Some time ago, container based data centers got good attention because of its time to market, availability of high power density and focused cooling. As recent as last December, Tier-1 indicated that there is little traction by this technology. Lately, a lot of buzz on this subject are in the media and from people I meet, including a 20 feet container recently announced from HP, VMware and ones mentioned in here.

There are several vendors providing container based DCs, including:

• HP
• IBM
• Verrari
• Sun/Oracle
• SGI

The benefits of container based DCs include:

• Short time to market
• Availability of dense server (and therefore, power) concentration
• Savings of packing materials, such as boxes, extra power cords and so on

Some vendor put up to 22 racks of 42U (924 1U) servers. If you put blade servers instead, you could pack several thousands of servers per container with effective cooling. Because it is already set up as a mini DC, time to market is very short and you do not need to unpack each server box to get rid of empty boxes and extra power and networking cords.

As I mentioned, with some exceptions like Microsoft and Google, the market traction for this technology has been slow because many people consider this technology being only for large data center operators. I have informally solicited for opinions from several data center experts. There was almost 50-to-50 split between positive and negative opinions. The people with negative opinions claim this type of solutions only work for large DC operators and they are not large in number. Initially, I was in the same opinion. But after talking to two experts with positive opinions, I am beginning to change my stance, although I am not totally convinced yet.

	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
		Energy Efficient Cooling for Data Centers: A Close-Coupled Row Solution
	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

The positive opinion is summarized as follows:

• Because of the capital crunch and time to market, more and more enterprises are moving towards co-location
• Co-location/Whole Sale DC operators are becoming bigger by acquisition and other means to exploit economies of scale
• Cloud computing will accelerate the above two trends
• Containers are not property but equipment and thus, are not subject to property tax. They can be placed anywhere, including rooftops and parking lots, eliminating the need for large space to accommodate large IT loads.

The largest negative factor on the container based DCs is the granularity. Only large operators can use the container based DCs is limited in number. As the proliferation of cloud computing continues, more and more enterprises will move their roll your own DCs to DC operators’ sites for time to market and cost. This trend will make DC operators bigger and consequently; let economies of scale work well. For small players, a container-full of servers are too big to deal with but for large operators, the container may become a reasonable building block. Based on some conversation with a DC expert, fortune 500-1,000 companies are beginning to move to cloud computing type computing and this is actually happening.

Finally, not all the container based DCs are made in the same manner. Some are good and others are not that great. All of these data and information make me research further on the container based data centers. More to come…

As I write my daily blog, some readers contact me via several media, such as LinkedIn and email. My blog covers the United States and Japan, but not Europe, for the simple reason that I live in the U.S. and originally came from Japan. When you cover a market of a specific country or region, you need to have a good idea of the market.

I received a message from Philip Petersen, CEO of Ad Infinitum Multimedia (AdInfa) in the U.K., who wanted to chat. His company develops the InSite software suite which, like several others I have written about in my blog, provides metering and monitoring functions to data centers.

Unlike my other meetings in this space, this interview was conducted via Skype. The eight-hour time difference worked very well (my 8 a.m. is his 4 p.m.).

Petersen’s concern about my not covering the U.K. was not a problem for me. On the contrary, it was a great chance for me to understand the energy efficiency space of the U.K. market. The U.K. and other E.U. countries are, in general, ahead of the U.S. in the area of sustainability. The U.K.’s Carbon Reduction Commitment (CRC) is to go into effect this April, and many companies need to start measuring and metering power consumption (converted into GHG emission quantity that is to be regulated).

As I interviewed several metering companies, I got pretty good at asking questions so that I could compare the companies later. Here are some of the highlights of my conversation with Petersen:

• There are very few metering companies in the U.K.
• Customers include Equinix U.K., whose parent company is U.S. based.
• InSite can measure at any IP-addressable point - device, sensor, power strip, UPS, PDU.
• InSite can collect data from network management tools in addition to power consumption, temperature and humidity, and present it through a single web interface.
• InSite’s dashboards aggregate and integrate the collected data and display them in an easy-to-follow, interactive fashion.
• The entry price point is moderate for SMBs and the solution scales to enterprise level.
• InSite can be deployed on a licence or managerd service basis

Let me elaborate on some points. In spite of the imminent imposition of CRC—in less than three months—Petersen does not see much competition in the U.K. market yet. Having limited competition is good and bad news at the same time. Usually, the existence of competition assures the existence of a market. As the deadline for CRC comes near with little competition, it looks like smooth sailing ahead for Petersen and AdInfa.

