. Other companies have taken a different and more difficult tack, choosing to develop tiers specific to backup functions (see "Attributes of backup service levels," PDF). This allows backup capabilities to be reprovisioned and rearchitected along tiered lines so that they can be served to the business units on an a la carte menu.
What the lines of business receive from their IT service providers is best described as an operating-level agreement (OLA). The OLA specifies the services to be provided by an infrastructure group as an "offering" to its customers. Customers use the OLA to better understand what products are offered by each infrastructure group and can then choose which services they need to support their apps while knowing the cost of the services.
So how does an OLA help a storage manager and not just add another administrative burden? For starters, the OLA lets everyone know what to expect from everyone else. Becchetti, for example, had an application that generated a large number of media files, mostly .wav files. The application has a feature that trashed these files based on various space-saving parameters, but the application's owner was unaware that, based on the backup settings in place at the time, only 30 days of backed up data was available--no archived copies. When the business required .wav files older than 30 days, there was no way to recover them. With an OLA in place, there are no surprises about how the data is--or isn't--protected.
Asked what he would have done differently based on what he knows today, Becchetti says, "I wish we would have written the SLA documents before we implemented the tiered concept and not after. Also, but less importantly, we should have identified the ramifications of handling the various one-off exception requests to each tier we received to accommodate the customer goodwill. These [exceptions] can often unravel the purpose of a service-level [agreement]."
|Backup service tiers|
|Click here for a comprehensive list of backup service tiers (PDF).|
Although every organization has its unique tiering definitions, they tend to follow broad outlines. The "Backup service tiers" chart (PDF) shows some of the components and attributes commonly found in a multitiered storage architecture. Noticeably missing is the degree and extent to which backup functionality is integrated into the different tiers of other services, such as networking, security and database admin.
Storage OLA tiers capture those services that can be rendered at various levels and then priced accordingly. Storage managers usually align their online disk- serving capacity by identifying storage services unique to each array type; for example, a tier-one service might be array-based replication.
To charge or not
Many OLA programs assume that costs can be efficiently identified and then allocated to user communities based on their consumption of IT services. Although implementing a tiered approach isn't dependent on having chargeback, a pay-for-services model keeps users focused on what's possible. Without this mechanism, business users gravitate to the tier with the most features--provided its cost isn't applied to their budgets. High-end backup costs can be steep. Take, for instance, backup tiers with zero RTO values and other advanced backup features such as disk-to-disk backup or long-term archiving, which can cost as much as $50/GB; this could mean that the cost of handling 10TB at a tier-one backup service level could exceed $500,000.
Some of the logic inherent in these calculations is a bit arbitrary because there are certain costs in the installed backup topology, media servers and base licenses that need to be amortized and attributed to all the tiers. How these costs are accorded to each tier can make all the difference in determining how much application owners will pay to receive their backup services.
Service tiers vs. service catalogs
David Kaercher, vice president of core services at Allianz Life Insurance Company of North America in Minneapolis, offers his business units a "catalog" of storage services with their associated costs. This type of catalog focuses more on describing storage capability in generic terms that can be understood by non-technologists. It contains none of the guarantee language usually found in an OLA or SLA.
Many storage managers have a difficult time envisioning how to separate, define and stratify backup functionality into service tiers. Backup service levels can become remarkably complex. There are many options a customer can take, even within a given service level. The temptation for the service manager is to simplify the choices to ease the burden of implementation, maintenance and reporting. Conversely, business units will want to expand their options within a tier. Of course, there's a delicate balance between cost, complexity and manageability. The key for storage managers is to help their users make good decisions by understanding the tradeoffs among the different service options.
Storage managers should consider the following before laying out their first tiered-backup service offering:
- Choose the right media. The media choice establishes the boundaries for backup and recovery times, and determines the ability to move data between business apps for other data protection purposes. The appropriate media type for backup is determined by the attributes of each data class, speed of reading data, physical storage parameters and document storage restrictions that may dictate a media that will maintain bit-level integrity for a given number of years, prevent rewrites or allow reliable deletion.
- Define the data protection infrastructure. Three interrelated factors--number, frequency and location of the copies--define the data protection infrastructure.
- Support the company's mission. Companies are increasingly realizing their core competencies through IT strategy, such as faster business cycles, sophisticated risk management schemes and more efficient information exchange. OLAs must support this corporate direction by accommodating changing workloads and investment priorities.
The following steps briefly explain how to start a tiered backup services program:
- Take an inventory. Identify, document and place a cost on deployed hardware, software, agents and staffing.
- Interview your customers. Discover what's most important to them in terms of the technology you have to offer. Ensure your questionnaires are in language that can be understood, so user feedback is meaningful. Pose questions in terms of cost/benefit tradeoffs so users appreciate that there's no "free lunch" in deploying different types of data protection solutions.
- Craft some pro-forma storage tiers. Stratify the features and functionality of data protection technologies currently in place. Identify the gaps where users want what you don't have. Present the services in each tier in business language, not arcane vendor verbiage.
- Develop formal OLAs that reflect these tiers. Convert the feature and service list, including things like "guaranteed incident response time," and extract the requirements most in need of codifying in a signed document.
- Build in iterative review. Make it clear to business units that your first attempt to build a tiered-backup service program will likely need modifications. As frequently as necessary, update the OLAs and the underlying service tier documents.
- Expand your services over time. As part of the review process, expect to gradually expand the number and types of services in each tier, as well as improve the cost-effectiveness of those services.
After your program is up and running, the tier levels will most likely require tweaking or perhaps even a major overhaul.
- If it's in the OLA, measure it. One of the biggest mistakes organizations make is to agree to OLAs and defined service levels, but fail to measure compliance with the agreements. Storage groups should not only detect events that breach the OLA, they should develop reports that analyze trends and try to detect problems before they become OLA violations.
- Formalize customer feedback. You may meet OLAs, but still fall short of expectations. The best way to ensure that this doesn't happen is to sample your key customers regularly to look for ways to improve the OLA model or the services offered.
- Connect OLAs to the procurement process. Organizations often keep vendors and their own purchasing department at arm's length when deploying service tiers. This is a mistake. These two groups can greatly enhance the success of your rollout or imperil the program's future by not backing your choice of tier levels or making it more difficult to purchase the products you need to support the various backup-tier service levels. It's therefore best to include them in the tiering scheme from the beginning.
- Service tiers for backup aren't a fad. If implemented wisely, different backup-tier levels will save your organization money and better protect data. As the lines of business become more involved in how data and applications are protected, the storage group will become more visible and recognized for the important service it delivers.
- Taming Hadoop: Storage Tiering for Big Data –Western Digital
- The Emergence of a New, Highly Flash-Optimized Tier 1 Storage Platform –Pure Storage