There’s some interesting information floating around about HP’s Virtual Connect.  They say that it simplifies server management and eliminates barriers to change.   Wow, that’s a stretch! The Virtual Connect switches still need provisioned uplinks. Perhaps a better idea is Dell’s FlexIO (with FlexAddress), which enables automated deployment configuration in advance, network stacking (making a number of switches appear as a single virtual switch), supports industry leading Cisco switches that can be managed using existing Cisco network management frameworks and tools, and of course it all comes from Dell as a single integrated solution.

Also remember that although HP tries to position VirtualConnect Ethernet products as not being an Ethernet switch, they truly are. In fact, the heart of the HP VirtualConnect Ethernet switch is the same family of Broadcom switch ASIC used in the Dell PowerConnect M6220 (actually HP uses a lower end version with fewer 10GbE ports). They put some software on top to hide that, but at the core it’s a commodity switch. So they can’t be managed by network tools in a Cisco network infrastructure: they must be managed separately using HP’s proprietary management interface. In fact, to manage more than one chassis of VirtualConnect switches, HP requires you to buy VirtualConnect Enterprise Manager at $9,000 per chassis!   Dell Cisco blade switches are managed like any other Cisco switch in a network infrastructure, which we think is easier.

HP promotes persistent WWN/MAC with VirtualConnect, but Dell’s implementation (FlexAddress) doesn’t require costly proprietary switches.  In fact, because Dell’s FlexAddress is enabled through the Chassis Management Controller it can be used to deliver the advantage of persistent WWN/MAC with any IO module, including Cisco, Brocade, PowerConnect, and even pass-throughs.  It’s not only more flexible, but also a much simpler and lower cost way to deliver this important functionality.  Finally, to get persistent WWN functionality with VirtualConnect, HP requires customers to buy BOTH a VirtualConnect Ethernet AND VirtualConnect Fibre Channel switch. So, if you don’t want to go down the proprietary switch route for Ethernet, HP has no way to deliver persistent WWNs for Fibre Channel.

Dell’s FlexIO technology is fully compatible with a Cisco infrastructure and enables Cisco stacking functionality (known as Virtual Blade Switch).  This reduces cabling, internal cross charges for networking, and the use of distribution layer ports. It also allows the switches internal to the blade enclosure to be managed using the same knowledge and tools as customers use to manage the rest of their network environment. So who is working to simplify IT?  It isn’t HP.  If you disagree, let me know.

It is kind of funny when I talk to customers about blade servers.  They are really jazzed about packing more servers in the same space.  But with blades comes other issues — managing power and cooling. Denser environments usually mean more power and more cooling, rather than less. So are these ideas – blades and managing P&C – mutually exclusive?  Aaron Hanson, senior manager PowerEdge Blade servers at Dell, doesn’t think so, as he says in his podcast.  Is he right?

What’s up with the P?

What’s up with the P?

Wed. Apr. 23, 2008

by Jeff Johnson, Enterprise Strategist

Technorati Tags: HPC,Dell,Servers,clusters

What I am learning from places like InsideHPC.com and from the guys at Dell is that HPC is rapidly scaling from supercomputer centers down to much smaller organizations. It seems a more apt term than High Performance Computing – and one that is being embraced for smaller class of clustered systems is High Productivity Computing.

Bringing the huge compute power and expertise of large scale sites to smaller packaging, easy to deploy systems is a common enough story, but it’s still an exciting thing to witness.  This is how HPC technology is going to become more useful to many more customers:  by reducing the costs and complexity of the technology and by making it easier to manage and use. 

I spoke to Tim Carroll, Senior Manager HPCC at Dell about what’s happening. What I found is that HPC and clustered systems can be affordable and manageable – even for small organizations. Cool stuff. Listen here and share.

http://direct2dell.com/insideit/archive/2008/04/23/what-s-up-with-the-p.aspx

I’ve been ranting about the complexity of IT and the need to simplify it.  And when most organizations talk about it, they expect the technology industry – companies that make servers, software, services, and devices – to do something about it.  Dell has gone on record as saying we are going to focus our efforts, resources and process-driven approach to do just that.  What doesn’t get as much chat, however, is the staffing side of the equation.  In a speech recently at the Gartner Data Center Conference, Jim Soderbery, senior vice president, Storage Foundations Group at Symantec talked about what seems to be an accident waiting to happen.  As IT gets more complex, the ability for IT managers to recruit, train and retain expert talent is getting much harder. 

After the speech I checked out the report “State of the Data Center Research report” In the study by Symantec, datacenter managers said:

n  52 percent of respondents report their data centers are currently understaffed

n  86 percent of respondents have difficulty finding qualified applicants

n  68 percent report staffing is challenging because data centers are too complex to manage

n  60 percent believe staff skill sets are too narrow

n  57 percent indicate that employees’ skills do not match their current needs

What does this mean?  It means that unless IT is simplified, organizations are headed for a train wreck. Fewer and less experienced people are being asked to manage increasingly complex IT.  The question you have to ask is whether your IT infrastructure is headed for the same disaster.  Are you seeing it coming?  I’d be interested in hearing what your organization is planning to do about it.

Data centers are massive users of power.   A single data center can use as much power as a mid-sized town.  So anything that can be done to reduce power use can have a huge impact on the environment.  Gartner estimates that power consumption by computers accounts for 2 percent of global carbon dioxide (CO2) emissions. That’s roughly equal to the carbon output of the entire airline industry.  The words “sustainable” and “data center” usually don’t travel in the same circles.  Irrespective of the desire to be more sustainable, data center managers focus on three primary goals:  Meet the computing needs of the organization, keep it reliable, and meet budget parameters. Being greener is a nice side benefit, but often not the primary goal.  Building a more sustainable or “green” IT has more to do with data center management and strategy than with specific “greener” products, although they coexist in the same conversation. Purchasing more energy-efficient equipment is certainly part of the equation.  But managing the power consumption of that equipment can have the greatest impact in sustainability.  Data center sustainability is really the convergence of power use (number of watts or efficiency), performance (amount of compute per watt or effectiveness), and product lifecycle management (how often you buy and discard).  What complicates this picture is that there are a lot of technologies that all contribute to the sustainability picture:  power-efficient servers and cooling, data center design, virtualization, and others. 

The easiest thing to do is to focus on using less power, but there are lots of ways to do this.  You can start with the equipment.  As an example, Dell M1000e Energy Smart servers consume 19% less power yet generate 25% more performance per watt than competitors.  They can save 3,200 watts per rack per year, which saves more than $30,000. More importantly they don’t generate 18 tons of CO2 emissions, which equal about four acres of pine forest.  You can meet the increasing demands of the organization and still use less power.  Also make sure your power supplies are at 80% efficiency, rather than the 60% – 70% efficiency we commonly see. Processors with greater power efficiency are meaningless if you’re losing efficiency before the power gets to them.  Consider changing the way you cool your data center. Cooling gobbles up 40% – 50% of the power in a data center.  An increase of five degrees can lower power use by 5%.  New and different kinds of chillers are significantly more energy efficient because they do a better job of providing reasonable air inlet temperatures to the servers and disposing of hot air.  In the end, it is the temperature of the air going into the servers, not the temperature of the air leaving them that is more important.  Even then, the cold-aisle doesn’t have to be a meat locker.  With reasonable inlet temperatures and hot exhaust temperatures, server fans don’t run unnecessarily and CRACs don’t work as hard – energy is saved in both cases and the servers are still perfectly happy.  Of course, you can’t forget desktops and notebook computers, especially Energy Star certified ones.  There are configurations today that consume 70% less power, through a combination of lower power components and aggressive power management tools. According to the EPA, if all businesses were to purchase only Energy Star-certified equipment, they would save $1.2 billion over the life of the computers. Less energy waste means greater sustainability.

 Also make sure you don’t leave performance on the table.  It is widely known that servers in most data centers run at 5% – 15% utilization.  A server uses 60% of its maximum power sitting idle.   But studies show that increasing utilization to 50% can only increase power use by less than 5%.  The best way to increase utilization of course is though virtualization.   Virtualizing to make one machine work as 15 or 20 does two things for sustainability and accomplishes one major goal:  it more efficiently uses power, it reduces the need to buy more servers, and it helps meet the increasing computing demands of the organization.  Virtualization can also extend the life of your current hardware, because you are keeping up with demand and using what you have in a more effective and efficient way.Even an efficiently run data center will hit the wall sooner or later.  And that wall could come in the form of a maximum power envelope, a space limitation, or simply the end of the lifecycle of a piece of equipment.  At this point you have the opportunity to design for sustainability rather than simply retrofit pieces of what you already have.  Here is where you can think about the issue more holistically, rather than piecemeal.  One study shows you can get a 97% increase in productivity if you combine energy-efficient servers, virtualization, different kinds of cooling and some basic improvements in data center layout to improve air flow.   

Using less power, buy less, and using things longer means using less resources, in manufacturing, shipping, operation and disposal.  Using less resources means a more sustainable environment – but more importantly, a more sustainable IT.   I’m looking for agreement or disagreement with these ideas.  Better yet, we’re looking for your ideas, either below or on IdeaStorm.  Let us know what you think and don’t forget to visit www.Regeneration.com.

OK, it’s not a smackdown like on TV, but it is a power struggle, so to speak. Data centers consume a lot of power in companies. In one corner you have the IT department who is tasked with keeping and meeting the IT needs of the organization, scaling and staying within what is usually a flat budget envelope. In the other corner you have the facilities manager that is trying to manage overall costs, power use, space utilization, and the needs of the organization as a whole. A couple years ago the two rarely needed to talk, except if the data center manager needed more space.

The new smackdown is about power. Not political power but electricity. Globally, electricity prices have risen nearly 60% in the past couple years. Gartner says that 50% of data centers will lack sufficient power by 2008.

IT managers haven’t really had to care too much about power consumption for a couple of reasons. Rarely has power used the data center or any other technology been isolated, so it hasn’t been tracked or measured. The primary focus has been on eliminating business disruption, building IT agility, and meeting business needs. And for most managers, 70% of their budget is spent on maintenance and managing an aging, heterogeneous infrastructure. It’s all about performance.

Facilities managers worry about the physical side of the equation: power costs, getting more compute power out of the same space, and cooling. They wonder where they will get the space, if the utility is going to be able to deliver enough power, if they have enough breakers for new equipment, exceeding rack capacity, and a lot more. For some, the only way to add a new server is to take one out.

So the two sides are forced to talk. The IT manager needs performance. The facilities manager needs control. And what happens if they don’t? Inefficiency, more complexity, and the problems will certainly get worse. But if they do collaborate great things can happen. Here are some examples of what is available today:

• Client desktop systems that use 70% less energy.

• Energy efficient servers that use up to 25% less energy, but deliver the same performance – saving up to $200 per year per server.

• Increasing compute capacity by being able to put five servers in the same power envelope as four.

• Spot cooling solutions that focus on hot spots – saving 30% or more –not on inefficiently cooling the entire room.

• Holistic solutions that reduce energy consumption by up to 75% with the same power usage, or twice the performance in the same space, or twice the number of servers in the same power envelope.

Are there reasons for the two sides to talk? Absolutely. Can the two sides, with the right outside partner, lower costs, increase performance, manage space and meet each others’ needs? Absolutely. Some companies have done a lot of work to simplify this process. But what I’m more interested in are your stories. How is the relationship between facilities and IT? How do you break down the barriers?

http://climatesaverscomputing.blogspot.com/feeds/posts/default

Photo of me at 6am