I spent this week at Dell Storage Forum in Boston. Dell's announcement of the EqualLogic Blade Array is generating a lot of interest, and I think it signals the introduction of the second wave of converged solutions. There are a few vendors riding this second wave, and it may leave current converged solutions all wet.
The goal of converged infrastructure solutions is to provide lower time to first Virtual Machine (VM), reduce ongoing operational costs, and reduce the infrastructure footprint by providing a fully integrated virtualization infrastructure. These converged solutions typically include multiple compute, network, and storage in a single system. The first wave of converged solutions comprised primarily off-the-shelf solutions that were pre-implemented prior to arriving at your site. In many cases, all three components came from three separate vendors. There was not much customer innovation or true integration in the first wave of converged systems.
However, the first-wave systems did, for the most part, achieve the goal of lowering time to first VM. They might have even reduced operational costs if the vendor provided some additional management software. But most systems failed to reduce the footprint of the infrastructure. The footprint should be judged not only by how many physical rack units the system will consume, but also how many virtual machines can be supported per rack unit.
[ Learn more about how eliminating data redundancy can pay off. Read The Value Of Data Deduplication. ]
As I discussed in my recent article "Rack Density--The Key to Maximum Virtualization ROI", the number of virtual machines you can support per rack unit will become a key measurement in determining the long-term ROI of a virtualization project. Achieving a high VM per rack unity density will require more than just prepacking disparate hardware solutions; it will require innovation that rethinks how compute, network, and storage are integrated so that they provide maximum performance with minimal space utilization.
To achieve this will require the intelligent integration of solid state storage into the environment so that the I/O blender of server virtualization can be easily handled. It will also need to be able to leverage flash as a virtual swap area since the use of DRAM will need to be minimized due to space and cost constraints.
That means you must shift from thinking of these systems as preconfigured systems, in which the initial engineering work has been done before it gets to your site, to systems that are engineered specifically to reduce time to first VM, reduce operational costs, and reduce data center floor space consumption. Further, the systems must do all this while maximizing VM density per rack unit. This will increase the ROI of the virtualized investment and give the data center a fighting chance of keeping up with your organizations' rapidly growing IT demands.
Follow Storage Switzerland on Twitter
From thin provisioning to replication to federation, virtualization options let you reclaim idle disks, speed recovery, and avoid lock-in. Get the new, all-digital Storage Virtualization Guide issue of Network Computing. (Free registration required.)