Diligent Technologies' ProtecTIER Goes Under the Microscope; My Skepticism Lessens

I have long been skeptical of inline approaches to deduplicating data. My skepticism is not based on whether or not inline deduplication works. Anyone who knows anything about how most of today's deduplicating appliances are configured knows that the majority of these offerings perform inline deduplication. My real concern is "How will these inline deduplication appliances scale when deployed in enterprise environments?"

I have expressed skepticism in the past about Diligent Technologies' ProtecTIER™ in light of the fact its primary go-to-market strategy is the enterprise open systems and mainframe environments. This strategy prompts me to exercise extreme diligence about their technology and architecture before endorsing it. The reason? The speeds and feeds that ProtecTIER is likely to encounter in enterprise shops are unlike what inline deduplication appliances will experience in small and midsize businesses and enterprises.

To help convince me it does perform in enterprise environments, Diligent Technologies was courteous enough to set up an interview with one of its customers in the UK. The interview was set up on the condition that I could not disclose the name of the administrator or company due to the customer's corporate PR policies

This specific customer's environment is what I consider a midsize to large enterprise configuration. ProtecTIER was initially deployed about 12 months ago in an environment with about 110 host servers that it needed to protect. The ProtecTIER software was installed on two Dell 2950 servers with dual quad-core CPUs and 16 GBs of RAM. These servers used multiple 4 Gb FC connections to back end EMC CLARiiON storage systems that had 300 GB FC disk drives and about 7 TB of usable capacity. This same configuration was deployed at two sites (for offsite data protection).

Since then, these 110 host servers have experienced about a 40% rate in native data growth and another 80 - 90 host servers were added to the backup pool. This growth in host servers and data required the addition of another 2 TB of useable capacity to the EMC CLARiiON at each site.

For the sake of brevity, there were two main points I learned about ProtecTIER and its architecture from this administrator that I think are of interest to companies considering ProtecTIER for their enterprise backup environments.

First, companies need to be smart in how they deploy ProtecTIER. Though this administrator was sending over a hundred backup jobs to ProtecTIER, the backup jobs he was streaming to ProtecTIER were those that were too slow and caused shoe shining if they were sent to tape. Using ProtecTIER he consolidated and expedited these backup jobs by backing them up to disk. However the administrator did concede that if he attempted to add his most performance intensive backup jobs to his current ProtecTIER configuration, he is not sure it could handle their backup data flows in its current configuration (this admission by the user goes back to my scaling concerns about ProtecTIER).

Second, ProtecTIER's ability to scale is better than I expected. Some inline deduplicating architectures are not upgradeable so when a new appliance is brought in, it needs to start the whole deduplication process all over again. ProtecTIER overcomes this by decoupling the deduplicating index from the server hardware. In this way, companies can continue to bring in off-the-shelf hardware as they need it for better performance without needing to start the deduplication process all over again.

The conversation with this customer helped to remove some of my skepticism about the use of Diligent Technologies ProtecTIER in enterprise shops. However for enterprises that are in the position where they want to backup everything to disk and not just their slower performing applications, it looks like it might be worth waiting until this summer. At that time, Diligent Technologies plans to make a clustering solution available that may take it the final mile to address my concerns about its ability to scale performance and capacity in high performance enterprise shops.