The Data Center is the Computer: Step 5 of 9

Step 5 of my 9 step process:
Step 1: Measure
Step 2: Develop an action plan
Step 3: Prioritize Opportunities
Step 4: Create "Program of Improvement" with ROI and timeline
Step 5: Obtain approval for your initial project (from the Program of Improvement plan)
Step 6: Implementation Strategy
Step 7: Implementation Partners
Step 8: Communication
Step 9: On-going Operational Improvement based on a "Total Improvement Plan"

Obtaining Approval:
If you've done most of what's required in steps 1 - 4, the approval process should be a snap. The Executive Summary of your work to-date should include the following information

• Need & Vision: What were the driving factors behind this project and what's the short and long term vision post project completion.
• Benefit: What will the benefit of a successful project be for the business? It's very important to spell the benefits out in business terms; how will this reduce the bottom line costs of IT, enable high availability and improved Disaster Avoidance and Recovery Plans (DARP), provide for rapid system deployment and allow for IT to report on the availability of "Data Center" capacity as opposed to individual application infrastructure capacity.
• Risks: Risks of project failure and the risks if the project isn't approved. Be realistic here, most execs will applaud your honesty.
• Timeline and Resources: Be very clear and detailed with what you think the timeline is (in bite sized chunks) and about the resources (People, money, etc) that you'll need.

Don't be afraid to show the proposal to more people. If it can't stand the light of day, then you're probably doing something wrong. Be prepared to backup and justify the project with your personal conviction. In the end, most execs want to know how you really feel about the project and whether you're passionate about it. Lastly, don't be surprised if you have to repeat yourself. It's quite common to have to present projects 3 or 4 times to the same people before they really begin to understand and buyoff on it (they actually start evangelizing it themselves as their idea).

until next week....

Step 4 of my 9 step process:
Step 1: Measure
Step 2: Develop an action plan
Step 3: Prioritize Opportunities
Step 4: Create "Program of Improvement" with ROI and timeline
Step 5: Obtain approval for your initial project (from the Program of Improvement plan)
Step 6: Implementation Strategy
Step 7: Implementation Partners
Step 8: Communication
Step 9: On-going Operational Improvement based on a "Total Improvement Plan"

Now that you've prioritized your opportunities you need to create a "Program of Improvement". This process can be something as simple as a high level project plan to something much more comprehensive. My recommendation is to build your plan based on a combination of factors;

Complexity -

Are there geographic issues, how many partners do you have to work with, are there corporate business related drivers that need to be accommodated (product release or end of quarter, etc.)

Cost vs. ROI -

How much do you have to spend in time and money to recover how much? Building out a solid proposal for action will also help ensure you have the before and after picture covered. The before and after is critical to getting the credit and recognition your team deserves.

Team size and skill sets -

If you have a small team and relatively simple systems architecture than you can probably get away with a simple plan that covers primary milestones and critical path issues. Don't forget that "investigation/planning" and "communication" are the most important parts of any good project plan. If you have a larger team and systems with multiple external dependencies then you really should put together a detailed plan that accommodates the majority of potential risks.

Recognition and rewards -

Plan for breaks in the action. It's very important to celebrate your successes before moving on to the next opportunity or challenge.

Communicate, Communicate, Communicate -

You can't over communicate your plans. The more people understand about what you're doing the more support and understanding you'll get in return.

Until next week >>>

Step 3: Prioritizing Opportunities

I'm on Step 3 of my 9 step process:
Step 1: Measure
Step 2: Develop an action plan
Step 3: Prioritize Opportunities
Step 4: Create "Program of Improvement" with ROI and timeline
Step 5: Obtain approval for your initial project (from the Program of Improvement plan)
Step 6: Implementation Strategy
Step 7: Implementation Partners
Step 8: Communication
Step 9: On-going Operational Improvement based on a "Total Improvement Plan"

Now that you've got an action plan and you're ready to start improving your infrastructure and saving money of the bottom line, how do you decide where to start. The first key is that regardless of what you decide to work on first, the end solution should fit into your larger solution framework. In other words, don't change out your server platform without determining what is appropriate for your entire enterprise once the entire project is complete.

