Greg Ness at Archimedius has raised a very important concern from the cloud
computing perspective. He is arguing that the current network infrastructure and diverse
range of protocols cannot serve the needs of cloud computing and it needs
Infrastructure 2.0. His argument is extremely valid and I strongly urge the
readers to check out his post to learn more about why network infrastructure
needs an upgrade. There is no denying to his argument that the infrastructure
upgrade will greatly accelerate the enterprise adaption of cloud computing.
I fully agree with his arguments about the need for more dynamic network
infrastructure but I disagree with his conclusion. He seems to argue that cloud
computing cannot take off until Infrastructure 2.0 is in place. I don’t agree
with it. Cloud Computing in enterprises can definitely take off and I would like
to present the following two arguments to support this.
My co-author Ben Kepes talked about a very important point in his post sometime back. Unfortunately, this point got
sidetracked when the focus of the post and the ensuing debate, in the comments
section, got shifted towards the criticism of the demands of US consumers. In
his post, he argued that
All of these issues are immensely important to SaaS vendors – your market is
everywhere – in the US with fantastic speed and bandwidth, in New Zealand with
moderate speed and bandwidth and in the third world where connectivity is dicey
at best. There are two commandments you must adhere to;
- Build it as lightweight as possible
- Build in offline access
This is a very important point pertinent to the faster adoption of SaaS
technologies. SaaS vendors have realized the importance of lightweight
applications and are slowly moving towards it. As more and more SaaS vendors
shift their focus to building and releasing light applications, the strain on
the infrastructure will also greatly reduce, leaving some breathing room for
enterprises to start using Cloud Computing even before they make plans to
implement Infrastructure 2.0.
My second point is that as more and more enterprises understands the benefits
of Cloud Computing and join the bandwagon, maybe with an intermediate path of
hybrid On-Demand & On-Premise Computing, the demand for better
infrastructure will see a huge increase. This increased demand will result in
lower infrastructure costs and hence a higher and faster enterprise adaption of
Infrastructure 2.0.
I don’t have the access to the kind of data Greg has and, hence, my arguments
are more qualitative than quantitative. Even then, the low cost of SaaS for
enterprises combined with lightweight apps and lower cost of infrastructure
upgrades, will ensure that enterprises will move to Cloud Computing without
waiting for Infrastructure 2.0 to happen. What do you think? Do you think we
will see enterprises jumping into Cloud Computing before they build
Infrastructure 2.0 or do you think Greg is right and I am just showing the
emotions of a Cloud Computing evangelist? Feel free to add your take in the
comment section.
Krishnan:
Thanks for mentioning my Archimedius blog. I think you are correct that cloud will take off regardless of Infratsructure2.0. Clearly it already has. I think I2.0 will help it accelerate at a faster pace, by enabling more mobility and more robust services.
Thanks
Greg
Krishnan,
I agree with Greg’s post – cloud computing requires Infrastructure 2.0, absolutely. For the most part, a lot of the application delivery infrastructure out there is already Infrastructure 2.0 “ready” – it’s flexible, highly scalable, and able to be integrated via interoperable standards.
The bigger problem is getting cloud computing providers to recognize the importance of infrastructure in assuring availability and performance needs of customers are met. Many do, but many still have their “head in the cloud” regarding this aspect of infrastructure architecture.
DNS and DHCP, as Greg points out, are also often overlooked, and the inherent scalability issues associated with TCP cannot be underestimated.
Cloud computing capable of supporting enterprise class applications – whether lightweight or not – requires more than just slapping together a couple web servers and some virtualization technology. It requires an investment in infrastructure that’s capable of scaling up with demand, and being flexible enough to provide monetization opportunities through additional value-add offerings (acceleration, security, etc…) while simultaneously having a mechanism through which that infrastructure can be integrated into the processes that make the cloud work.
I agree we can start now, with what’s available, but as you also say providers need to be aware that in the long run it’s going to take more awareness of the infrastructure in order to make this a viable (and for them, profitable) computing model.
Lori
I think Greg Ness’s infrastructure 2.0 piece is one of the most insightful commentaries forecasts on infrastructure and IT I have read in some time.
Infrastructure virtualization is a key next step toward 2.0. In support of Greg’s position, abstracting network devices and the services that run on them (routing, switching, FW, VPN, QoS, etc)in order to automate their configuration ideally in response to end-user requirements is critical. For example, when VMs are moved physically or virtually the infrastructure services required to provide access, security and connectivity to VM resources should be automated. 2.0 needs a new architecture capable of abstracting users, applications and how they interact (associations) at a “business level” into a policy statement that is not bound to but has knowledge of infrastructure at a technical level (configuration). This architecture achieves true infrastructure virtualization (and automation which = reduced TCO). For example manage the relationship between users and VM applications via a policy “top down” rather than “bottom up” and automate from the policy statement to infrastructure services and elements. Security is implicate because infrastructure services are configured on the basis of selective inclusion (policy definitions)…..no one is on net unless defined by a policy VS selective exclusion (today’s “static” infrastructure 1.0) where-by everyone is on net unless ACLed out by a legion of technicians sitting on the command line. Clearly the latter will not scale technically or economically.
Hats off to Greg for fielding an important editorial for all of us in the information age. Great piece.
Harley