Cloud Computing: An Escapist Fantasy?
Cloud computing will need to automate IT processes and kludge
By: Greg Ness
Feb. 27, 2009 03:30 PM
Before we get to carried away with the cloud computing dream lets talk about the fundamental challenges in the network (manual labor and kludge) that may have a substantial impact on the shape of cloud computing, at least in the enterprise.
Newspapers were originally fed on presses manually one sheet at a time. Telephones were once connected via legions of operators connecting callers (one at a time) with cables and plugs; and today computer networks are still managed by legions of manual administrators who configure network appliances and manage IP addresses as endpoints are added or moved or networks are acquired. Yet can cloud computing really take that pain away?
How cloud evolves will be partly driven by technical limitations and partly by business case economics; and the way in which cloud is adopted may have a significant impact on the market caps of a host of networking and IT powerhouses, from Cisco and Juniper to Microsoft, IBM and F5 Networks.
Is Cloud an IT Escapist Fantasy?
They may be the cloud dream's most promising market: those so out of touch with the micro expenses and delays within their network that everything else looks cheap and flexible. For them, the promise of cloud offers an escape from the nightmare of a thousand cuts.
The question is really to what extent the cloud leaders, including Google, Amazon and others have really addressed these problems. Or are they simply writing off these costs as insignificant because the revenue generated is still insignificant?
All of these common network tasks (and many others) require ongoing human intervention and thus drive up network costs as networks continue to grow. They also impact a networks availability and ability to address new requirements. As long as networks are islands of manual, anachronistic tools to what benefit are mobile servers and endpoints? They cannot move very far without manual intervention, pigeonholing the very benefits of virtualization... in a very un-cloudlike manner.
The Free Hypervisor Illusion
While these virtualization vendors play with predatory pricing, the real payoffs are perhaps delivered by the partner ecosystems. That's why hypervisor-based pricing may miss the point. It's really what you can do with that hypervisor that drives the value. That's why VMware has an enviable position, despite the Microsoft and recent Citrix price-centric marketing moves.
VMware is also further along in technology, features and deployments. VMware's relationship with Cisco and the other ecosystem power players could be a second advantage and be a driver behind the Microsoft and Citrix alliance of necessity. Then where does that leave Juniper and IBM? Perhaps in a similar alliance.
If the likes of VMware, Cisco and F5 Networks (and others) can deliver the connectivity intelligence that would allow networks to automatically keep up with the status of a VM, it will be a massive payoff that could impact the market caps off all involved. They could, in effect, drive the manual expense out of the network and establish new economies of scale and performance.
It seems likely that connectivity intelligence will require the cooperation of multiple players, including Infoblox (my employer). Core network service automation may be step one; yet step two would likely involve the deployment of IF-Map.
The bottom line is that the more expense that can be driven out of the network the more gear the vendors will sell and the lesser the business case for the coming cloud disruption with or without an economic rebound. Free gear could cost more in management and consulting and merely hide the core problems even further.
It's Not the Economy, Just Economics
These costs grow every year under the canopy of business as usual while the common denominator is really an ironic inability to deliver automation where it could really make a difference... on the network. That's right; the network may be one of the last bastions of manual labor in the modern enterprise; and it is a critical factor in the proliferation of many types of cloud computing.
While systems pros dream about the day when virtual machines can follow the moon (e.g. VMs move from one cloudplex to another chasing cheap evening electrical power around the world) networking pros are already under siege from more endpoints, more manual processes and more business operations risks. It is this disconnect between the cloud computing buzz and the reality of life managing a network (where most endpoints don't yet move) that ignited the Infrastructure 2.0 conversation.
Dreaming Isn't Free
Those who grasp the potentially threatening gap between system and endpoint mobility and static networks requiring ever higher rates of manual intervention will either drive for core network services automation or keep the cloud genie contained within dozens (if not hundreds) of VLAN bottles, called virtualization-lite. The latter scenario I've called The CIO Shell Game.
That's why I think cloud computing is really a code word for dynamic systems, storage and infrastructure. And you need all three. Cloud computing without a network capable of keeping up is merely a fantasy propped up out of frustration with the "kludgenomics" of today's increasingly expensive infrastructure.
As much as the headlines about the economy may convince us otherwise, cloud is more likely an artifact of rising management costs more than a weak global economy. I think the vendors who minimize those costs via automation will win. Those who simply try to discount their hardware as a way to make room for the operational pain will lose.
I am a senior director at Infoblox. You can follow my comments in real time at www.twitter.com/archimedius.
Latest Cloud Developer Stories
Subscribe to the World's Most Powerful Newsletters
Subscribe to Our Rss Feeds & Get Your SYS-CON News Live!
SYS-CON Featured Whitepapers
Most Read This Week