Other considerations you should make before getting started:

• What's the initial investment cost vs time to implement and end result benefit?
• What's the complexity of the activity (will staff need training or will lots of consulting be needed) vs. the time span and ROI?
• What fix will help make subsequent fixes easier to tackle?
• Validate your technology strategy and know how all the parts will fit in your final architecture

Now it's time to start picking the low hanging fruit:

• Server consolidation will probably be the first and best option. You can make the changes will little to no impact to production environments and it's fairly low cost as far as investments go. Also, when you've made some headway in the consolidation of your servers, you might find that working on your other areas of opportunity just got easier.
• Example: If you have inefficient cooling it's very difficult to make serious changes to your cooling strategy while your Data Center is running at or near full capacity. If you've reduced your server count by 50% or more the work will become much easier.
• This process is fairly complex so don't be afraid to bring in someone with infrastructure virtualization experience. The benefits you're looking for from virtualization won't be realized if you don't have the appropriate network and storage infrastructure to support it.
• Depending on whether you're working in existing DC space or have the luxury of building into a new space your opportunities and ability to work on them will vary. The benefit of a new facility is that you can build the appropriate infrastructure from ground up without having to worry about customer impact.
• Additional potential opportunities:
• Dramatically reduced power consumption through airflow management of your DC space. This could be something as simple as plastic curtains that seperate hot and cold.
• Determine the efficiency of old electrical gear some transformers could be costing you 5%+ through inefficiency. This inefficiency directly translates into additional heat that you have to extract from the room. UPS units are another great place to look. If you have high power costs you might be able to justify UPS replacements on the power savings alone.
• Reduce your network footprint and the associated infrastructure of cable and cable management.
• The list goes on.

As mentioned in "Step 1" be sure to capture all the information you can about your current environment before embarking on the change. Having measurements of your space, power, cooling, people & hardware (networking, storage & servers) overhead before your project starts and after it's complete is crucial to getting the recognition your team will deserve. You might even be able to get Public Relations to commit some resource to helping document the transformation. Remember to include your local power company as well, they will probably be more than happy to provide your company a rebate for any power savings you can demonstrate as a result of your project.

Until next week. Happy computing.

Mark

Step 2: Develop an Action Plan

Please don't consider this an exhaustive playbook, but rather a high level "strategy"

Developing an Action Plan is critical to defining your goals and objectives for the "Data Center is the Computer" project.

The Action Plan should include an overview of what you expect the end state or vision to be (What does done look like). While I'm not a big believer in being "finished", it is always wise to have a target that is measurable in it's completion.

Ensure you've captured baseline information (previous post) that will allow you to report and market your improvements.

Create a team - The data center and it's systems aren't generally owned by any one person, you're going to need the help of the following folks;

- Application managers

- CIO

- CFO

- Facilities

- Green or Sustainability team

There are unique ways of gaining each groups support, you can't use the same message with the CFO that you used to win over the application team.

Each of these groups has a stake in the current environment and will either put a drag on your project or facilitate it moving forward, but it will all depend on how you get them involved.

Review your environment for opportunities - My first recommendation here is to run something like VMware's Capacity Planner against your environment to obtain up-to-date utilization of servers. This utilization can help drive your choices for consolidation and virtualization and the information can be an argument winner.

Review your network infrastructure - How will your future environment be supported by the current network. In the long run even a fairly minor improvement in server utilization rates will overcome any short term costs for network upgrades. You should consider the technologies that your team is most suited to support and or have the most flexibility for adapting to a "Virtualized" environment. At a high level you should be considering 4GB or better connectivity so that you can get the most value out of a network consolidation and improve the performance of consolidated infrastructure.

Review your storage infrastructure - If you don't have a SAN solution in place, now is the time to consider one. The benefits of having a SAN back end for your virtualized environment are overwhelming. You don't have to do Fiber Channel, but you really do need to have the SAN.

The strategy here is not a simple one, but if approached correctly it will create a great starting point on your road to Nirvana. To me Nirvana is when your data center is truly managed and run like one large computer with lots of mostly commoditized components.
- Your front end is expandable/retractable processor and memory capability that can be measure as a total, not by individual servers or applications
- Your middle is a single network that handles basic ethernet and FCoE or iSCSI
- Your back end is a virtualized multi-tier storage environment that allows you to move servers from one location to another dynamically and to allocate space based on the total available, not on some pre-determined value. Your storage environment should also allow for the remote replication of data and largely tapeless backups.

Steps in getting closer to "The Data Center is the Computer"

Step 1: Measure
Step 2: Develop an action plan
Step 3: Prioritize Opportunities
Step 4: Create "Program of Improvement" with ROI and timeline
Step 5: Obtain approval for your initial project (from the Program of Improvement plan)
Step 6: Implementation Strategy
Step 7: Implementation Partners
Step 8: Communication
Step 9: On-going Operational Improvement based on a "Total Improvement Plan"

Step 1: Measure

Develop a baseline of what you’re using today:
- Measure total power delivered to the Data Center
- Divide total power by the power used by your computer infrastructure to come up with your current PUE (Power Utilization Efficiency)
- This initial PUE baseline will be what you measure any design improvements against

- Get your local power company involved.
- Many power companies are offering rebates and incentives for changes or upgrades that result in the reduced use of power
- It’s important to get the power company involved before your kickoff any project

- Meet with your corporate Green Team and or Sustainability Manager
- If your company doesn’t have a Green Team or Sustainability Manager it might be an opportunity for you to help get the team or role started

- Run a measuring/monitoring tool against your infrastructure to determine how your hardware is being utilized
- A good tool I’ve used is VMware’s Capacity Planner. The main reason I like it is because it doesn’t require an agent. However, the output in generates is extensive and extremely informative. You can obtain complete inventory of your environment all the way down to the number of Emulex cards or dual core Opterons you have
- Regardless of which tool you decide to use, the information you gather will give you specific opportunities to shut systems down, improve sharing and identify opportunities for further/new virtualization

It will surprise and amaze you when you realize that before buying anything you’re able to shutoff or otherwise repurpose 10-20% of your existing gear.

Data Center Efficiency - We're still just scratching the surface!

I hope my title is correct, because I love the opportunity that technology brings to the world, and I'd hate to think that at some point we might have to decide between a safe planet and more data centers.

What is a data center? Based on the following facts It's really a computer;
- A shell like a PC case is the building that houses your systems and protects them from the elements
- Power (much more of it) coming in from the utility, just like the plug you put into the wall for your PC at home
- Large HVAC systems protect a data center from getting overheated, just like your fan on your PC
- There's storage in the data center, there's memory in a data center and there are processors and applications there, just like in your PC.....

So what's different?

On your PC you have a group of applications sharing an environment and controlled through a single interface and a minimal number of input devices. The data center has many separate devices that often have just one application running.

We need to make the computing resource in a data center more like a single large computer with replaceable components. The opportunities associated with having a "Computer" for a data center (I coined the phrase "The Data Center is the Computer") are so numerous I'll have to create several posts to try and do them justice. Here's a short taste of what I'm thinking.

- Data Center managers and their managers should be able to say "I've got XX.X percentage of DC capability left" in response to a question from the CFO on current capability.
- When the DC manager is asked to add a new application to the environment his next step would be to determine the performance requirements of the application. He would then compare the requirements against his available data center capacity. If he determined that the new application would take him over his threshold of "available" capacity then he would simply buy the appropriate amount of disk, storage or processing required.


The above model is very different from today where a data center manager would have to look for a specific architecture of hardware and software to install the new application. He would also have to determine what extra equipment he needed to buy to accommodate additional “SAN” or “Core” network capacity and extra servers for “test & dev”. We must get to the point where the data center is truly converted to a large resource of "brain" (processing), "heart" (1 network) & "stomach" (storage).

The benefits of the "Data Center is a Computer" model are numerous. Lower cost of ownership (you reduce the overhead associated with running many silos of servers that are supporting single applications) and you make the resources you have optimized for the amount of compute power you actually will use.

I’ll write more on the opportunities and provide some thoughts on how you can get to this “Nirvana” in upcoming posts.

Bridging the Gap between DC Operations & Facilities

Data Centers always seem to be what I come back to. I guess I have a strong opinion about how a resource as big and expensive as a data center should be handled.

What do I mean by "Bridging the Gap?"

Recently more and more Bloggers & Analysts have started jumping on the bandwaggon of the need to develop staffing roles that fill the gap between what the average Data Center Operations manager knows and what a Facilities person knows. I've been worried about this problem for some time. In October last year I was able to develop and justify a new position "Global DC & Lab Space Efficiency Manager". Funny how these things work, but within days of hiring this person no one could figure out how we had managed without it.

We're several months into the project to build a new Data Center and this new role has been crucial to filling the void of information between the traditional IT & Facilities roles. Following are some of the areas that have proven this Gap theory;
- Carbon Emissions measurement
- Outside Air or Water Economizing for HVAC systems
- Finding efficiency improvements in the electrical system
- Identifying current DC standards initiatives in Europe & the America's (EU DC Standards & Green Grid are examples)

Any or all of the above are things that might normally be missed or ignored in a Data Center project. Yet, when you're building a facility that will eventually have 30-40 MW of power and 20,000 plus systems any small improvement means real dollars and a greener Data Center.

I'll be speaking about these issues in several upcoming events.
- Uptime Institute (Orlando April 27th - 30th)
- Emerson DC User Group (Tampa May 1st)

I'd love to hear your thoughts on this issue. I will surely be writting more in the near future.