Let me start out by saying that I see more discussions about greenfield deployments of products than I do of migration/integration by vendors. Personally, I think there is no such thing as a greenfield deployment unless the organization is just starting out, creating a brand new data center, or perhaps has money to waste. In most cases, what is defined as greenfield is really just a grain of sand on an island of technology that still needs to integrate into the greater organization. As such, that integration should be the foremost thought when products are developed. But instead, it is not, and effort goes into becoming that island or into a replacement install that is still an island.
Let us look at The Virtualization Practiceâ€™s environment. We use a part of a 100% virtualized environment that extends into the cloud. As such, anything that is placed into that environment must work within the virtual environment or the existing cloud environment, whether the item placed there is a security device, converged infrastructure, storage, management tool, performance tool, or even a new hypervisor. These must all integrate into an existing deployment. Most other enterprises of all sizes are the same way. They may drop in a Vblock, but it needs to integrate into existing management and performance tools, security, and network devices. Understanding how these tools and devices integrate is crucial to the development of new products.
What is also important is the buying cycle. Most organizations add capability as they need it: they do not replace everything outright. They upgrade hardware in spurts, getting as much as they can afford while still making some use of older hardware, until it is too old to work with the code base they have. This is usually a lengthy process. To get where The Virtualization Practice is today, we went through four upgrade stages to convert many different physical boxes and tools into a virtual environment that extends into the cloud. Part of each step was addressing how to get closer and closer to a 100% virtual environment. Once there, we do not have to worry too much about hardware, but rather we must focus on the applications we need. However, now that we have tools such as VMware NSX and VSAN, we need to consider yet another upgrade process if we want to make use of these new tools. We will need to consider the following:
- Our existing local disk deployments
- Our existing network deployment
- Our existing hypervisors
- Our existing performance monitoring tools
- Our existing management tools.
When we look at all of these, a plan emerges on how to upgrade in an existing deployment. We could, of course, buy absolutely new gear, but then we would be faced with a migration deployment. Migrating from old to new is still not, in my opinion, a greenfield deployment. Yes, the kit may be new, but the software in use will not be. We are expanding our functionality with new gear, not just replacing what is already there. This is the normal method of adding functionality, I have found. We add to our existing environments. Granted, we may be a few seed units or we may update a few units to be those seed units. But seed units are still not greenfield deployments. They are seeding the upgrade or install of new software, hardware, etc. Once more, this is migration/integration.
I was recently at the GPU Technology Conference to discuss NVIDIA GRID technology, and I realized that this is another technology that requires specific hardware (mainly for cooling reasons), bigger power supplies, and other necessary hardware to make these new 3D graphics boards work within an environment. Because of the restrictions, NVIDIA is allowing the OEMs to pick the hardware that is required, blessed, etc., and then is adding in the GRID boards. Even if these are new boxes, they will be brought in either as an island of technology that must integrate with what is already there, or as upgrades to specific hardware that meets the GRID requirements.
Do you think of buying an island of technology? If so, have you asked how it will integrate into the network, storage, and management that already exists within your data center? Will it? Or is this a brand-new deployment that requires you to change your underlying tools? Which approach do you take? In either case, vendors need to start providing either migration instruction or integration instruction; even a little will go a long way. Vendors should also provide reference architectures for existing common deployments and to assist others in integrating with newer technologies. In these cases, a spreadsheet does not cut it. Vendors can reuse existing architectures, but they should tell us how they differ from past products and what hardware is required to make them work.
Do you plan on using VSAN, GPUs, VBlocks, new storage, etc.? Does this fit your existing architecture? Has the vendor provided any help in this arena? Which vendors are better than others?