If I take a look at power consumption beyond data centers, my current thinking is that the market will follow this sequence:

1. Mitigation with power consumption
2. Mitigation with energy consumption
3. Carbon footprint awareness
4. Renewable energy
5. Smart grid

Among the above, I think the mental transition from #1 to #2 is one of the biggest changes. Power is power no matter how it is generated. However, once you realize that power is the transformation of energy from many sources, the whole picture changes. It opens up the discussion of how power is generated and how much GHG entered the atmosphere because of it. Then, what are the remedies for that? A clear solution seems to be to exploit renewable energy sources and construct new infrastructure, namely smart grid. I have blogged on these issues and plan to continue to do so.

At the same time, as an analyst, I should be aware of the current state of the data center market. When I’m conducting research and talking with data center operators who deploy bleeding-edge technologies and practices, I tend to miss what the current market is.

The current data center market still suffers from many of the problems discussed by many people and covered in the press repeatedly:

• Uncontrolled increase in power consumption and, thus, power cost
• Lack of cooling capacity to accommodate intense heat by IT and facilities equipment
• Lack of space
• Mismatches between power consumers and the power bill payer
• Low utilization ratio of servers and other IT equipment (low virtualization adoption rate)

If you did just one thing to cope with those problems, what would it be? You would monitor and measure power consumption at data centers. It sounds so easy, but there are still a lot of data center operators who do not want to do so. People cannot understand what they cannot see. Without monitoring and measuring, you cannot see how much power is used when and where and how. I hate to repeat the well-known adage “you can’t control what you can’t measure,” but it is really true.

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Deploying High-Density Zones in a Low-Density Data Center
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

There may be multiple reasons why data center operators do not monitor. One could be that they lack the expertise. Recently, I visited the Redwood City headquarters of  Sentilla. The company develops and markets solutions for monitoring and measuring power consumption at data centers and industrial buildings.

My questions focused on:

• Who they are
• What their differentiations are
• Where they are going

Bob Davis, CEO, a former senior vice president at CA and a veteran of several startups, answered my questions.

Who They Are

Sentilla was founded in 2003 and started monitoring the power consumption of industrial buildings. They got funded in 2006 and entered the data center energy efficiency segment in 2008. They are headquartered in Redwood City and have a branch in the U.K. The European Union, especially the U.K., has more severe environmental laws than the U.S., and power is more expensive in the U.K. than in the U.S. So the U.K. branch makes sense, with a tail wind (like CRC) to capture the market with little competition.

Remember that there are two kinds of measuring and monitoring companies. One kind installs its own sensors, collects/aggregates data, and displays the result. The other kind does not deploy its own sensors but aggregates data and displays the result. Sentilla is a hybrid of the two. It collects information readily available from the following sources to visualize the power consumption at data centers:

• System software
• EMS (enterprise management system)
• Service operations
• IT equipment
• Facilities equipment

The data are collected more or less at each rack level. But if you need more detailed monitoring for each piece of equipment, they sell three kinds of hardware for monitoring:

• Wireless PAU (power analysis unit)—attaches to each piece of equipment
• Wireless gateway—aggregates data collected by PAUs
• Appliance for mounting a web-based user interface (UI)—runs on Linux

The third kind is optional, and they can install their UI software on a customer’s Linux box.

Their first customer shipment was January of this year, and they are working with multiple customers.

Their UI example is given here:

Three servers’ power consumption information display with power cost and GHG emissions information

- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
Humidification Strategies for Data Centers and Network Rooms
- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Differentiation

In my  report, I covered several companies in this segment. From 30,000 feet up, it is not easy to see the differences among those companies and what matters most.

Davis emphasized their software architecture and design. Their software is written in Java whether it is in the appliance, embedded PAUs, or gateways. Because of the design, it scales nicely. Currently, their software deals with hundreds of nodes, but towards the end of the year, it will be able to deal with thousands of nodes. And in the future, it should be able to handle tens of thousands of nodes. I was wondering if configuring the UI to receive data from multiple sources takes a lot of time. Most professionally run data centers keep their equipment information in a CMDB (configuration management database) and/or an LDAP database, and it is straightforward to set this up. On the average, the most it takes is one or two days without service interruptions.

Another difference is Sentilla’s IT-centric view as compared with the facilities-centric view of others. The best way to reduce power consumption is to save power consumption on IT equipment. For that, the IT-centric view is useful.

Future Directions

The discussion with the Sentilla folks made me realize that companies like Google and Microsoft can afford brand-new data centers with their bleeding-edge technologies to run most energy efficiently, but most other data centers are not equipped with any tool to visualize power usage. Selling is never easy, but with logistics taken care of, Sentilla’s sales pitch is straightforward because the ROI is very clear and can be visualized.

For several more years, this segment will stay independent and will be alive and kicking. But over time, Sentilla needs to team up with other management tool companies like CA (Unicenter), HP (OpenView), IBM (Tivoli), and BMC (Patrol) to provide a wider scope of management capabilities.

Tags: Data Center, data center power consumption, sentilla