Can Software Define a System?

A system can be anything. It can consist of only software pieces. In that case, we already know it is doable. But what if a system consists of software, hardware, and infrastructure, like a data center? The data center market is unique and many IT people, especially software applications and middleware people, tend to think it is not directly related to them. But any new trend, like BYOD, mobility, and the proliferation of sensor technologies, cannot exist without the backend. The frontend is fancier, sexier, and easier to see and understand. But without a solid backend, no new trend can be sustained. Moreover, some new technologies, like virtualization, came from the backend.

I have written a few blogs relating to “software-defined XXX.” At a recent Critical Facilities Roundtable technology group meeting, the software-defined data center (SDDC) was discussed.

Critical Facilities Roundtable hosted at Oracle in Pleasanton, California

The meeting consisted of a presentation followed by a panel discussion. I only had time to listen to a presentation by Krish Shivakumar (Manager, Product Management at Cisco Systems) and skipped the panel discussion. It is always nice to listen to experts in the subject matter and observe how they present the subject, including their presentation materials. I think Krish used a very effective set of slides to articulate what SDDC and software-defined network (SDN) were.

Krish Shivakumar

In Slide-1, he showed the transition of the rate of IT spending relative to GDP growth normalized to the average in the US. As new IT technologies were introduced, IT spending fluctuated, as shown in the picture. Notice that when the Internet was introduced and moved into the mainstream, IT spending decreased. But it is creeping back again, probably thanks to smart phones, which have fueled BYOD and mobility.

Slide-1

Then he blew up the leftmost part (the years between 2010 and 2016), as shown in Slide-2.

Slide-2

His remarks on the slide are noteworthy.

  • Virtual Desktop Interface (VDI) and Bring Your Own Devices (BYOD) are on the rise, while traditional clients are on the decline.
  • Service providers’ mobility capex are on the rise while wireline capex are on the decline.
  • Cloud (private, public, and hybrid), *aaS, and open/choice are on the rise, while traditional IT is on the decline.
  • Big data are on the rise and structured data are on the decline.

Then he defined SDDC as in Slide-3.

Slide-3

As shown in his definition, the IT components comprising a data center need to be virtualized. The high-level view should look as it does in Slide-4.

Slide-4

When we need to configure a data center according to our specifications, a particular combination of those three components is selected to form one, as shown in Slide-5.

Slide-5

Also, we need a set of APIs to control these components, as shown in picture Slide-6.

Slide-6

Krish then jumped into Cisco’s solution for SDN, Cisco Open Network Environment (ONE). I will not go into that. For those who are interested, this is a pointer to information about ONE.

He concluded by noting the current data center trends that go with the notion of SDDC:

  • Higher capacity utilization efficiency
  • Applications level resiliency and distributed applications plus smaller form-factor devices with denser packing
  • Increasing data center workloads that continue to drive power and cooling requirements upwards
  • Dynamic space, power, and cooling requirements controlling placement of workloads within/across data centers

Overall, his presentation was well designed and delivered. If you are new to the concept of SDDC, this is a good tutorial.

One thing that was missing from his talk was a reference to virtualization of facilities components, i.e., cooling and power. Cisco is an IT company and concentrated on the IT side, although they understand that both cooling and power play a big role in data center operations. Without the proper amount of cooling and power, we can lose space for placing ICT gear even when enough physical space is available, leading to stranded capacity.

Krish acknowledged that facilities should be plugged into his model. But it is not simple or easy. I discussed virtualization of cooling and power in previous blogs. See here (cooling virtualization), here (power virtualization – part 1), and here (power virtualization – part 2). How cooling and power can be factored into Krish’s model remains to be seen. But if this works, we can automate the design and placement of devices and equipment at data centers. A day may come soon when we can do this by creating an easy-to-produce configuration file and feeding it to SDDC software.

Zen Kishimoto

About Zen Kishimoto

Seasoned research and technology executive with various functional expertise, including roles in analyst, writer, CTO, VP Engineering, general management, sales, and marketing in diverse high-tech and cleantech industry segments, including software, mobile embedded systems, Web technologies, and networking. Current focus and expertise are in the area of the IT application to energy, such as smart grid, green IT, building/data center energy efficiency, and cloud computing.

, , , ,

No comments yet.

Leave a Reply


*