When it comes to IT, change is constant, and today’s biggest change is the ever-increasing amount of data that enterprises need to back up every day. However, as the volume of data grows, the backup window grows—at a time when business needs require a shorter backup window.
This article looks at what is impacting the performance of backup and recovery, and offers some best practices for ensuring that backup performance scales and doesn’t degrade over time.
Backup Performance Headaches
There have traditionally been two options for data center backup storage: primary disk or deduplication appliances. Although primary disk is fast, it becomes incredibly expensive when more than 10 total copies of retention are needed for all of the daily, weekly, monthly and annual backups.
As for inline deduplication appliances, although they reduce the amount of primary disk required, they also have their drawbacks, including:
- Extremely slow backup. Inline deduplication is a very compute-intensive process that uses up a lot of CPU, memory and time to complete.
- Scalability challenges. Inline deduplication appliances are scale-up, so as the amount of data being backed up grows, the backup window also grows.
- Forklift upgrades. Once the backup window gets too long, the deduplication appliance requires a bigger and faster controller to replace the existing controller. These forklift upgrades are expensive, disruptive and wasteful.
Recovery Performance Headaches
Looking at the restore side of the equation, primary storage once again has a performance advantage but is too expensive to maintain the 20 or so backups that most enterprises retain.
Deduplication applications fare far worse on recovery. Since data is stored only in deduplicated form, it must be “rehydrated” for every recovery request, which dramatically slows restores. As a result, restores can take up to 20 times longer than restores from primary storage. A virtual machine boot from an appliance can take an hour or more, compared with just a few minutes from primary storage devices.
Best of Both Worlds: Tiered Storage
The simple fact is that both backups and restores need to be fast, and a deduplicated repository is required to reduce the storage footprint and the cost of long-term retention for multiple backups.
What is needed is a tiered backup storage approach that has both a front-end disk cache for high speed and a deduplicated repository tier. Why both? There are several reasons:
- There is no single inline deduplication appliance that supports backups at the speed of primary disk. Inline deduplication is slow.
- On a tiered storage device, jobs are written directly to a disk-cache landing zone for backups that are as fast as writing to disk (without the deduplication penalty) and the most recent backups are kept in an un-deduplicated format for restores, which avoids the data rehydration penalty.
- The second retention tier is required to lower the overall costs for all long-term retained backups in a deduplicated data repository.
Tiered storage offers the best of both worlds: deduplication that doesn’t impact backup performance; more efficient recovery by avoiding data rehydration; and a reduction in the overall amount of storage required, yielding lower overall costs.
ExaGrid offers the best of primary disk backup performance and deduplication in a single, integrated system. ExaGrid’s unique scale-out architecture scales all elements—not just disk—to ensure backup windows stay fixed-length as data grows. Best of all, every ExaGrid system ever made has a future-proof design that enables mixing and matching of newer and older models, which can each share the load of backup and deduplication.
The benefits to enterprises are clear: high performance, low costs, no product obsolescence and no forklift upgrades. Also, every ExaGrid customer receives five-year price protection, which translates to predictable costs and peace of mind. For more information on how ExaGrid can solve your most pressing disk backup challenges, please visit ExaGrid and review the articles and resources on this site.