Features

Jayshree Ullal, senior VP data center, switching and services group, Cisco, is responsible for driving the direction and execution of Cisco's switching, security and the company's expansion into data center product lines. She believes that the next generation of data centers -- dubbed data center 3.0 -- will give enterprises more flexibility and reduce capex and opex.

Can you tell us what data center 3.0 means to the enterprise?

The main three areas of focus of data center 3.0 are consolidation, virtualization and automation. The trend in data centers is to go from many data centers in an enterprise to fewer. The main drivers are power, cooling, space constraints, and the massive proliferation of underutilized machines. The focus of our consolidation is to bring greener data centers and improve capex and opex.

What are the chief concerns of CIOs?

CIOs are worried about making their data centers greener, and in reducing the latency of their networks and their applications.

What is wrong with the way data centers are managed today?

Let me give you a comparison -- today, when you go to see a doctor, you don't straight away go to a heart specialist. You first go to a general practitioner, and then he sends you to a specialist. Today, in a data center, everybody is an isolated expert on servers, or storage, or networks, or power. What is lacking is a general practitioner for a data center.

What is the primary push for green data centers?

There is a social responsibility to have clean sources of power. The other factor is cost savings. Let me allude to the savings by giving an example -- in my house, if I have one switch that automatically turns on nine lights, it will not be the best way to use power because invariably my children will leave this switch on all the time!

This is where the Nexus 7000 comes in. In the Nexus 7000, the way we have designed the power supplies is something we've never done before. The power supplies are placed at the bottom of the switch -- we actually have three of them. You can keep all three on, but we have an auto usage and shutdown system that turns off the power supply when it is not required. This can save a lot of power.

A typical large data center with around 50,000 servers consumes around 20 to 30 MW of power. As a ballpark figure, you could say that 10,000 servers consume around 10 MW of power. Using the technique for power saving that I described, I can save 10 percent of that 10 MW. If I save 1 MW, think about the cost of power and how many millions can be saved.

In the heart of our Nexus 7000, we have built a lot of IP capability including nine new ASICs. We design our own chips and this is how we get performance and reliability. In the ASICs, we are making sure that the power consumption is low.

In a Greenfield data center, one has a chance to literally build one's house from scratch! If you are an existing data center, you can't do that -- instead you should focus on the hottest things and see what can be done with limited space, power and budget. Probably the most common things that I've seen getting deployed in existing data centers are virtualization and application delivery.

What other areas should CIOs concentrate on?

Application delivery is something that CIOs need to concentrate on. Suppose I send you a file and you make some changes, and this goes back and forth a thousand times, and each file transfer takes 4 milliseconds, I can spend 40 minutes for the exchange of one PowerPoint presentation. Instead, if I consolidate on what is called the wide area application services (WAAS), my network looks into your content and instead of sending the entire file, and it sends only the changes back and forth. Now, in this case, I'm using my bandwidth more efficiently. We have built this into our networks.

One more area of importance is compression. Today, all sorts of traffic like file traffic, Web traffic, exchange traffic, and video traffic flows through the network. Compression can make the network respond faster and give huge savings. It will also allow remote users to "feel" local by giving them very good response time. Think of the benefits for areas like finance, where one millisecond can represent millions of dollars.