Application Sensors and Intelligent Power Management Part #2

IPMPlus provides built-in application sensor for most commonly used applications. However if you have custom applications  you can define your own application sensors. To define the application sensor policy  you need to identify the applications which are most often used (know their process/executable names) and then determine if it is an interactive or background application. 

Desktop interactive applications will require the monitor to be turned on when the application/user is active. For example Powerpoint user on a laptop - while doing a presentation the monitor should not be turned off even if the user is not active on the keyboard. The disk can be turned off (not to worry, when ever the system needs to read or write data it will automatically turn the disk on). If the user is away from the desktop and PowerPoint is not running - the monitor can be turned off.  Here is the example for PowerPoint.  Imagine the situation if you didn't  have application sensors, if you enable power savings on Windows with the default power policies, while you are doing a presentation the monitor turns off if you are away from the keyboard for a while and interrupts your presentation and irritates your audience.

 Applications sensor enable fine-grain power management. For each application the administrator can specify which components of a system can be selectively turned off or not and finally whether the entire system can be moved to standby/shutdown or not. Multiple applications can be selected and the  fine-grain power control can be specified for each of the applications. The policy can be enabled or disabled for each of the applications. The following example show the application sensor definition for running virtual machines using VmWare Player.

Now let us look at  another scenario of a server which is used most of the time in the day and at night it can be moved to a standby mode when the batch job is completed. In the application sensor policy define the name of the batch job and specify that the system cannot be in standby/hibernate/shutdown state when the batch job process is running. After the batch job completes the power policies will move the system to the low power state.

Multiple application sensors can be defined simultaneously, each application sensor policy can be associated with one or more power policies. However at a given time - which ever power policy is enforced will select the corresponding application sensor policy. In the following diagram you can see that the Power Scheme is referring to the default Application Senor Policy.  You can edit the existing Application Sensor Policies or Create New ones.

To summarize Application Sensors provide a mechanism to enable fine-grain power savings while allowing the respective applications to function and enable the user to work uninterrupted.

What will you do with 1 TeraWattHour of Electric Energy?

DatacenterDynamics reports that in 2012 data centers across the globe have consumed 322 TeraWattHours of electric energy. [Even experts seem to be confused between units of Power and Energy. We'll stay puritan-istic in this regard and measure Electric Energy in Watt-Hours and Electric Power in Watts.]

If we assume that this is the energy in its aggregate form drawn from the grid, the sad news is only about 100 TWh would've really got used - the remaining having been wasted in various stages. [Refer Schneider's calculations in this regard on how energy is wasted in data centers.]

That being the point, I was just wondering how much is really 1 TeraWattHour of Electric Energy?

Firstly, what would it take to generate that much of electricity? Here are some options...

• Nuclear generation: Run one of the two 1,000 MegaWatt units of Kudankulam Nuclear Plant at full capacity for 1,000 hours = 42 days on a 24x7 basis
• Thermal generation: Burn 149x3 ~ 450 Kilotonnes of Coal [assuming 6.67 KWh/Kg of energy per Kg of coal and 30% as the thermodynamic efficiency of thermal power plants] 
• Solar generation: Roughly 220 Sq.Km of Solar Panels [assuming 70 MilliWatts/Sq.Inch]

What can I do with it if I weren't running data centers? Well, to begin with, I can burn a tubelight or run a laptop at full power for 2.8 million years on a 24x7x365 basis. In other words, 5.6 Million houses can have at least 1 tubelight burning throughout the year for 12 hours a day - enough for 10 Million children to study if we assume each house has 2 children !

We invented the Internet and then Mobile telephony. Somewhere in between the bankers said - "High Availability and Zero Data Loss". Haven't we made the world a little too complicated in trying to make it a smarter planet? Now you know why few of us are spending sleepless nights over reducing power consumption of IT infrastructure. People today question us on return on investment when we offer IPMPlus at highly affordable prices - tomorrow there may be no choice :)

A case for regulating mobile device power consumption

We all love our smartphones, don't we? The more the merrier. Retina display, HD videos, GPS, Gaming - these are good, no doubt. But do we think what will happen when these devices proliferate in large numbers in an energy starved economy? In little drops we are draining away a mighty ocean.

