The announcement by Cisco/EMC joint venture Acadia offering Vblocks—a closed wolf's stack in open sheep's clothing—is a computing delivery model that creates a feeling of déjà vu. Where has this model been tried before?
Oh yes. Proprietary systems such as mainframes from Unisys and IBM have used this model very successfully, especially in high volume transaction environments, an area VMware covets. Wikibon has been talking with senior people about the comparative costs and benefits of different delivery models and it is instructive to review those in light of the Vblock announcement.
The major advantages of the closed stack computing delivery model (or “mainframe” as it is traditionally called) is that all the parts are written, integrated and tested together. The most pure form of this model today is the Unisys mainframe. The advantages that Wikibon found for this model are that the costs of delivery are lower, even though the cost/MIP is higher. The reasons for this are lower operational costs, lower cost of providing high-availability systems, and reduction in the number of components that have to be integrated and managed.
The second major advantage of the integrated approach is that time to change is faster. Applications written on this platform can be changed and the new code tested twice as fast compared with an open systems approach. The reason is again that the integrated stack means less pieces to change, less testing and easier introduction of new versions of application software.
The disadvantages of this approach are the initial higher cost of hardware and the lack of choice in components in the stack. Unisys offers the DMS-II database, an excellent system but without some of the development features of (say) Oracle. Vblocks offer Ionix as the management framework, where other organizations may base their systems management on BMC, HP, or IBM management suites.
Comparing the V-Block stack against the Unisys Mainframe stack is most illuminating. The Unisys stack is a complete delivery mechanism including database, transaction computing, communications, and systems management as well as operational automation. The V-blocks provide the hardware, virtualization layer, and some systems management, but do not provide an application layer.
Wikibon found that mainframe integrated stacks are most successful and cost effective where high complex transactional systems have very high availability requirements, and where the rate of change within the transactional system is high. The lesson of history is that to be successful with Vblocks in the high-end, VMware, Cisco, and EMC will need to provide a more complete and robust stack. Will VCE be able to get Microsoft and Oracle to join in their stack and provide the software infrastructure, or will these companies build their own ecosystems?
Early indications are that Oracle especially is shunning VMware, indicating its customers don't need it. With Virtual Iron and Sun, Oracle will build its own containers for database applications. Microsoft will push Hyper-V and only reluctantly support VMware leaving the market fragmented.
Action Item: Users should review and monitor the development of integrated computing and infrastructure stacks. However, it is too early for users to be considering Vblocks for applications that are currently on mainframes. These stacks will need to significant extentions in scope and functionality before they can be considered as serious contenders.
Footnotes:
