Enterprises face a battle. How do they address an ever growing data protection and retention requirement while at the same time using less of a shrinking backup window? Five years ago this challenge was met by using purpose-built backup appliances (PBBAs). These products leveraged the speed, reliability and random nature of disk to help meet backup windows and improve restore performance.

The challenge today is that the amount of data has continued to grow, in some cases exponentially. That rate of growth has overwhelmed those first generation disk backup appliances. Enterprises need a new standard in disk based data protection and as a result, need to take a new approach in selecting a PBBA.

In the data center of five years ago, adding almost any vendor’s disk backup appliance would improve performance, even if not to the maximum potential. Since that time the rest of the environment has increased its ability to transfer data. Servers are more powerful and can identify data to be protected and move it to an interface more quickly than ever. Networks are faster and can leverage multiple trunks to transfer that data to a backup server faster. Backup servers are also more powerful and backup software can have multiple media servers simultaneously transferring data to the backup target. Software, in the case of Symantec  (Veritas) NetBackup, has even created more efficient network protocols to make this data transferral more efficient.

The net result is that the capabilities of the DBBA matter more than ever. Its ability to ingest data, optimize that data and to scale to meet upcoming growth needs are now critical determinants in product selection. The goal for the enterprise is to select a near permanent solution that is cost effective today but can scale to accommodate future requirements. To meet these demands and goals the IT professionals involved with the purchase process must be more exacting in their evaluation.

The Art of Discovery

Potentially the most important step is to know the state of the environment right now. It’s essential to have an understanding of actual data volumes, operating systems and hardware that are in use. While to some extent this seems obvious it’s often overlooked or done ad-hoc. A hands-on inventory of the environment almost always uncovers details that come as a surprise to the IT department in the form of unprotected servers, unprotected volumes or unexpected capacity.

This discovery can be done in several ways. In some cases the prospective PBBA vendors may offer to provide a detailed inventory at a reasonable rate. After all they need this information to size their hardware recommendations. There is also the option of the IT staff manually inspecting each server and capturing the information in a spreadsheet. At Storage Switzerland we recommend using a third party tool that can automatically discover the servers, clients and storage systems in the environment. This provides an automated and objective third party view of the environment and is something that can be easily updated as needed.

Simulate Production

Most evaluation scenarios are no where near representative of the real-world environment. Instead they’re often small amounts of data, more akin to samples than actual production. This results in what’s often called a “beauty contest”, more of a demo scenario that focuses on GUI and features instead of a performance evaluation. While ease of use is very important, as is the quality of the user interface, a PBBA is typically bought to solve a specific problem. Most often that problem is meeting backup and recovery windows in given timeframes. This means the performance capabilities of the devices need to be well understood.

The evaluation scenario should use high file counts and files sized to match the current environment. Backing up 1TB of 100MB files is an entirely different problem then backing up 1TB of 100GB files. Also, it’s important to conduct backups as they would be performed in production. For example don't run a full every night if that is not what’s done in practice. All this will do is artificially inflate the system’s deduplication effectiveness and may inaccurately report backup performance.

The goal should be to push the PBBA to its limits so that those limits can be understood. A best case scenario would be to bring the PBBA into production during the evaluation timeframe. This doesn’t mean putting data at risk and cutting over to the new system but instead running the old environment and the potential new environment side by side. There is no better indicator of production data than actual production data.

Stress The System

As stated earlier stress the PBBA to its absolute maximum. Make sure this load is done with the features enabled that are going to be counted on by the organization. Measure performance with deduplication, replication and other capabilities fully engaged. This will clarify how far into the future it will have the performance to meet the organization’s needs and how well each tested PBBA will perform under similar conditions. There is a balance that most organizations need to reach between cost and performance.

Stressing the system will also provide the opportunity to test "what's next". Once maximum performance is met ask the vendor to provide a solution to overcome that limit. Examine how the vendor will get you over the next hurdle. Will there be a performance ‘wall’ with the existing solution that forces you to rely on the release of a higher performing unit before your projected requirements are met? Will that solution require a complete change-out to the next system in the line? Or, as is the case with SEPATON's scale out capability, will getting to this next step merely be a matter of acquiring an additional node that provides performance, bandwidth and scalability to the overall PBBA?

Knowing how the vendor will overcome the performance wall, and how much it will cost,  are potentially the most critical aspects of right-sizing the data protection purchase. The one constant in enterprise storage is that data will grow and backup windows will not. Upgrades are going to be a fact of life. Understanding if each upgrade must be done all at once, with a forklift, or incrementally, with additional nodes, will prevent future budget surprises.

Look For and Evaluate Enterprise Features

The enterprise needs more than a disk at the end of a wire that can deduplicate data and replicate that data to another location. New capabilities exist that enterprises need and should demand. The first of these is multi-tenancy, obviously important for storage and compute providers but also for more traditional organizations that need to pool and segregate data. For example, there may be a need to keep separate data from different backup applications or business units. This should be done without requiring a separate PBBA.

Another enterprise feature to be considered may be selectively turning deduplication on and off, by a given set of data. If there is a data set that won’t deduplicate or one for which deduplication may violate a compliance concern, this feature can be a detriment. The goal is to meet the needs of the organization without increasing management complexity, which in some situations means excluding a ”cool” feature.

Another important ability that can impact compliance is to quickly and securely erase data. For example, some systems even though a delete command is given, will not actually remove the data until the next maintenance cycle, which could be a week after the files were marked for deletion. For specific data sets this could be a violation of corporate governance and leave the organization open to litigation.

Finally, it is important that the Enterprise PBBA use deduplication to monitor status, efficiency of backup and effectiveness of deduplication. Some PBBAs are surprisingly lacking in this regard and instead point the backup administrator to the backup application for answers. The problem is that PBBAs by their very nature, will typically consolidate multiple backup processes. Since the backup applications obviously don’t know about each other they can’t provide the comprehensive status information the user needs. Only the PBBA, as the consolidator of all these backup jobs, can provide an accurate reporting of what is going on.

Summary

Many reports have surfaced over the last few years about an expected turnover in the enterprise backup market. The assumption is that much of this turnover will consist of customers replacing legacy backup applications with other, more modern backup applications.

Storage Switzerland predicts that the turnover will not come in the backup application market, but instead will involve the backup target devices. Vendors like SEPATON are well positioned to support the needs of this market as the enterprise data center looks to either augment tape and/or upgrade from an early, non-enterprise disk backup investment. Differentiating between enterprise PBBAs and mid-market PBBAs will require evaluating systems in real production environments, not just in production simulations.

For more details on right sizing your next data protection purchase please download Sepaton’s new white paper: “Ten Steps to Right Size Your Next Data Protection Purchase”

Sepaton is a client of Storage Switzerland