India has 900 million mobile devices (roughly). For ease of calculation, assume each consumes 1 Watt-Hour of electricity a day. The more the talk time, the more the games and videos - more is the power consumed. This means every day we are spending 900 Mega-Watt-Hour of energy by using mobile phones. When we consider that we consume 2 Watts in effect instead of 1 Watt due to transmission losses and poor mobile designs, this is an alarming 1,800 MWh of energy consumed a day - roughly equal to running both units of Kudankulam nuclear plant for 1 hour at near full capacity.

Most device makers have very limited understanding of software-hardware interaction. They simply assemble and give you products to make quick bucks. The average margins in the market are between 30-50%. So why bother about power optimization?

In next 3-5 years as we move towards LTE/4G, power consumption in these little devices is going to increase dramatically from the current base. Device makers are likely to trade-off performance to power optimization. Hence it is essential for all of us to wake up to the issue of power wastage in mobile devices and look for better ways of power management. At Vigyanlabs, this is what we are working on. Try out our first Android offering - IPMPlus Touch2Sleep Beta.

Application Sensors and Intelligent Power Management Part #1

What aids powers Intelligent Power Management is the concept of Application Sensors. Just like we have temperature, light, proximity sensors which sense and aid the instruments in better control, an equivalent component in the IT Infrastructure is the  Application Sensor. Application Sensors sense which applications are running and what components of the hardware they use and provide the intelligence to selectively turn on or turn off hardware components in a system. Multiple applications can be running on a given system, however what is important is the critical list of applications. Let us consider the case of a Laptop, while doing a power point presentation (or any other presentation application) - the display output (rather the external projector/display) has to be on, whereas other components like disks can be powered off if they are not in use.  When the presentation is completed, after power point is closed, the display output can be turned off if the user is not using the system. Consider the conventional power management where we tend to disable the power management entirely while doing a presentation.  Similarly if a background application is doing a file download or transfer, while the user is away from the desktop, the display can be turned off, cpu can be in a lower power mode. If the download completes and the user is still away from the desktop, the system can be allowed to go to either standby or hibernate thus saving power.

IPMPlus provides few built in application sensors for applications which do download, file copy, presentation (power point). In addition user can define his/her own application sensors and selectively turn on/off hardware components and thus save power.  To summarize Application Sensors provide the intelligence to selectively turn-on or turn-off hardware components of a system while the system is in use and helps us to save power without interrupting the users work.  In the next article I will talk about how to define application sensors.

The concept of Intelligent Power Management

Power management has been an ubiquitous feature in all aspects of IT infrastructure for a long time. If so, what is new about Intelligent Power Management? Read on to understand the subtle differences.

In the traditional approach to power management, power optimizations are generic in nature. The  optimizations are not based on understanding the application-OS-hardware interactions. Hence these optimizations aren't always the best.

Power management in the traditional approach relies on time-based policies - even so, no solution exists for enterprise-wide administration of such policies. IPM as defined by Vigyanlabs is not time-based. Instead it is based on application sensors - a technology patented by us. The power management policies are based on applications and are customizable as per the application needs. A good IPM platform allows such application-sensor based power schemes to be administered in an enterprise-wide manner.

Traditional power management approaches are device specific and consequently, not scalable. A truly intelligent power management platform is scalable and should be implementable across a wide spectrum of devices and operating systems in varied operating environments.


Traditional approaches are intrusive. They either trade-off user experience (e.g.: your laptop goes into standby when you are presenting) or performance In our IPMPlus approach, we ensure non-intrusiveness and do not trade-off power management against either performance or user experience.

Finally, intelligent power management is about quantifiable power savings which is not possible in traditional power management approaches.