Blog Host: Carrie Higbie - Global Network Applications Market Manager, The Siemon Company READ ENTIRE BIO

• Modular, upgradeable and temperature-neutral cooling concept.
• Max. 12 kW cooling output, with three cooling modules possible per cooling rack.
• Approx. 4 kW cooling output per cooling module at 800 m³/h air throughput.
• Optimised air routing: Cold air is blown in at the sides, so that the cold air is evenly distributed in front of the 482.6 mm (19″) level.
• Optionally 1 or 2 server racks may be cooled -- stand-alone and bayed sitting possible.
• Bayable to TS and PS server racks, even retrospectively.
• Active condensate handling.
• Optional control and monitoring of the cooling rack via Rittal CMC-TC.
• Two-piece modules; for easy installation in narrow passageways.

These units can also provide supplemental cooling, because they take advantage of chillers that are underutilized.

Adapting cooling to current equipment and the changing power environment

Overhead supplemental cooling
The cornerstone of adaptive cooling approach is the as-needed addition of spot or zone supplemental units. Supplemental cooling works in concert with traditional under-floor systems by cooling at the source with advanced compact heat exchangers. This method provides the necessary incremental, 100 percent sensible cooling for the added heat load with minimal to zero floor space consumed. It is applicable for both retrofit and green field sites and leaves the requirements of humidity control and filtration to the base cooling system.

Responding to equipment trends
In the future, when considering new or updating of CRAC units, data center managers must consider new approaches to cooling. Features that deliver higher reliability and efficiency and lower total cost of ownership include:

• Variable capacity: Cooling systems should operate effectively at varying loads. Units should have compressors that can step unloading or provide variable capacity without cycling on and off. This approach reduces starts and stops, which are a leading cause of compressor wear.

• CRAC unit-to-unit communication: Communication between system units also increases total cooling efficiency. This is particularly important in rooms with a few high-density loads, as some zones may be operating at a much higher temperature than others. Integrated communication keeps units from fighting each other, for example, by having some dehumidifying while others are humidifying. It also provides the ability to direct specific cooling to high-heat zones, thus increasing energy efficiency.

• Service organization availability: As heat load increases, the margin of error in a cooling system design becomes critical. A solid preventive maintenance program will keep equipment in top operating condition.

InfraStruXure fully integrates power, cooling and environmental management within a rack-optimized design. The InfraStruXure on-line BuildOut Tool automates the process of designing the optimum system. These customized systems are then built from standardized components, which are modular, manageable and pre-engineered to work together. This modular flexibility enables you to install only what is required today, yet [is] easily scale to meet future demands. InfraStruXure increases system availability, while decreasing design and install time from months to days, while eliminating the need for raised floors in many applications.

These racks are Blade Ready (a trademark of APC). To see them, you should check out their Blade Ready Seminars. They also have some excellent white papers and application notes on the subject. This is above and beyond the data center university I blogged about the other day.
Posted by Carrie Higbie Supporting blade technology -- APC
08 MAR 2006 15:06 EST (20:06, GMT)
APC has been very active in supporting blade-based servers on a variety of fronts.

APC's Data Center University offers industry-leading education for IT professionals, facilities managers, engineers and others involved with the Network Critical Physical Infrastructure (NCPI) of data centers. Our courses are developed by actual data center executives with decades of combined experience running some of the world's most advanced data centers. The courses address the core competencies required for Data Center Design, Build and Operations (DCDBO). Data Center University's curriculum centers around the education you need to solve real world issues in the data center. Courses are online and on demand, providing you with all the knowledge you need, when you need it and where you need it.

On another front, they have also started the BladeReady program. This program is designed to answer questions surrounding the adoption of blade servers and provides educational forums as well as a great online site with answers and support. In their words...

Server compaction, such as that experienced through Blade server adoption, can create power densities that exceed the power and cooling capacities of virtually all data centers. They also present unique challenges to the interoperability of your network-critical physical infrastructure (NCPI). That's why APC, along with trusted industry Partners, developed the BLADE READY Program. It unites complementary NCPI technologies to deliver comprehensive interoperable products, solutions and services for blade NCPI environments. To ease your blade implementation and enable operational excellence, look for industry vendors that carry the BLADE READY logo.

