Physical Infrastructure Adapting to the Cloud
The design of the physical layer and the devices that populate it can have a significant effect on the performance of cloud-based architectures.
Conventional thinking holds that once you move data environments to the cloud, underlying physical infrastructure no longer matters. This is only partially true, however. Both the design of the physical layer and the devices that populate it can have a significant effect on the performance of cloud-based architectures.
This is evident in the large number of servers, storage systems and networking components hitting the channel that are "cloud-optimized." To be sure, a lot of this is simple cloud-washing, where you take a standard data product or service and slap it with a cloud moniker to make it seem new and edgy. A quick look under the hood, however, will usually get to the truth of the matter considering true cloud products will stress not only broad scalability and high throughput, but energy efficiency as well.
For example, F5 has added new application delivery controller (ADC) blades to its Viprion line aimed at handling the increased traffic loads and dynamic service environments that clouds produce. The 4480 and 4300 blades in particular are tailored to demanding application service environments through support for 40 gigabyte per second (Gbps) Ethernet, plus high L4, L7 and SSL throughput. They also provide upwards of 80 Gbps of hardware compression to support powerful application acceleration functions.
Elsewhere, you can find highly scalable storage solutions from companies like SolidFire, which utilizes a self-contained node design and 10 gigabyte Ethernet (GbE) networking to quickly scale capacity into the petabytes. As well, there are new network virtualization platforms like Nicira, which add fully abstracted layers between end hosts and legacy networks to more accurately pool resources to meet shifting data and application needs.
Cloud-driven changes are drilling all the way down to the silicon level, as evidenced by Intel's new Xeon E5-2600. The device is chock full of multi-thread management capabilities and high-I/O support designed to bolster hardware's ability to thrive as the age of the siloed data center comes to an end.
It would seem, then, that the cloud, or at least cloud-like functionality, is coming to your enterprise whether you want it or not.
On a more macro level, the cloud is looking to change the very nature of data architecture itself, from the server-centric, distributed format of the past decade or so to a much more modular, integrated and imminently scalable approach geared toward quick provisioning and reconfiguration.
The U.S. Army, which is usually out front when it comes to new technologies, is eagerly embracing HP's container-based "Performance Optimized Datacenter" (POD) systems as a way to boost readiness at home and abroad. The program is part of the Army's Private Cloud contract that aims to boost the use of services and distributed applications even is it reduces hardware and facility redundancy.
Physical infrastructure will never be as flexible or dynamic as virtual or logical ones, but that doesn't mean they are the modern equivalent of the ancient pyramids. Adaptability has been built into the server, storage and network elements that make up the three legs of the data center, even if it tends not move a little slower than upper-level data architectures.
Arthur Cole covers networking and the data center for IT Business Edge. He has served as editor of numerous publications covering everything from audio/video production and distribution, multimedia and the Internet to video gaming.