One of the highlights of EMC World 2012 was the presentation by Stephen Manley (CTO, EMC Backup & Recovery Systems - BRS) of a new model for backup. Manley came from NetApp and was involved in development of the pioneering SnapManager and SnapVault products, and he has set a new, profoundly different vision for EMC's Backup and Recovery Group.
According to IDC, EMC's share of the purpose-built backup appliance (PBBA) market has consistently hovered around 66%, an astounding feat considering the number of established competitors in this area. The big question on everyone's mind is can EMC maintain that share. With this new, extended vision, we believe EMC is covering the one flank where alternative approaches could hurt its market share. Specifically, if EMC executes on this vision we believe it can maintain its top leadership position. This research note summarizes the EMC vision and provides our opinion on what it means for customers.
Contents |
The Need for Change
Wikibon has extensively written that today’s model of backup is broken and that data protection as a service is the fix. The concept of one backup process for all data is too expensive and not working any more. For example, most customers perform daily incrementals and weekly fulls as the sole backup service for all data. Some might add an expensive remote replication capability for very-high-value data, but often that's it. Techniques such as using EMC Data Domain data de-duplication have helped ameliorate the problem of backup windows and provide easier recovery of more data. But the pressure of meeting backup windows and of providing a service appropriate to the business value of the application remains.
Data Protection as a Service
The new EMC vision is tied to the increased use of storage-based, application-consistent snapshots of application data and remote replication of just the changed data in the snapshot. The SLA for the application (expressed as RPO and RTO) can be tailored by adjusting how often the snaps are made, the size of the pipe moving the data offsite, the speed at which the remote data can be restored, the time the historic data is kept, and the provision of data hardening technologies for data in flight (e.g., RecoverPoint and Axxana). The level of RPO and RTO required can be dialed up and down for different applications and can be provided as a level of service defined by the business.
This vision was first put forth by Wikibon several years ago and referred to as the "Time Machine for the Enterprise", referring to Apple's technique of taking snapshots over time to perform backups on its personal computers (see Figure 1). A main criticism of this approach is the overhead involved and the CPU cycles required to execute. We believe however that the massive increases in performance from x86 systems, embedded into purpose-built appliances, will easily overcome this objection.
Breaking the Decades-long Backup Software Lock-in
Microsoft has been an important enabler of this capability by providing VSS services to enable application-consistent point-in-time snapshots of data. Oracle and other database vendors provide similar services, and they are becoming ubiquitous across all applications and operating environments, including the all-important hypervisors from VMware and Microsoft Hyper-V.
One important byproduct of this approach is that backup data can be held in its original format and not in the format of the backup software. This will enable much better use of backup data for archiving and move towards a “single truth” for historic data, with enormous savings on storage costs and improvements in data accessibility. In the long run it will mean less tie-in for users to keep backup software, just because the historical data is kept in that format.
Metadata Management is a Key to the Future
Manley pointed out that the key to differentiation in this new model is the multi-site catalog of all snaps and replications made across the data environments and the management services that are provided. Within EMC there is competition to see which product will provide the catalogue. Wikibon suspects that the Data Domain catalogue will win as the starting point but could lead to opportunities for competition if it is tied too closely to the provision of Data Domain products. Tape will need to be included as a low cost option, and the only cost effective way to transfer very large amount of data quickly over long distances.
One of the areas that snapshot techniques need to borrow from traditional backup and tape systems is increased resilience to data corruption, especially long term data corruption. Manley calls this "Data Invulnerability," the systematic addition of checksums, multiple headers and index entries that is strong and (mostly) built-in to traditional backup systems. Disk-based technologies error rates combined with data travelling long distances over telecommunication lines fundamentally designed for voice needs a large dose of invulnerability technologies.
Although the snapshot ideas have been available in other solutions for sometime, this is the first time that a major backup vendor has put forward a complete vision so clearly. Wikibon believes this is an important and far reaching announcement. Assuming that EMC follows through on its promises and executes well (historically EMC has been strong on both counts), this vision will radically change the backup products, processes, and procedures used by enterprises over the next decade. EMC will be in a prime position to leverage its dominance in the backup appliance market (66% revenue share in 2011 according to IDC) into this new model.
Customer Implications
Customers are facing ever-shrinking backup windows. Combined with more demanding RPO and RTO requirements, the backup and restore problem is becoming untenable. Add to this the fact that virtualization is sopping up the free cpu cycles that used to be available for backup, and conventional approaches to data protection will not likely satisfy next generation system and storage architectures. By taking a service-oriented approach to data protection, customers can dial down or up data protection by application and break the one-size-fits-all mentality of most backup processes today. Key issues remain with regard to leveraging (versus replacing) existing backup software, vendor lock-in with de-duplication appliances and de-duplication algorithms that speak different languages. Nonetheless, the vision EMC is putting forth is, we believe, the correct one for customers pursuing an IT-as-a-Service strategy where backup and restore is a fundamental part of the service catalogue that can be applied as needed across the application portfolio.
Action Item: CIOs, CTOs and other senior executives with responsibilities for backup and recovery should follow this EMC strategy closely and should encourage EMC and other vendors to accelerate progress. As the vision matures, organizations should be aggressive in moving to the new model as the best way to meet backup windows cost-effectively, allowing the line of business to decide the RPO and RTO requirements and expenditure by application and improving the ability to leverage historic data. The business case for change is overwhelming.
Footnotes: Video Spotlight on New Models for Backup from EMC World including a segment with Stephen Manley