Enterprise data centers environments are continuously faced with ever growing volumes of backup data. Many organizations have embraced purposed built backup appliances (PBBAs) utilizing deduplication (dedupe) to help drive operational efficiencies and reduce reliance on tape as a primary backup medium. While dedupe is an essential element of next-generation backup technology, many large IT environments need solutions which can also efficiently scale within a single system to support high performance and large storage densities for current and future backup requirements.
Despite their advancements, the challenge is that most PBBAs are limited in terms of how much storage capacity and throughput can be supported within a single system image. In these types of “scale up” storage architectures, once the frame is filled to capacity with disk, end users either need to upgrade to a higher-end controller unit or add a second frame to meet backup growth demands – potentially resulting in multiple islands of backup infrastructure.
Hopping on the Grid
In addition to management complexity and reduced dedupe efficiency, a “siloed” backup strategy can be a costly proposition. This is because when purchasing PBBAs, IT buyers need to invest in the full processing and memory resources of the appliance controller “head” up front, regardless of the actual storage capacity required for the backup workload. Furthermore, the processing and memory resources can’t be shared or load balanced across heads.
On the other hand, “scale out” backup systems like Sepaton’s DeltaScale grid architecture can be configured with as few as two nodes and 24 TBs of storage capacity and scale to 8 nodes and 2 PBs of usable storage in a single system. Sepaton’s grid architecture is more like a “just in time” approach for provisioning backup resources. End users can add nodes as additional processing and/or storage capacity as required and avoid overbuying processing and memory upfront.
With the recently announced version 7.0, a fully loaded, 8-node Sepaton S2100-ES3 appliance can deliver an impressive 80TB/hr of backup throughput and provide multiple PBs of deduplicated backup storage capacity in a single 42U rack enclosure. This provides plenty of performance and storage scalability for large data center environments while reducing data center footprint and operational costs.
Time to get off line?
Data deduplication can be performed, “inline”, meaning it runs before the data is written to disk, or it can run “post process” as a secondary task once backups are complete. Sepaton offers a third option, concurrent deduplication, which starts the deduplication process as soon as the first backup job completes and continues concurrently with the rest of the backup.
When backup speed is of paramount importance, inline dedupe can be a limiting factor. The overhead of dedupe operations inserts latency into the backup process, thereby reducing the effective transfer rate of backup data to storage. Likewise, when data is being recovered it must first go through a “re-hydration” process back to its native format before it can be presented to the requesting host or application entity.
Comparatively, in a post-process implementation, dedupe takes place after the backups are completed. Sepaton’s concurrent deduplication process delivers the best of both worlds, allowing for a much shorter backup window than inline or post-process dedupe. Additionally, concurrent systems like Sepaton’s, reserve a copy of the previous night’s backup in native format (non-deduped), thus avoiding the added latency of re-hydrating data during the recovery process.
Interestingly, many users adopted inline dedupe, despite the overhead it placed on backup throughput, mainly because it enabled the fast off-siting of data for DR purposes In fact, some implementations of post process deduplication can take 2-3x the amount of time required to electronically vault backup data compared to similarly configured inline platforms.
Sepaton’s high performance architecture, however, greatly enhances time to safety by providing for backup throughput rates that are in some cases, nearly 3x as fast as top end in-line dedupe systems. As a result, concurrent dedupe and replication jobs can complete faster than their inline dedupe counterparts.
Database Dedupe Efficiency
Another important attribute of Sepaton’s dedupe is its ability to support backups of very large databases by performing data reduction across multiple backup streams and across multiplexed volumes. It provides deduplication down to the byte level without slowing ingest performance. In contrast, some dedupe systems can’t analyze data in segments smaller than 8KB, which is not granular enough to capture the small changes that occur in database systems like Oracle, SQL or DB2.
By peering into data at the byte level, Sepaton can provide significant backup efficiencies, which may translate into big savings for data centers running large databases and other data intensive applications.
The need to secure sensitive business data and to meet compliance regulations is driving organizations to implement encryption solutions for data across the enterprise. Data stored on disk or tape backup repositories (referred to as “data at rest”) requires strong levels of encryption like AES256 (Advanced Encryption Standard). The new Sepaton S2100-ES3 Series 2925 backup appliance delivers AES256 encryption for data at rest without a performance penalty while requiring minimal system overhead.
Users can encrypt data across the entire system. Encryption keys can be managed through third party KMIP-compliant key management frameworks, like the RSA DPM 3.x and Thales e-Security keyAuthority 4.0 appliances, enabling end users to avoid key manager vendor lock in.
Storage Swiss Take
Various industry sources estimate that data is doubling approximately every two years.
With big data analytics and cloud computing initiatives on the rise, the massive proliferation of data will continue. While data deduplication has improved backup efficiencies, its deployment in monolithic, scale-up storage appliances hasn’t provided the storage density and performance that’s needed to stay ahead of the data growth curve. In fact, scale-up deduplication appliances can even make backup storage less efficient since storage, memory and processing resources cannot be easily shared across backup silos.
Sepaton’s S2100 modular backup platform leverages high performance grid computing in conjunction with data deduplication to address this problem. Its scale-out architecture helps data center managers increase capacity and performance to stay ahead of backup growth. Furthermore, it allows them to deploy backup resources as they are needed, leading to improved efficiency and enhanced investment protection.
With its new data-at-rest encryption capability, the S2100-ES3 also enables enterprise data centers and managed service providers to meet compliance requirements, safeguard sensitive business data and elevate backup quality of service (QoS) while lowering operational costs.
Sepaton is a client of Storage Switzerland