The choices, as I stated before, are plentiful. The decisions are not quite as clear. In data centers where existing storage is a factor, this may govern your choice for a chassis. In new systems -- well, they sky's the limit. Companies like QLogic and Emulex have introduced faster storage host bus adapters, and the storage switch manufacturers are providing fast and furious options for storage and retrieval. Probably the two most common platforms are fiber channel, iSCSI and Infiniband products, which run on copper (15m max -- twinax -- and you thought twinax was dead?). The new 10GBASE-T standard will provide a 10-gigabit option over copper.

So how do you pick? Well, first, you will want to decide if you want directly attached storage, enclosed storage or an attachment to a storage area network (SAN) or network-attached storage. To do this, think of your applications and which of them need to share data sources. If you use network-attached storage and you are limited to network speeds of 10/100, that will be your storage limit. If you use a host bus adapter that runs at, say, 4GB, this would be a better option for larger storage shares. I am not aware of any nifty little tool that maps out the best one for a variety of situations, but if you know of one, email me. I would love to share it.

What exactly do you want to do with your storage? If you want to have virtual PCs and provide boot from SAN, check out 3UP Systems and some of the other vendors that have made this a reality.

For backup and data protection check companies like Kashya and others that bring a variety of environments under single control. They have really great products. I recommend that everyone talk to more than one vendor. The big guy might not be the best, especially when you are dealing with multiple options on the same network. Sometimes the big guy may be the only guy. EMC has the corner on this market right now, with alliances with all the major blade manufacturers, but they may also be cost-prohibitive for some of the smaller networks. Don't assume that there are not other players in the markets -- there are. Don't blindly go with a vendor unless you have a strong allegiance to that vendor. In other words, don't go by the last article (or blog) you have read to make your decision. In reality, you may find that a variety of vendors are the best solution, or only one. In a single-vendor solution, it sure helps with troubleshooting and patch management and certainly in support. But there may be some areas where other options make sense to your organization.

Don't assume that fiber is the best when lots of copper options exist -- and make sure you understand any technology completely. If your sales guys is not getting you the answers, go for their technical assistance. If that doesn't work, escalate. It is your money and your data. After all, we really wouldn't need networks or servers if it wasn't for the data.
Posted by Carrie Higbie How to select a blade manufacturer
06 MAR 2006 17:38 EST (22:38, GMT)
Not all blades are alike between manufacturers and not all blade servers are alike even in a manufacturer's line. They vary from entry level to enterprise level and all have a variety of functionality, depending on what you want to do and what your server needs are. Your decisions are also going to vary depending on whether this is a new deployment or a retrofit situation.

One thing to check is backplane speed for the chassis and switching modules if they are used. This will determine how quickly the blades can talk to each other and what future speed increases the blades will support without a fork-lift upgrade.

The second thing you will want to determine is the management functions. This is critical. Blades should save on administration time, so look for the features and especially security features. Which functions can be tied to a user's rights?

Next check out what types of storage each supports and with which code. For instance, you may have one vendor that supports a certain storage system and another that would require custom code to do so. This can hinder performance and increase your total cost of ownership when patches come around. You may have to wait for patches to your applications if your storage is not supported. What speed is the storage interface? And does it run on copper or fiber (the cost will vary)?

Check the redundancy and failover options. Do you fail over to another card in the same chassis? Or can you mirror one chassis to another? This, of course, will depend on the level of redundancy you want to build in to your server infrastructure. Is it automatic or will it require manual intervention?

Look at the power loads for day two costs. You may even be changing out some power. A lot of the newer electronics are moving to 220 power because it is more efficient.

Examine the footprint of the server and look at how many blades can be contained in a chassis. What are your blade options? Can you add networking ports or storage ports? What is standard or base vs. what is optional?

Your cost comparisons should include all the features you want, as well as all the Day 2 costs, warranty particulars and fine print. This should include the servers and their environment, too. In a retrofit situation, don't use lower cable counts as a savings, as the cables are already run and that is not a real savings. Beware of stock ROI documents. While they do provide a guideline, you may not realize all of those savings; make sure that they are applicable to your environment.

