|By Larry Carvalho||
|July 14, 2012 12:00 PM EDT||
I recently participated in a panel discussion, titled: "Infrastructure as a Service (IaaS) endgame - commoditizing the data center" at the GigaOM Structure 2012 Conference. While software is definitely defining the use of IaaS through Platform as a Service (PaaS), the underlying need for IaaS does not go away. In my opinion, the current pace of infrastructure and data center build out cannot be maintained, leading to a future need of resource sharing in this space.
Resource sharing includes everything that contributes to a data center, especially the physical infrastructure. Several Software as a Service (SaaS) providers, especially in the collaboration space, innovatively optimize their infrastructure to serve their specialized purpose. Facebook, for example, has open sourced their hardware design through Open Compute. Skype, as you know, routes calls and video through their users' bandwidth, reducing the need for back-end infrastructure.
With an increasing focus by companies on environmental responsibility, it would be much better to productively re-use physical data centers (and leverage what we've already got) rather than convert them into basketball courts like I have seen in a recent HP advertisement.
Cloud computing is evolving, and, with the transition, many companies are asking this question: Why should I move IT infrastructure to the cloud when I already have invested in my own data center? Existing data centers are equipped with a lot of capabilities required for high availability, like uninterrupted power supply (UPS) and network bandwidth. Companies moving existing workloads to the cloud may find more capacity than they need in the data centers they already own. At the same time, more cloud service providers are looking for space to expand their capacity for new customers.
This can provide an opportunity for everyone to improve resource usage.
As an analogy to what I'm talking about, we can look to power companies. These utilities have been around a long time, supplying power to customers via a massive electricity grid. Increasingly, there is significant investment going into financing solar panels and windmills that can contribute excess capacity back to the grid. In fact, this is becoming popular among both businesses and consumers, who can reduce their power bills and earn credits.
Why can't we apply this same concept to cloud computing, where everyone "contributes" to the grid?
My proposition is for cloud service providers to not invest in new infrastructure but to evaluate the option of leveraging existing data center facilities. In an environment of cloud computing driven by software management, cloud service providers could control the usage of IT resources from multiple locations, improving the value proposition.
Service providers that already provide peer-to-peer sharing capabilities include Spotify, Freenet, Gnutella, BitTorrent and Kazaa - but these are very consumer-focused solutions. Compute sharing enabled by BOINC, the open-source software for volunteer computing and grid computing, powers a number of grids, the largest being World Community Grid primarily targeted to home users. Many of us have also provided home computer resources to help SETI in its search for extraterrestrial life in the universe.
We need to take these consumer examples and apply them to the business market. If IaaS can be considered to consist for the most part of compute, storage and network resources, I find it devoid of solutions available to businesses, with the exception of storage sharing by Symform. There is a wide open opportunity for software innovation to enable network and compute sharing in the enterprise space that would be very attractive not just for costs but also to enable us to truly "go green".
Cloud service providers need to target complete IaaS sharing by leveraging existing data center resources. This will enable infrastructure re-use and also encourage large data centers to embrace the cloud computing paradigm shift without wasting existing investments. We could see mini-Amazons in place enabled by cloud service providers while improving high-availability. I foresee several companies building on the storage sharing concept pioneered by Symform to include all IaaS components.
The main challenge is to complete the software components capable of managing shared resources in a secure manner (and most likely an open API infrastructure) required to pull this off.
• • •
This post was first published on Symform.com. Republished with permission.