In many data centers, electricity now represents as much as half of operating expenses. Data centers are not using their resources effectively. The resultant power unleashed if operations were optimized could amount almost to a new data center that had been "hidden". The average power consumption per server rack has doubled in the past three years.
Over the next five years, 90% of businesses will see their data center options halted by power failures or limits on power availability. A third of companies are planning projects in the next 12 months...roughly 50% were planning (data center) expansions in the following 12 months. Demand for data center services will increase at 13% a year. You've heard the hype, we're ready to take action
Do we spend a lot of money on exotic solutions? Or do we step back and take a more practical approach? Here are some tips to help you approach this complicated problem and ensure the greatest potential for success.
The Outline for a Practical Approach
1.
* Where it's being used in the data center
* How effectively your power distribution and cooling systems are being used
2.
* Plot the historical consumption levels
* Then overlay plans for new services and their impact on consumption
3.
* Isolate the "Hot Spots" and re-distribute the load optimize the cooling system
* Balance the power load across the power distribution system
4.
* Now that you know where you are, you can plot a course forward
* Compare and contrast relevant technologies and solutions
* Build a business case!
Selecting the Right Tool for the Job
* You could do this on spreadsheets but, you might grow old doing it
* Better to use a solution that will automate as much of the process as possible
* What are the requirements for such a solution?
Where's the Data?
Centralized data storage & controlled access
* It should put all of the data in a relational database such as SQL
* Once data is entered into the database and future changes should be changed controlled
* It must have role based security
* Read only access should be provided through a thin client, i.e. web browser
The goal is to make the information:
* As accessible as possible, while maintaining it's accuracy
* Create and maintain relationships between the data that will allow it's modeling and analysis
Visualization
The ideal solution should provide visual access to the dataset Data that can not be easily understood or accessed has little value Get the big picture and then drill through to detail
* A top level view should be provided that summarizes all of your datacenter assets
* That view should provide configurable summary and status information
* The user should be able to drill through to more detail
What views should it afford?
* A geographical view that shows me all of my locations
* Location views that display top down views of my data centers
* Rack level views of front and rear elevations
* Device views that display consumption information
Customizable Overlays and Status
* The solution must be able to visually represent user defined thresholds
* Display coding by business service, department, or customer will also be extremely useful
Documentation
The system must document the data center physical infrastructure:
* Equipment layout, location, and placement
* Equipment Consumption of space, power, heat generation, and weight
* Power distribution system
Relationships between all elements of the data center must be tied back to business services, departments, end users of IT services, or customers
* Devices, power connections, applications must be identifiable by the end IT service they are associated with
* Each element should be able to be associated with multiple IT services
* Current contact information for the end user of the IT service should be maintained
* Warranty, maintenance, service contract, service life, information should be maintained
Modeling and Analysis
Basic Modeling
* Where is power be consumed and heat generated by (Location, Data center or room, Row, rack, or device, etc)
Advanced Modeling
* Adding or removing devices debits or credits the power system for available power (Circuits, PDU, UPS, Power Source)
* Analyzes which devices are redundantly powered
* PDU, Circuit and Equipment Levels
* Simulation of failing of power infrastructure components: Determines effected devices, applications, business services
Historical and Trend Reporting
* Must be able to produce reports that graph the consumption trends in the data center
* Should be able forecast future consumption based on past consumption and expected expansion existing services and addition of new services
Management
Now that we know where we've been and where we're likely to go we can analyze how new technologies might help us
It's about the business case
* The solution must be able to use the existing infrastructure model as a base line to compare/contrast against a proposed solution
* Example: If we replace the legacy servers in the data center with blade servers
* Basic Analysis:
How much energy is saved?, Does heat generation go up or down?
Is heat concentrated in specific locations vs. being evenly distributed?
Is space saved?
* Advanced Analysis:
Economic impact
Analysis of the relationship between CAPEX and OPEX
Return on Investment (ROI)
Business case and purchase justification
To save the world and be green you must first:
1. Know at a detailed level where you are
2. Optimize what you have
3. Look where you've been so you can determine where you are going
4. Consider likely growth in determine the future
5. Build a business case to justify the green plan
The boss loves saving the world As long as it makes him money
follow us on twitter (Rackwise)
[http://www.rackwise.com]