My favorite advice of all: Check the press releases for people who purchased the technology or ask around local user groups. Call some of these folks and find out how smooth the installation went and how well they are performing now. This will help you evaluate the vendors and plan for any additional installation or follow up costs as part of your cost comparison.

Then -- install and have fun -- it's really great technology!
Posted by Carrie Higbie The blade server market
03 MAR 2006 16:16 EST (21:16, GMT)
We are all creatures of habit. One habit is to ride the knowledge of others where possible. Meaning: We like to learn from others' mistakes and successes before sticking our collective neck out. Since blades accounted for roughly 25% of server sales, I thought it might be helpful to learn about the blade market -- where things are today, where they are going and who is getting the lion's share of the market. According to IDC, in the third quarter of 2005, IBM garnered 42% of blade server revenue, with Hewlett-Packard second at 32%, and Dell in third with 9%. Each has a variety of chips that can be used. According to a recent article in CNET news:

HP's BladeSystem can use blades with Advanced Micro Devices' Opteron chip, Intel's Xeon or Intel's Itanium. Sun Microsystems' second-generation blade, due to ship later this year, will accommodate both Opteron and Sun's own UltraSparc. For IBM, the choices are Opteron, Xeon and IBM's Power processor family. For Power chip fans, IBM will begin offering a new choice of blade: the PowerPC 970MP-based JS21. This 970MP processor has dual-processing engines, compared with the single core of the PowerPC 970FX used in the current JS20 blade. The PowerPC 970 MP also includes a significant new virtualization feature that lets several operating systems run on a blade simultaneously. That's a major trend as computer users try to wring more use out of existing systems to cut costs and reduce growing electrical power demands.

One thing is for sure -- we are storing some stuff! Industry predictions say that there will be an average one terabyte of data stored for each human on earth. And to think -- I am tickled with my 40GB iPod! Monday's blog will cover storage options -- stay tuned!
Posted by Carrie Higbie Sizing power for blades and a data center
02 MAR 2006 13:17 EST (18:17, GMT)
One of the most misunderstood and overrated things in a data center is power. A common method is to add up all the wattage requirements from every device and assume that this is the proper sizing for a UPS. Think about this, though. Typically that power draw is at startup. When you fire up a machine, the disks spin, the processors kick in and the power supply spins. This is called instantaneous power. If your whole data center had to spin up after a power failure to run on the UPS then you would need this requirement. However, the idea of a UPS is to sustain a load until a generator can kick in. When you size for a blade server and you look at the highest voltage, you may be providing for overkill. If you have systems with redundant power supplies, they may perform load balancing or one will draw lower power until the unit fails. If you plan for full power on each supply, you will double your needs or at least overestimate them. I mention this because power equates to cooling and one mistake that happens often in data centers it that as technology changes, the only thing communicated to the facilities guys is that another circuit is needed.

It is a good idea to realistically size your UPS and cooling systems based on what you are really using. An ampmeter will give you a good idea of what you are really drawing out of a piece of equipment. "A large fraction of the energy consumed in data centers goes to waste," said Bob Sullivan, a data center design expert from the Uptime Institute. In a survey of 19 data centers, 1.4 kilowatts of power are wasted for every kilowatt of power consumed in computing activities, the research consultancy found. Consult with your manufacturer to be sure that you have the right kind of power as well. Much of the newer equipment is 220 3 phase, which really won't work on a 110 circuit. But in either event, power and cooling should not scare you off. Knowledge is power! (The kind that doesn't require cooling.)
Posted by Carrie Higbie Getting SMASHed -- hardware management regardless of vendor for blades, storage and servers alike
01 MAR 2006 21:27 EST (02:27, GMT)
Quoted from the DMTF (Distributed Management Task Force) SMASH forum Web site:

The SMASH Forum is dedicated to the interoperability of the DMTF's Systems Management Architecture for Server Hardware (SMASH) initiative specifications.

The SMASH initiative -- which includes the widely-implemented Server Management Command Line Protocol specification -- is a suite of specifications that deliver architectural semantics, industry standard protocols and profiles to unify the management of the data center. Through the development of conformance testing programs, the SMASH Forum will extend these capabilities by helping deliver additional compatibility in cross-platform servers.

With many end users embracing grid and utility computing initiatives, server management continues to play a key role in controlling costs in the data center. The purpose of the SMASH Forum is to ensure that consumers of SMASH-based server management technology -- both end users and the ISV community -- receive interoperable management technology for multi-vendor server systems. As the SMASH specifications become the ubiquitous standards for server management, the work of the SMASH Forum will provide users with additional assurances of interoperability in their cross-platform implementations.

Here is the info from DMTF; I encourage you to check here often for news on management-related topics.

With more than 3,500 active participants, the Distributed Management Task Force Inc. (DMTF) is the industry organization leading the development of management standards and the promotion of interoperability for enterprise and Internet environments. DMTF standards provide common management infrastructure components for instrumentation, control and communication in a platform-independent and technology neutral way. DMTF standards include the Common Information Model (CIM), communication/control protocols like Web-Based Enterprise Management (WBEM) and the Systems Management Architecture for Server Hardware (SMASH) initiative.

First and foremost: Get rid of the extra unused cabling under the floor. Many data centers that have been around for a long time have old bus and tag, twinax and point-to-point cabling connections that have been abandoned under the floor. This creates an air dam and the air can't move through a big wad of cable. If you don't have room for the air to flow, you might as well skip it. Next, most manufacturers have some great retrofit products, and companies like APC have cooling zone products like their InfrastruXure product line. Rittal has liquid-cooled cabinets, and Liebert makes some retrofit chill boosters to help with the cooling as well.

In reality, 10 years ago, data centers consisted primarily of a mainframe and maybe a few PC servers. This is far different than what is in a data center today. We have Cisco 6500 series switches or maybe Force10, Extreme or Foundry switches -- whatever the manufacturer, they are probably large chassis-based switches unless you are a smaller enterprise. While power was probably added, cooling may not have been. The best favor you can do yourself is to resize your power and cooling, remembering the one-to-one rule above.

You will also want to ensure that you have blanking panels in your cabinets where there is no equipment. This keeps you from creating an air plume and the air intake in the front of your equipment from pulling the heated air in from the rear of the cabinet. If you are planning an expansion, pre-cable more than you will need so that you can accommodate growth, but not have a bunch of haphazard cables running in areas you don't want them. Be sure to run your cables in the hot aisles where they won't obstruct the air flow. Consider putting your blade servers in a small section of the data center if that will allow you to provide additional cooling in a more cost-effective manner. If you have sufficient air flow and floor space, consider spreading the blade servers out so that they are not in one concentrated spot, as this will allow the natural air flow in the room to dissipate the heat.
Posted by Carrie Higbie Benefits of blade servers
27 FEB 2006 07:06 EST (12:06, GMT)
In selecting a topic for this run, I suggested blade servers. There are many benefits to blade servers and some of the older shortcomings have been addressed, making these platforms even stronger. The first benefit is in management. It is much easier (with the newer tools) to manage several servers with one connection than having to administer each separately. There is also new feature-rich software that will allow administration to occur based on permissions. For instance, you may have permission to one blade in the chassis and not the others. There is also a benefit that can be received by using blades-based PCs so that the thin clients that have been around for some time but have not been as successful will work. Companies spend a lot of money on desktops, the ability to finally realize a thin client architecture due to the increase in network bandwidth could be wonderful in the right situation.

Another advancement is in the storage systems that connect to the blade servers and their networks. Booting from a SAN is now a reality. Copper and fiber 10G gear now exists, allowing greater bandwidth than we have had in some time. You have either greater or at least equal processing power in a chassis or two than with all of those two RU servers. This means a smaller footprint in the data center for the same basic power. Granted it is hard to fill a rack with blades, but there are new chips in the works that would use less power, which will help with the heat factor. There are also liquid-cooled cabinets to assist. I would have to say, however, that many data centers are running at less than peak efficiency due to abandoned cable under the floor, which results in an air dam.

In this series, I will reach out to other members of the BladeSystems Alliance to find out what is new and what is cool. For more information on the member companies, go to BladeSystems.org. I will also reach out to our friends at DMTF.org for information on SMASH -- tune in tomorrow to learn what these acronyms mean. Also check out the other articles written about blades here on TechTarget. Feel free to ask any questions!
Posted by Carrie Higbie