Virtualization and cloud computing are promising to change the way in which IT services are delivered and, in effect, transform computing as we know it today. I think the promises are likely to come true, if and only if critical technology issues are addressed.
Nicholas Carr told a recent audience at IDC Directions that “Cloud computing has become the center of investment and innovation.” While he is not a technologist, his sometimes shocking insight into the transformation of IT have been prescient, even if he doesn’t sweat the details of how complex IT infrastructures can morph into the equivalent of today’s public utilities.
To his credit Carr has predicted the rise of the cloud computing press release, multiple cloud conferences and panels and even the SaaS repositioning exercise. He also foresaw the rise in Amazon and Google cloud announcements, perhaps years ahead of profits and/or material revenue.
Here is an example of a billion dollar repositioning exercise as covered in CIO Magazine’s online blog:
In late February 2009, Salesforce.com CEO Marc Benioff proclaimed that the software-as-a-service CRM maker was “proud to be the first billion-dollar cloud computing company.” – Thomas Wailgum, CIO Magazine March 9 2009
Along with his vision it would be refreshing for Carr to address some of the technological and microeconomic issues that will need to be resolved for his vision to take hold. And by take hold, I’m talking about viable models of cloud computing as he describes them; not software as a service or a rack of hypervisors on a private VLAN or a management tool that enables some limited pooling or processors for a single application. IT as a service is much bigger than these early accomplishments.
The cloud computing models that Carr describes will require fundamental breakthroughs in the way networks, applications and endpoints interoperate. It will require unprecedented automation inside the network as well as an unprecedented scaling of network capacity. I wrote earlier about The Three Horsemen of the Network Revolution (virtualization, cloud and netbook computers); the common theme was the rising tide of demands on the network and the opportunities it would create for savvy network vendors.
After all, Carr’s cloud will require a robust network, or dynamic network infrastructure, something he doesn’t cite as significant. Yet the network is more than just a mesh of cables; application delivery is more than a power switch and applications have more complex and ongoing interactions with endpoints than the most complex power grids.
I think it is this disconnect between a systems-centric view of IT and the reality of the state of the enterprise network today that has fueled the rampant, hyperbolic speculation about the ascent of cloud computing and the resulting rush to be relevant. So let’s talk a little bit about at least one element of what Carr is missing and the implications for CIOs.
The VMotion Paradox
VMotion is the ability of a virtual machine image on a hypervisor to move to another hypervisor without having to swap cables and/or physically move a server. This capability was especially powerful in development and test environments, where images of software could be easily saved, updated, moved online and/or offline or even across multiple distinct hypervisors with minimal effort.
As VMotion moves from DevTest to production environments it enables new levels of consolidation and utilization. Servers and desktops can be spun up as needed and moved with minimal effort. This automation of once manual tasks produces impressive tactical breakthroughs in power consumption and system automation, and is one of the key drivers of virtualization in production data centers today.
In a heterogeneous data center environment (and especially a cloud of data centers) that level of movement is a double-edged sword. If VMs could easily move from one security zone to another there would be an even larger, strategic payoff; yet today’s manually managed network isn’t ready to support such powerful innovation.
Networks simply aren’t equipped with the connectivity intelligence required to keep up with dynamic systems and endpoints. Many enterprises are already feeling the pain of rising (per unit) IT diseconomies. Adding higher velocities of change would drive costs even higher, offsetting any system automation and/or energy gains. Automating systems attached to manually managed networks is, after all, the CIO Shell Game.
That’s part of the reason analysts and pundits will be paying greater attention to cloud revenue and margins relative to Google and Amazon and other “cloud” players. Whoever cracks the VMotion code (of security, scale and network automation within a heterogeneous IT environment) first will have considerable advantages.
Microsoft’s knowledge of enterprise requirements and applications put the company in a unique position if it doesn’t get bogged down in “installed base” application and OS dilemmas. Yet it isn’t a networking player. There is a chasm between the endpoint and the system that the software vendors will have to cross.
This gets us back to Cisco and Juniper in the once irrelevant cloud network space. Both seemed intent on making strategic partnerships a core part of their strategy (versus Microsoft’s frontal assault on established leader VMware). Cisco has been faster to embrace Infrastructure 2.0 (or dynamic infrastructure) and Juniper has recently aligned with IBM, who recently held its own Dynamic Infrastructure forum.
The decision by Cisco to invest in VMware and partner I think will be a force multiplier in the cloud relevance and delivery battle. If Cisco can tackle the network management and automation issues inherent with VMotion it could unleash incredible cost and energy savings (which could drive a new business case for the network upgrades required) and deliver on Carr’s promise; as Carr continues to pamper Google and Amazon with incredible, unproven visions.
F5 and others are also strategically positioned when it comes to application delivery requirements. Rather than sheer customer footprint, those who embrace automation and intelligence in the network will have a (yet misunderstood) strategic advantage when it comes to monetizing and optimizing on Carr’s vision. While Google, Amazon and even Salesforce position themselves as cloud players I think it will be the network vendors who end up selling the picks and shovels to the miners.
Scaling IT will require Unprecedented Automation
Cisco is strategically placed to deliver other breakthroughs in network and system management via a renewed emphasis on “unified computing” partnerships. I think that will ultimately include a range of strategic players on the team who would ultimately join to be the “cloud standard”. I think that standard will likely have to include dynamic network intelligence regarding endpoints, switches, core network services like DNS, DHCP and IP address management. After all, cloud CIOs would need real-time views into all IT assets.
That standard will be a competitive advantage if Cisco pulls it off.
Cisco would also need dynamic policies that would follow the movement of assets instead of being decoupled by VMotion. They would need networks capable of transporting large amounts of data in bursts across the mesh, while delivering on the promise of high availability and quality of service. All of these factors, including the diseconomies of scale mentioned earlier need to be addressed before Carr’s Big Switch will be flipped on.
Carr’s cloud is a head fake for CIOs who don’t fully understand the network effects of system mobility. While Amazon and Google deserve recognition for making cloud computing into the buzz of 2009, it was VMotion that ultimately unleashed the promise of cloud computing by breaking the VLAN barrier and introducing massive potentials for server and desktop mobility. Yet until VMotion can be unleashed on motion-ready networks, it will be well-grounded in short term, tactical gains. Hence the questions about the scalability and economics of large, multi-tenant cloud environments on today’s kludge network managed by armies of clerks.
That’s why I think the Cisco and VMware relationship is strategically significant, and why VMware’s recent vShield Zones announcement may be a harbinger (that virtualization and cloud will steam ahead and surprise today’s “thought leaders” with unintended consequences and revitalize the network industry or at least those who get it).
You can get my thoughts in real-time at www.twitter.com/archimedius. I’ll also be moderating a cloud panel on May 18 at Interop and a dynamic infrastructure panel at Future in Review. If you’re attending either event feel free to stop by and say hello. Or you can simply join the conversation here.
I am a senior director at Infoblox.