It seems counterintuitive to suggest that the virtualization market hasn’t already crossed the chasm1. After all, over the last three years we’ve seen a stellar VMware IPO, Citrix’s heady acquisition of XenSource, and Microsoft’s much celebrated entrance into the virtualization market. VMware’s Q3 revenue was $490 million.
What’s not to like? After all, virtualization has been a bright beacon of hope in an otherwise lackluster IT era. It is perhaps the most powerful enabler of true cloud computing, enabling new scales of efficiency and elasticity. Virtualization has been “all that” and more.
And Gartner recently predicted more problems for data centers in 2010. Couldn’t virtualization help to address these problems? If it does then it could become what Dr. Moore calls a de facto standard. We’re certainly not there yet, despite the size of the virtualization market.
Curing VMotion Sickness
Readers of this column won’t need to be taken yet again down the virtualization-lite path from 2008. Instead let’s take things further to their logical conclusion and discuss what happens when virtual networks (VLANS) eventually get over packed with VMs and start to create a kind of “VMotion sickness” condition.
Before you chuckle, think about the core issue with VLAN-centric virtualization: over time the network becomes increasingly out of touch with the assets attached to it. Movement within VLANs is a nonevent because the virtual networks are each (usually) managed as a single entity. While system movement takes place within VLANs, the static network is essentially unchanged.
Motion sickness from Wikipedia:
When feeling motion but not seeing it (for example, in a ship with no windows), the inner ear transmits to the brain that it senses motion, but the eyes tell the brain that everything is still. As a result of the disconcordance, the brain will come to the conclusion that one of them is hallucinating and further conclude that the hallucination is due to poison ingestion.
Most physical networks have limited visibility inside a VLAN. When virtualization evolved into the production data center the network teams were often the last to know; the initial business case for virtualization was so compelling that it created a kind of velvet revolution within IT which allowed operations teams to automate systems at unprecedented levels while network teams were shackled by manual configurations, committees and manually-updated spreadsheets.
Increased density and movement will create problems as it crosses a threshold and network connectivity is ultimately degraded. The VLANs that enable the capex revolution then become an impediment to flexibility and management because they create isolated islands of automation. Until those islands are connected, the virtualization business case is contained within increasingly dense and complex VLANs.
The IT Locomotive now Needs New Track
Locomotives created the need for railroads. Just as the locomotive that preceded the “golden spike”, virtualization creates a compelling case for a more evolved network. Without the transcontinental railroad and other linkages, the business case for the steam engine was limited to incremental gains over manual/horse transportation. The greater the distance the higher the value of the locomotive (versus animal/human-assisted transport).
Today we’re in the period where virtualization has demonstrated the potential of automation/movement, but the network vendors haven’t quite figured out how to fully enable that potential. The result is a kind of dissonance between moving systems and static networks and the inability for IT to bridge these incompatible worlds.
Thus the term “VMotion sickness” seems as apt as any to describe a condition more teams will eventually face as more locomotives (VMs) are packed onto disconnected tracks (VLANs). And here’s the punch line: if this issue is addressed/solved by the physical or virtual network the genie is out of the bottle and the IT industry will be permanently transformed.
That is why I think it will be solved within the next two years and that the solution will enable a new era of IT investments and capabilities. That new era will be marked by reduced power and operating costs and increased scale and transport capabilities per application/user/system. This theme is the core of the infrastructure 2.0 blog.
Public Companies who Could Benefit
If you think of IT as being strategic to business, then any strategic disruption could have significant effects across industries, economies and even energy consumption demands. It’s a vendors dream to be designed into the new standard.
From a virtualization vendor standpoint (I think that) VMware is still furthest along in collaborating with networking vendors, and the recently announced Acadia venture with Cisco and EMC seems to validate this viewpoint. This also speaks well of Cisco with its new servers and first generation UCS. VMware’s leadership and relationship with Cisco best position Acadia as a potential standard. Yet there are plenty of issues that will need to be resolved before victory is assured.
Application delivery leader F5 Networks is strategically placed near the middle of this issue, as their solutions impact the transport and load balances for networks. They are also active in the stealth infrastructure 2.0 working group and the infrastructure 2.0 blog.
The HP, IBM, Juniper coalition could also make for some interesting potentials, if the elephants can dance together with their own unified and branded stacks and containers. HP’s recent 3Com acquisition introduces interesting potentials to the herd as well. Juniper’s new tagline is, coincidentally, “The New Network”. Other interestingly positioned players include Citrix and some of the progressive service providers.
Why the Chasm will be Crossed
The first branded container/stack to address the VMotion containment issue and orchestrate stateful, elastic VMotion across (VLANs) today’s isolated islands (including great geographical distances) will have a strategic first-mover advantage in the race to virtualize the data center as virtualization will then set the standard. Until then expect continued incremental improvements and market positioning exercises.
1For those unfamiliar with Geoffrey Moore’s Crossing the Chasm, Moore argues that there is a large gap between the product requirements of visionary (early adopters) and pragmatic (early majority) customers. If a solution can cross this chasm it can create enough momentum to establish itself as a de facto standard in the market.