A Cloudy Future for Networks and Data Centers in 2010
December 23, 2009 Leave a comment
The message from the VC community is clear – “don’t waste our seed money on network and server equipment.” The message from the US Government CIO was clear – the US Government will consolidate data centers and start moving towards cloud computing. The message from the software and hardware vendors is clear – there is an enormous investment in cloud computing technologies and services.
If nothing else, the economic woes of the past two years have taught us we need to be a lot smarter on how we allocate limited CAPEX and OPEX budgets. Whether we choose to implement our IT architecture in a public cloud, enterprise cloud, or not at all – we still must consider the alternatives. Those alternatives must include careful consideration of cloud computing.
Cloud 101 teaches us that virtualization efficiently uses compute and storage resources in the enterprise. Cloud 201 teaches us that content networks facing the Internet can make use of on-demand compute and storage capacity in close proximity to networks. Cloud 301 tells us that a distributed cloud gives great flexibility to both enterprise and Internet-facing content. The lesson plan for Cloud 401 is still being drafted.
Data Center 2010
Data center operators traditionally sell space based on cabinets, partial cabinets, cages, private suites, and in the case of carrier hotels, space in the main distribution frame. In the old days revenue was based on space and cross connects, today it is based on power consumed by equipment.
If the intent of data center consolidation is to relieve the enterprise or content provider of unnecessary CAPEX and OPEX burden, then the data center sales teams should be gearing up for a feeding frenzy of opportunity. Every public cloud service provider from Amazon down to the smallest cloud startup will be looking for quality data center space, preferably close to network interconnection points.
In fact, in the long run, if the vision of cloud computing and virtualization is true, then the existing model of data center should be seen as a three-dimensional set of objects within a resource grid, not entirely dissimilar to the idea set forth by Nicholas Carr in his book the “Big Switch.”
Facilities will return to their roots of concrete, power, and air-conditioning, adding cloud resources (or attracting cloud service providers to provide those resources), and the cabinets, cages, and private suites will start being dismantled to allow better use of electrical and cooling resources within the data center.
Rethinking the Data Center
Looking at 3tera‘s AppLogic utility it brings a strange vision to mind. If I can build a router, switch, server, and firewall into my profile via a drag and drop utility, then why would I want to consider buying my own hardware?
If storage becomes part of the layer 2 switch, then why would I consider installing my own SAN, NAS, or fiber channel infrastructure? Why not find a cloud service provider with adequate resources to run my business within their infrastructure, particularly if their network proximity and capacity is adequate to meet any traffic requirement my business demands?
In this case, if the technology behind AppLogic and other similar Platform as a Service (PaaS) is true to the marketing hype, then we can start throwing value back to the application. The network, connectivity, and the compute/storage resource becomes an assumed commodity – much like the freeway system, water, or the electrical grid.
Flowing the Profile to the User
Us old guys used to watch a SciFi sitcom called “Max Headroom.” Max Headroom was a fictional character who lived within the “Ether,” being able to move around though computers, electrical grids – and pop up wherever in the network he desired. Max could also absorb any of the information within computer systems or other electronic intelligence sources, and deliver his findings to news reporters who played the role of investigative journalists.
We are entering an electronic generation not too different from the world of Max Headroom. If we use social networking, or public utility applications such as Hotmail, Gmail, or Yahoo Mail, our profile flows to the network point closest to our last request for application access. There may be a permanent image of our data stored in a mother ship, but the most active part of our profile is parsed to a correlation database near our access point.
Thus, if I am a Gmail user, and live in Los Angeles, my correlated profile is available at the Google data cache with correlated Gmail someplace with proximity to Los Angeles. If I travel to HongKong, then Gmail thinks “Hmmm…, he is in HK, and we should parse his Gmail image to our HK cache, and hope he gets the best possible performance out of the Gmail product from that point.”
I, as the user, do not care which data center my Gmail profile is cached at, I only care that my end user experience is good and I can get my work done without unnecessary pain.
The data center becomes virtual. The application flows to the location needed to do the job and make me happy. XYZ.Com, who does my mail day-to-day, must understand their product will become less relevant and ineffective if their performance on a global scale does not meet international standards. Those standards are being set by companies who are using cloud computing on a global, distributed model, to do the job.
2010 is the Year Data Centers Evolve to Support the Cloud
The day of a 100sqft data center cage is rapidly becoming as senseless as buying a used DMS250. The cost in hardware, software, peopleware, and the operational expense of running a small data center presence simply does not make sense. Nearly everything that can be done in a 100sqft cage can be done in a cloud, forcing the services provider to concentrate on delivering end user value, and leaving the compute, storage, and network access to utility providers.
And when the 100sqft cage is absorbed into a more efficient resource, the cost – both in electrical/mechanical and cost (including environmental costs) will drop by a factor of nearly 50%, given the potential for better data center management using strict hot/cold aisle separation, hot or cold aisle containment, containers – all those things data center operators are scrambling to understand and implement.
Argue the point, but by the end of 2010, the ugly data center caterpillar will come out of its cocoon as a better, stronger, and very cloudy utility for the information technology and interconnected world to exploit.