Quantum's DXi Deduplicating Portfolio Delivers Needed Choices for Corporate Backup
In a previous blog entry, I took a look at some of the challenges created by deduplicating and replicating backup data from remote offices to a home office and then restoring it again at the remote office. In this blog entry, I'll examine how the architecture of Quantum's DXi backup portfolio addresses these specific challenges by offering greater flexibility with more deduplication choices.
One of the more agonizing decisions that some companies face when looking to implement the same deduplication scheme across the enterprise is quantifying which version of deduplication to use: inline or post-processing. From a purist's viewpoint, inline (deduplicating data as it is ingested) is sometimes viewed as the best approach since data is deduplicated immediately as it is ingested. However scaling an inline deduplication architecture to satisfy enterprise ingest throughput levels (500 Mbps or greater) during peak backup times becomes problematic at best. In environments where these high ingest levels exist, post-processing (deduplicating the data after it is stored to disk) is generally viewed as the best approach.
Quantum's implementation of deduplication across its DXi products takes a unique approach in this respect. Rather than placing its DXi family firmly in either the "inline" or "post-processing" camps, the DXi3500/5500 appliances use Quantum's adaptive approach to deduplicate data as the data is ingested while the new DXi7500 offers a choice of deduplication approaches based on user policies. This strategy allows users to match specific deduplication approached with different backup jobs to maximize both performance and disk resource utilization.
Let's first take a high level look at how Quantum's DXi3500 and DXi5500 appliances deduplicate data using its "Adaptive" deduplication:
- As data is ingested, the data stream is written in segments onto a disk buffer
- As this buffer is filled (default is 256 MBs), the DXi begins immediate deduplication and replication
- If ingest speeds increase, it can dynamically increase the size of the buffer to use more available disk space and avoid backup throttling
The Adaptive deduplication process is available on all of Quantum's DXi products but, depending on which model a customer is using, the choices that a customer has available to it differ. For instance, the DXi3500 and DXi5500 appliances only offer adaptive deduplication and deduplication is always enabled whereas using the DXi7500 system companies can configure deduplication as "Adaptive," "Fully Deferred" or simply turn deduplication off based on each backup job's requirements.
Quantum gave the DXi7500 these additional options since it is intended for deployment in enterprise environments that may need to satisfy short backup windows; backups that contain a high percentage of new data where deduplication is less effective; the requirement for the fast creation of new tapes exists; or when rapid restores are needed.
However that is only part of the story. Each of Quantum's DXi products also supports the creation of logical partitions. The DXi3500 can support up to 8 partitions; the DXi5500 can support up to 16 partitions and the DXi7500 can support up to 64 partitions. Each partitions companies can configure as either NAS or as a virtual tape library (VTL).
Using this option, remote offices can configure partitions for their specific backup environment. If attaching a partition to a LAN, the remote office can configure it as a NAS to facilitate LAN-based backups. However if a remote office happens to have an iSCSI SAN in place, companies can configure another partition as a VTL for improved backup performance. The good news is that regardless of how the partition is configured or how many partitions exist, all of the data is globally deduplicated across all of the partitions on the DXi. (This is only the first step in Quantum's end-to-end de-duplication and replication strategy which I'll dissect further in an upcoming blog.)
The DXi7500 also supports all of these configuration options for partitions plus it offers one more deduplication option - "fully deferred". Using fully deferred deduplication on specific partitions, companies can deploy the DXi7500 in high performance backup environments since companies can direct backups to the "fully deferred" or post-processing partition. These different deduplication options on the DXi7500 coupled with the availability of partitions on all of its models give companies a number of powerful new approaches to deduplicate and replicate data across the enterprise.
In a future blog entry, I'll examine more deeply how the DXi family achieves a form of global deduplication and what options they offer to replicate deduplicated data between appliances located at remote offices, central data centers and disaster recovery sites.