We also chatted about where to measure and meter power consumption. The EPA’s Energy Star specification for data centers is in the final stage of development and is scheduled for release in April. The measuring point is at each UPS because most data centers measure power consumption at the UPS level. InSite can monitor power consumption at any point where there are metering devices; often customers want to do it on the PDU, which is more accurate than at the UPS.

Also of interest is the number of data points. In addition to information about power consumption, temperature, and humidity, there are much more data we could collect, especially from the building management system (BMS). This is what OSIsoft does. Sentilla collects what is available from network management tools. Arch Rock pushes IP as far as possible with a few prominent protocols from BMS. AdInfa takes an approach similar to Arch Rock’s and interacts with SNMP and others. It is not always true that more data points are better than a small but well-selected set. Personally, I think data from power consumption, along with that for temperature and humidity, should be combined with data collected from network management and some main BMS sources.

Finally, AdInfa is in an interesting position. The market for metering power consumption for containing GHG emissions is ripe and real in the U.K., and there is little competition. U.S. metering companies are mostly startups, with the exception of OSIsoft. Those U.S. companies are busy with the U.S. market first and will not concentrate their efforts on the U.K. or other European markets anytime soon. What AdInfa may want to do is to obtain as many U.K. customers as possible before U.S. companies enter the E.U. market. The metering space is in its infancy and will be too crowded in a year or two. After mergers and acquisitions begin to take place, the surviving companies will be sure to enter the U.K./E.U. market. AdInfa wants to have a majority of the U.K. market before that. If it does, it would be in an ideal position to select the best future course of action.

Although my main focus is the markets for the U.S. and Japan, I am more than happy to cover other territories in the world. If you have a great technology or an energy efficiency company, please contact me at info at alteterra dot net.

I met Jose Iglesias of Symantec when I hosted one of the sessions at the Green Software Unconference in August 2009.I talked to him briefly then and found out that his team was doing a lot in the green software area as well as in the overall sustainability work at Symantec. I finally had a chance to visit him recently to find out more about his team’s and Symantec’s sustainability efforts.

Jose is VP of Global Solutions and reports to the CTO, whose boss is the CEO. This position gives Jose easy access to the CEO. He wears two official hats and one informal one when it comes to responsibilities. Jose was a software developer before, and it was very easy for me as a former programmer to connect with him on the same wavelength.

Two of his official functions are to support customers in Japan and China and to provide solutions for the higher-level problems in the corporate-wide business. When his day job is over, he wears an informal hat to turn himself into “Mr. Green IT Man” to work on making Symantec and its customers greener. Although his formal functions are equally important and interesting, I encouraged him to discuss his role as Mr. Green IT Man.

Symantec recently surveyed more than 1,000 companies and found that the average company spends between $20M and $25M per year for power. Jose’s and others’ efforts allow Symantec to provide solutions that can reduce power consumption by more than 20%. Applying its own methods, Symantec has cut about $3M from its own power bill.

How did they do it? Jose gave me the following examples.

Data Centers
Usually, to ensure continued business operations, standby servers mirror a cluster of primary servers. Symantec used to use one mirror server for one primary server (1-to-1 mapping), doubling the number of servers required. Then the company developed a way to mirror n-to-1 to lower the number of standby servers without degrading SLAs. In doing so, they reduced the number of servers by two-thirds company-wide.

They also worked on storage, thought to consume 25–30% of data center power, which is significant, just below server power consumption. They take a three-step approach:

1. Examine overall storage allocation: Find out how data are allocated in the storage devices.
2. Make effective use: Use storage hierarchy to balance the access requirements of data to optimize energy efficiency.
3. Pay special attention to data: Remove redundant data. By applying its software solutions, along the lines of DeDuplication, Symantec reduced storage requirements by 70%.

Also, because Symantec’s IT is centralized, its data centers are accessed by international locations as well as local ones 24×7x365. In a way, it is easy to keep power at the highest level because the power requirement never fluctuates.

Offices
In the office environment, Symantec manages and controls endpoints, the PCs and workstations on the desktop. Symantec has 17,500 employees worldwide and 38,000 endpoints. An agent is installed on each desktop to manage assets, patches, and power, which helps to make the company’s offices green in terms of power usage. In applying its own solutions to its environment, Symantec is “eating its own cooking,” according to Jose.

In general, Symantec is very active in enterprise sustainability and pledges to reduce its carbon footprint by 15% by 2012 (as compared to 2008). I asked how they became so keen on sustainability. Jose answered that it was because of both upper management’s mandates and grass-root efforts. The Mountain View campus has been LEED certified, and employees are interested in making their company sustainable. Jose believes that people today want to work for a company that is perceived as socially responsible. And thus, being green is a good recruiting tool.

	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
		Energy Efficient Cooling for Data Centers: A Close-Coupled Row Solution
	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Finally, I asked him about the software for energy efficiency. Symantec enforces best practices for software development, such as in the  Norton suites, to make the resultant software energy efficient. Because a metric to assess the energy efficiency of software is necessary, Symantec sends an expert in measurement and metrics to The Green Grid, which is working on a metric to define useful work per watt.

Like NetApp, Symantec is full of sustainability-conscious people of all ranks and titles. I’ve said before that, compared with hardware companies, software companies tend to be indifferent to sustainability.

Clearly, Symantec is an exception.

The quest for efficiency improvement raises questions regarding the optimal air temperature for data centers. The ASHRAE TC-9.9 committee has recently adopted an extension of the recommended thermal envelope for server inlet temperature and humidity. A popular hypothesis suggests that total energy demands should diminish as the server inlet temperatures increase. This paper tests that hypothesis through the development of a composite power consumption baseline for a mixture of servers as a function of inlet temperature and applying this data to a variety of cooling architectures.

Download white paper

The trend of increasing heat densities in data centers has held consistent with advances in computing technology for many years. As power density increased, it became evident that the degree of difficulty in cooling these higher power demand loads was also increasing. In recent years, traditional cooling system design has proven inadequate to remove concentrated heat loads (up to and greater than 20 kW per rack). This has driven an architectural shift in data center cooling. The advent of a newer cooling architecture that was designed for the higher densities has brought with it increased efficiencies for the data center. This article discusses the efficiency benefits of row-based cooling compared to two other common cooling architectures.

Download white paper

Traditional methodologies for monitoring the data center environment are no longer sufficient. With technologies such as blade servers driving up cooling demands and regulations such as Sarbanes-Oxley driving up data security requirements, the physical environment in the data center must be watched more closely. While well understood protocols exist for monitoring physical devices such as UPS systems, computer room air conditioners, and fire suppression systems, there is a class of distributed monitoring points that is often ignored. This paper describes this class of threats, suggests approaches to deploying monitoring devices, and provides best practices in leveraging the collected data to reduce downtime.

Download white paper

High-density servers offer a significant performance per watt benefit. However, depending on the deployment, they can present a significant cooling challenge. Vendors are now designing servers that can demand over 40 kW of cooling per rack. With most data centers designed to cool an average of no more than 2 kW per rack, innovative strategies must be used for proper cooling of high-density equipment. This paper provides ten approaches for increasing cooling efficiency, cooling capacity, and power density in existing data centers.

Download white paper

Despite advances in computer technology, power outages continue to be a major cause of PC and server downtime. Protecting computer systems with Uninterruptible Power Supply (UPS) hardware is part of a total solution, but power management software is also necessary to prevent data corruption after extended power outages. Various software configurations are discussed, and best practices aimed at ensuring uptime are presented.

Download white paper

As 2009 draws to a close, I would like to ponder each element in the title of this blog. The information sources are my blog posts and others’. I will talk about data center market trends in this blog and treat the rest of the subjects later.

Regulatory pressure on data centers
The U.S. House of Representatives passed the cap-and-trade bill in June, and the Senate was expected to pass it in the fall. As we know, that did not happen. As we wait for the United Nations Climate Change Conference (COP15) to take place, the Obama administration has issued an executive order for all federal buildings to report their GHG reduction goal by the year 2020. Another noteworthy development is EPA’s designation of CO2 as hazardous material. This means that EPA can go after any organizations, including data centers, that emit CO2. These two changes, even without cap-and-trade, could affect the data center market significantly.

Mike Manos of Digital Realty Trust has warned us that regulatory pressure is around the corner. He recently noted that it is not a matter of if but when real regulations will be imposed on data centers’ GHG emissions.

New Trends
Two trends are emerging:

• outsourced data centers
• a new market

Because of the credit crunch and the consideration of speed to market, many enterprises do not want to build their own data centers. It is becoming quite complex to construct a data center in this environment of changing power availability/cost and regulations. If your requirement is for 2–5 k sq. ft., it makes sense to outsource your data center to someone whose daily business is to construct, operate, and maintain data centers.

	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
		Solution Brief: Building MPLS+ Managed Services
	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

In addition, a larger data center tends to be more energy efficient than a smaller one, mainly because of economy of scale. Many colocation and wholesale data center operators have been pushing this point. A large data center consumes a large amount of power and other resources, which makes it a good target for attack, but a large data center that is well run provides IT services very efficiently. To help this argument, we need a metric like useful work per watt, as The Green Grid advocates.

Regarding the development of a new market, currently the only way to measure GHG emissions is by metering and measuring power consumption. Companies that provide metering and measuring technologies and products will do very well. Another thought is that if GHG emissions become an issue, it won’t be long before water and e-waste are considered equally important. This opens up a discussion of sustainability from the IT perspective.

In one of the panel discussions at Tier1 Data Center Transformation Summit, there was a discussion of what enterprises want from data center operations. One such thing is monitoring and recovery. This means that metering and measuring should focus on merging many different kinds of data from several different types of data center equipment, such as network and system management, power consumption, and building management system (BMS). It is hard to tell how much data should be gathered and collected at which interval. Each metering company does this in its own way, and no standards exist yet. This space needs shaking up as the requirements for metering enter the mainstream.

There are a lot of technologies and practices for greening data centers, but I am focusing on metering and measuring. I have talked to several companies that provide metering and measuring solutions. As with Tim Canning at Arch Rock, after bumping into Prajesh Bhattacharya a few times at data center–related conferences and meetings, I took it as a hint to visit OSIsoft.
Recently, I met with Martin Otterson (by telephone) VP of sales, EMEA; Mike Gallagher, director of operations; and Prajesh Bhattacharya.

I know that visiting an interesting company takes time, and that it isn’t very green to emit GHG (my car is neither hybrid nor electric), but it sure beats just talking to them on the phone. A visit puts me personally in touch with the real people behind technologies and services. It also really helps to visit and talk to each vendor in the same space to understand their differentiation and where the market is going.

Moreover, a company that has been in business for more than 30 years tends to have a great website with a lot of information. Sometimes, though, too much information makes it hard for me to grasp an overview of a company. Martin did a great job of giving me an overview of OSIsoft in five minutes, which was very useful for me. So let me summarize his summary in the following:

• OSIsoft has been in business for more than 30 years, and the founder is still CEO.
• OSIsoft started by providing process control monitoring software for an oil refinery company.
• OSIsoft employs about 600 people and has approximately 15,000 installations of its software worldwide.
• OSIsoft is a software-only company, providing three kinds of components:

*Over 425 Interfaces to collect data
*a time series database called PI
*Analytics and client software to make data into valuable information

• OSIsoft provides the software infrastructure, and its customers can develop applications and services on top of it. The software infrastructure is implemented on MS Windows.
• There are currently over 425 interfaces supporting various protocols, data formats, and sensors.
• PI is the name of the core software component, the data archive. Over the years PI has enjoyed strong adoption in many process manufacturing industries, and most recently in the data center and critical facility market..

	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
		Electrical Efficiency Measurement for Data Centers
	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

I asked Martin if IP would become the protocol of choice in this space. He said “it depends upon ROI and risk, as the data center market is risk adverse . It may not make sense to rip out all the existing protocol and data formats and start all over from scratch with IP when the economy is so bad. Because sensors for many buildings and factories tend to be proprietary to protect their own markets”, OSIsoft’s way seems reasonable for now. This is totally new to me, coming from IT, which is constantly attempting to come up with a set of standards.

I asked about the metering and measuring market. “I believe the market is a just starting to evolve due.There is definite desire to have the ability to treat the facility/data center as a holistic system – measuring IT, Power and Mechanical performance” Currently, the great majority of data centers (about 80%) measure IT power consumption at UPS units. So the market for metering at each rack or server seems to be several years away. Martin referred to the US EPA report that suggests that facilities should start measuring everything they can and the U.K.’s carbon reduction commitment (CRC), which forces everyone to measure and meter their energy use. OSIsoft has customers in the U.K. and is very keen on this. Jim Smith and Mike Manos of Digital Realty Trust have warned us that a similar regulation will come to the U.S. in the not-too-distant future.

Martin’s thought on the Facilities market today ‘There is a large gap between the BMS and EEMS systems within facilities especially when enterprise users may have a large portfolio of facilities. How does an operator grasp the use of power and the efficiency of the mechanical asset or IT asset today? How can users to compare when efficiencies project are measured for success if there is not a holistic benchmark of information easily accessible.”

OSIsoft’s core business is managing process control data. When the company sees a market opportunity, it partners with a company like Microsoft, IBM, Cisco or Skanska to extend its coverage to exploit its core system. With appropriate partners, OSIsoft could add other attributes, like e-waste and water usage, to provide sustainability metrics. In addition, OSIsoft is involved in the smart grid/microgrid because the PI system can provide some of the functions of smart grid. I’m interested in investigating this area further.

After talking to OSIsoft, I think I gained some perspective on the metering and measuring space. I am putting together a one-page slide presentation to describe this market. One difficulty I face is how to describe metering companies’ offerings in conjunction with BMS providers and sustainability players like HARA/CSRware. I also would like to add smart grid to the mix. Whenever I talk to a vendor in this space, I need to revise the picture.

Because I cover Japan as well as the U.S., I often talk with Japanese data center operators about their interests. In these Internet times, some U.S. news is delivered to Japan instantly, but they want to talk to real data center operators and experts to find out how things are in their day-to-day operations.

Their three main interests appear to be cloud computing, economizer, and tier. I have written a lot about cloud computing and its relationship to data centers. Some Japanese data center operators wonder whether the future of data centers and cloud computing will be heavily consolidated to a small number in the U.S. If cloud computing makes the location of data centers irrelevant, all the Japanese data centers will not be necessary, and their operators may be out of work.

In discussing data center operators and doing some secondary research, my conclusion is that this will not be the case. The speed of the electrons that carry a packet on the Internet cannot exceed a certain limit. The distance between San Jose and Tokyo is about 5,500 miles, and the round trip is more than 11,000 miles. Some applications are insensitive to the speed of response, but for other applications, a slow response does not work. For that reason, most experts and operators do not think that data centers will be consolidated to a small number in the U.S. or that data centers in Japan will be out of business.

	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
		Avoidable Mistakes that Compromise Cooling Performance in Data Centers and Network Rooms
	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Because cooling uses 35% or more of the power consumed by a data center, if you eliminate cooling altogether, you should be able to reduce overall power consumption. Economizer for cooling is getting a lot of attention in the U.S., and Japanese operators are also interested in it. Japan’s data centers are mostly in Tokyo and the region between Tokyo and Osaka, where the majority of the population and businesses reside. The cooler climate in Japan is north of Tokyo, where a smaller concentration of the population and business is. This area tends to lack power and network access, so having data centers outside the Tokyo-Osaka region is not a good idea.

Economizer is a good idea, but Silicon Valley and Tokyo have very different climates. Silicon Valley enjoys a moderate climate with low humidity. Some say that outside air can be used without cooling for data centers there for nearly 80% of the year. Tokyo and its industrial region tend to be hotter and more humid in summer. Using the same criteria, economizer is not suitable for Japan.

Intel and Microsoft did some experiments using only outside air for cooling. They concluded that there was no problem in cooling IT equipment with untreated outside air. On the basis of that conclusion, some people claim outside air is the solution for cooling data centers. I am a little hesitant to make such a claim. Both experiments were conducted in a nonproduction environment, and their duration was not very long. Furthermore, I have not seen any real application of this idea in any production environment since those experiments.

Christian Belady of Microsoft has an interesting point of view. If vendors like IBM, HP, and Dell manufacture a server that can operate at a higher temperature, the possibility of using outside air will increase. Each component of the server could be cooled individually, and a fan could be redesigned so that it does not increase its speed when the inlet air temperature goes beyond 75 °F. Server vendors like NEC, Fujitsu, and Hitachi may want to consider developing a server to implement this idea. If successful, they can expand their market in this new world of data center temperature.

Finally, Uptime Institute developed the idea of data center tiers in 1996 and revised it in 2008. Recently, industry experts like Mike Manos of Digital Realty Trust suggested having only Tier II data centers (no redundancy for power or cooling). They are insisting on redundancy at the inter–data center level. Of course, we cannot yet implement this inter–data center approach instead of Tier III or IV. But this is the direction of data centers in the U.S.

After I bumped into the same person three times in a row in October at data center–related conferences and other events, I decided that something or someone must be telling me I need to spend more time with him. Tom Canning is a data center consultant at Arch Rock, which manufactures sensors for buildings and data centers and makes a network technology and an energy management application to collect, aggregate and analyze data like power consumption, temperature, and humidity.

Actually, Arch Rock is not totally new to me. Several years ago, I met Dr. David Culler at his Berkeley motes meeting when he was still with UC Berkeley. Later, I learned that he founded his own company while taking a sabbatical or something like that. At that time, Arch Rock was targeting buildings and factories for its business, which did not interest me too much. Later, the company added data centers as target customers. After repeated meetings with Tom, I decided to pay a visit.

	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
		Electrical Efficiency Measurement for Data Centers
	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

Unlike other meetings I’ve had with other companies, this meeting was over lunch. I did not take any notes or record the conversation, so I am writing this from memory. Roland Acra, CEO, joined us for lunch.

Arch Rock is in the same general space as the likes of SynapSense, Sentilla, Sensicast, and OSIsoft—in the business of measuring power consumption, temperature, humidity, air pressure, and chiller performance. One thing different about Arch Rock is its insistence on the use of standard IP technology, including on the sensors. Equally, Arch Rock’s Energy Optimizer application is based on open IT standards that enable easy data import and export with other enterprise applications for analysis or budgeting or alerts. Certainly, IP has become the protocol of choice for several network types and is growing in its application areas. However, in the building management system (BMS) space, IP is usually not spoken. Arch Rock installs its wireless IP-enabled gear where new visibility enabled by battery-based wireless mesh networking brings advantages, such as around data center racks and aisles, and it provides an interface with other protocols such as Modbus towards legacy power or cooling equipment where IP is not implemented. Although IP may not be the most efficient protocol compared to proprietary approaches, Arch Rock’s FAQ states quite convincingly why IP is fine in its product.

As for data centers vs. other buildings, Roland felt that people in the data center space seem to have less of a mental barrier to installing their gear, mainly because they are more technically oriented. Also, from a business point of view, data centers require more sensors than other types of buildings and, thus, represent bigger sales volume.

Speaking of installing a large number of sensors, I asked whether Arch Rock employees install the sensors or whether the company partners with others. After having its employees install the gear, Arch Rock lets customers install it. According to Acra, installation goes smoothly. This is very important because it usually takes a lot of adjusting in the configuration of sensors. The IP-based approach used by Arch Rock means that there is a widely available skill set of IP-trained professionals to deploy the sensor networks, which – along with the auto-configuration and self-healing properties of the Arch Rock technology – makes installation easy and robust.

I have talked to the measuring companies mentioned above in the past and plan to talk to others in the near future.

http://www.altaterra.net/members/blog_view.asp?id=288668&post=75780

The more I talk to these companies, the more confused I get about the market they are in. But after bouncing ideas with Roland and Tom, I think I can say something a little more reasonable about this space.

Even though measuring power consumption at rack level is certainly a good way to obtain more-accurate information, the vast majority of data centers measure power consumption at the UPS level (if they measure it at all), according to the draft Energy Star for data centers. Unlike power, measurements for temperature, humidity, and air pressure need to be taken at each rack.

It appears that it will be a few years before this sensor-oriented measuring technology enters the mainstream. The market is in an early stage, allowing multiple entrants into the space and waiting for someone to break out of the pack and take a leadership position.

I repeat here what I told Roland. It is very difficult to develop a technology and make a product out of it. But measuring as a concept is simple and even trivial, and there are only so many ways to measure the limited number of useful data types. Once sensor-based measurement becomes the standard, it will become a commodity and will be integrated into a bigger system. Then differentiation should come from such things as execution, alliance, distribution, and ecosystem. As for sustainability extension, Roland felt that it would come after Arch Rock’s current technology entered the mainstream.

In recent data center conferences, one thing that attracted my attention was cloud computing. For most people on the facilities side of the data center, cloud computing is pretty new. One keynote speaker emphasized the importance of cloud computing, and data center operators should take notice of it.

OK, you take notice, and what do you do to prepare for this tsunami? I discussed this with people at some interesting companies, such as OpSource, QualityTech, and Grid Dynamics. OpSource provides its cloud computing services as if it were a private cloud (I named it private external) without a data center.

Once upon a time, a Yahoo research executive said that there would be only five clouds (computers) in the world because it is very capital intensive to prepare backbones of cloud computing, which require gigantic data centers, huge amounts of IT gear, software, support staff, and so on. This may have been true at the time, but times are changing. If you detach cloud computing from a data center (container) as OpSource does, your capex for creating cloud computing goes down substantially. In addition to cutting costs, this allows you to implement cloud computing at any geographical location you like. Contrary to popular belief, you cannot access any cloud at any location. Because an electron cannot travel faster than the speed of light, we still need to worry about latency. A combination of something like OpSource’s technique and large, well-designed data centers that take advantage of economies of scale would be ideal for implementing cloud computing energy efficiently.

	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -
		Electrical Efficiency Measurement for Data Centers
	- - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - - -

How about data center operators themselves? Is cloud computing a curse or a blessing to them? Those who do not understand the internal working of cloud computing claim that it replaces data centers, and therefore data center operators are doomed. As above, if a small number (but certainly more than five) of huge cloud players dominate the market, then large, medium, and small data center operators will be wiped out.

Some disagree with that scenario. Data center operators I’ve talked with see cloud computing becoming mainstream over time, but even if that is the case, demand for colocation will not disappear. Some customers will still want to colocate their gear in a nearby data center. Data center operators also see an opportunity to team up with managed service providers and cloud system integrators like Grid Dynamics. One such example is Terremark’s partnership with VMware to provide cloud computing at an infrastructure level.

Another point of data centers is worth mentioning. An interesting feature of cloud is that from the outside the way it is implemented is completely transparent. It does not have to consist of one data center. One cloud may be implemented by more than one cloud. One large cloud player may run out of computing resources when an unforeseen demand occurs. If it has some contracts with several smaller data center players to provide their excess capacity, those data centers can provide their capacity to the larger one to meet the demand.

In this way, the large cloud provider does not have to overprovision too much. Smaller ones also can make their otherwise idle capacity available. I have been wondering about the greenness of cloud computing from the provider perspective. Maybe this is the way to go to make cloud computing green from both customer and provider perspectives.

Recently, I had an opportunity to meet with Richard Dym, CMO, and Koss Yokota, VP Asia Pacific Business Development, at OpSource’s headquarters in Santa Clara, California, about their brand-new offering, OpSource Cloud.

Their web page is the best source of information about who they are and what they do. However, like any other company, they have several different offerings, each fairly complex, and it was hard for me to grasp their operations at the highest level. So Dym gave me a comprehensive presentation.

OpSource’s main business has been to provide comprehensive services at three operational levels to make sure clients’ applications run smoothly. These are the levels:
Technical—Provides technical infrastructures, such as servers, in a colo data center.
Applications—Tunes applications to run without any issues.
Business—Provides metering, CRM, etc.

Its latest offering, announced October 5, is OpSource Cloud.

Dym mentioned that although Amazon is a leader among the few cloud computing companies, it does not have many enterprise customers. This is because Amazon’s offering is weak in security, performance, and control.

OpSource Cloud provides:

• Security—Private firewall
• Performance—Multitier architecture
• Control—Many layers of passwords
• Standards—De facto standards
• Support—24 ´ 7 phone support

This is very similar to what I described in my previous blog. Here is a recap: Cisco’s James Urquhart introduced the notions of “internal” and “external” to cloud computing. His point was that two factors should be considered in the classification of clouds: control (public vs. private) and physical location (internal vs. external).

These factors create four possible combinations of clouds:

1. Public and external
2. Public and internal
3. Private and external
4. Private and internal

Numbers 1 and 4 are the original definitions of public and private clouds, respectively. Number 2 does not make sense and should be removed from the list.

I argued that when you run out of computing capacity in your private internal cloud, you burst the private cloud to exploit private external cloud, as shown in the previous blog. What OpSource provides is the private external cloud that it calls OpSource Cloud. My question is whether there is a minimum charge to reserve some computing power in the private external (outsourced) cloud. OpSource does not charge a minimum fee; it enforces the pay-as-you-go scheme 100%. This is great for clients, but how about for OpSource? It is hard to foresee IT demands and the right level of IT gear–provisioning ahead of time in the private external cloud.

Dym said that this is the secret sauce for forecasting and provisioning IT gears and power at a right level. The cloud offering is hosted at an NTT data center, and NTT gives OpSource power as needed. This is too good an arrangement. Why? NTT is the most recent investor.

This visit was very interesting from the following points:

• Private external cloud is actually implemented.
• Private external is targeted to enterprise customers.

This was the first time I heard about the actual implementation of private external (outsourced) cloud. With real private external cloud implementation, cloud computing may enter the mainstream of enterprise, although Dym thinks outsourced cloud has not crossed the chasm yet, á la Geoffrey Moore.

Finally, I think OpSource plays its game well without having its own data center, as opposed to the likes of Amazon and Rackspace, which do. OpSource can pick and choose data centers to implement their outsourced cloud offerings in selected geographical areas. They leave the data center facilities management to colo players like Equinix and NTT. In the previous blog, I wrote that only a small number of companies (five, according to a Yahoo research executive—IBM, Microsoft, Yahoo, Google, and Amazon) can afford to be a cloud provider because it takes an abundance of money to implement a cloud offering with very expensive data centers. But the way OpSource plays its game, there could be more cloud offerings with data center operators like Digital Realty, Terremark, and other large colo players.

Andy Parham, CEO of Bick Group, delivered the keynote speech at Data Center World.

In short, he said that a tremendous number and extent of changes are coming to the data center segment. To emphasize his point, Parham used two examples. The first was alternating current’s takeover of the power market from direct current, carried out by Samuel Insull, an apprentice of Thomas Edison. The second was a clip from the movie “Poseidon,” showing the ship being hit by a giant rogue wave.

Parham listed five Cs to further analyze the smaller waves in the big wave:

1. Cost: The cost of all IT gears (servers, storage, and networking) is declining.
2. Consumption: Developing countries like China, India, Brazil, and Russia are starting to consume a lot of IT resources.
3. Convenience: Ease of use
4. Componential: Systems are increasingly constructed with smaller modularized standard components.
5. Cloud