Originating Author: David Floyer with contributions from: Dave Vellante
Contents |
Introduction
Pat Gelsinger has been appointed the new CEO of VMware in a management swap that sends Paul Maritz to EMC as the head of strategy. This move signals a new era in enterprise computing, where VMware becomes the center of the enterprise universe. Maritz put VMware on a path to build the "software mainframe", and Gelsinger is being appointed to take that vision to the next level. The vision of a software mainframe is highly ambitious, with integrated systems capabilities and performance, availability, and security attributes that can handle any and all applications. Those under 40 might prefer the analogy of the cloud operating system, which implies better economics and greater flexibility than traditional mainframes. Owning the cloud OS for the coming generation of platforms is the mother of all value creation plays, and VMware is clearly in the driver's seat.
Gelsinger has indicated there's a need for more storage integration with VMware, and he has observed that most active data will be held in flash and predicted that the days of high-speed disk drives are numbered. Last year, Wikibon discussed the VMware storage innovation ideas at VMware 2011. The basic idea is to create a storage container including the data, the metadata, and the policy data. Logical channels connect the storage to the external storage ports (think the IBM System z multiple logical channel subsystem, SMS and VTOCs, plus object storage). VMware indicated a three-to-four-year roadmap to this world of storage without LUNs.
The key questions are: Where will Gelsinger take VMware? What strategy(ies) could Gelsinger deploy to integrate VMware, storage, and flash to improve the performance and reach of VMware and benefit storage customers generally and EMC storage products specifically? What does this move mean to the hardware ecosystem around VMware?
IO-Centric Architecture
Introducing flash into the storage hierarchy has its challenges. The flash-cache approach is useful for some historical systems but suffers from the occasional very long waits for IO on disk; IO latency variance goes through the roof. For high-performance database systems, the cumbersome SCSI stack is an anchor round the neck. The idea that the management stack could be managed from an array-based tiered storage manager such as EMC FAST is pure illusion – it has been shown time after time that a slow peripheral cannot manage a fast processor. In addition, we believe that flash will be deployed at every level of the storage stack. Some solutions will be deployed sooner, while others will take time to evolve. Those solutions that don't disrupt existing IT processes and procedures (e.g. requiring guest OS agents) will be implemented first and deliver immediate value.
Wikibon has discussed future architectures in Real-time IO Centric Processing for Big Data and has concluded that a top-down architecture is required that will process transactional and analytical applications in real-time.
Figure 1 shows such a layout.
VMware Potential Strategies
VMware is on a path to deliver core infrastructure, management, middleware, and applications. It essentially is taking a Microsoft-like strategy - applied to 2012 and beyond - competing within the entire stack from systems software to applications. Despite it's ownership by EMC, we believe it is in VMware's interest to act as a software player with the goal of making hardware as simple and cheap as possible and grabbing as much function as makes sense and placing that into the Cloud OS discussed above. That said, we believe it will leverage its relationship with hardware vendors (with EMC in the catbird seat). To wit:
One simple strategy for Pat Gelsinger would be to introduce EMC VFCache and VMware kernel APIs to drive a top-down data acceleration and management strategy. This tight integration could provide atomic write capabilities to the flash card and management of metadata and the total storage stack, enabling both transactional and analytic big data applications. More APIs would allow key file and database systems to integrate with VMware and flash and provide a seamless integration to the shared flash-only storage array layer based on the XtremIO acquisition. This goes right at the heart (for example) of Fusion-io's value proposition. Despite Fusion-io's enormous lead, we believe VMware will move down this path. The potential performance of these systems would be in billions of IOs per second, enabling a new generation of applications with far higher data integration. The result would be applications that deliver far greater productivity to users and better utilization of resources. This IO-centric era would also result in higher overall spending on IT.
We believe that VMware must balance its EMC relationship by working with all major hardware players. However, we expect it to use EMC as the reference architecture/sounding board for many of its storage innovations. Over time, we expect VMware to co-opt more storage function, including metadata management and storage management across all arrays. So, for example – speaking to an all-flash array and any disk array or jbod, VMware will own the protocol between the hypervisor, the flash layer and spinning disks. VMware will control the creation of snapshots, H/A, replication -- essentially all storage function -- via the protocol between the file system, the databases, the metadata, the flash arrays and the spinning disk. We believe the direction is clear -- i.e. that control of the stack will come from VMware, not from the array.
This result will be a system that is so efficient, essentially the underpinning of the new data infrastructure, that hardware partners will have to "play ball" or risk being punished in the marketplace.
Conclusions
The constraints to achieving a modern software mainframe vision are significant. The legacy array businesses of EMC will fight it tooth and nail, and management will have genuine discussions of the priorities of pushing VMware downward rather that upward. Cooperation from database vendors such as Oracle and Microsoft will be difficult to achieve. However, Gelsinger has the opportunity to take on these challenges and turn EMC into a full systems company with VMware as the core operating system managing an entire top-down big data stack. As did its predecessor platform companies before it (e.g. IBM and Microsoft), VMware will be calling the shots and capturing the most value in this high stakes game.
Action Item: CIOs and CTOs should expect bold new architectures and ideas from IT vendors to enable the potential of new IO-Centric applications. EMC and VMware could be one company to provide both the vision and delivery of integrated products.
Footnotes:
