SchedUling on heterogeneous Mobile Multicores based on quality of ExpeRience

Lead Research Organisation: University of Edinburgh
Department Name: Sch of Informatics


Users want mobile devices that appear fast and responsive, but at the same time have long lasting batteries and do not overheat. Achieving both of these at once is difficult. The workloads employed to evaluate mobile optimisations are rarely representative of real mobile applications and are oblivious to user perception, focussing only on performance. As a result hardware and software designers' decisions do not respect the user's Quality of Experience (QoE). The device either runs faster than necessary for optimal QoE, wasting energy, or the device runs too slowly, spoiling QoE. SUMMER will develop the first framework to record, replay, and analyse mobile workloads that represent and measure real user experience. Our work will expose for the first time the real Pareto trade-off between the user's QoE and energy consumption. The results of this project will permit others, from computer architects up to library developers, to make their design decisions with QoE as their optimisation target. To show the power of this new approach, we will design the first energy efficient operating system scheduler for heterogeneous mobile processors which takes QoE into account. With heterogeneous mobile processors just now entering the market, a scheduler able to use them optimally is urgently needed. We expect our scheduler to be at least 50% more energy efficient on average than the standard Linux scheduler on an ARM BIG.LITTLE system.

Planned Impact

According to Gartner the global smartphone market is worth more than £400 billion annually in end user spending, with more than 1.3 billion smartphones sold every year. Battery life and performance are critical factors to end users, who want more powerful, smarter devices that last longer than a few hours. Providing tools to support and enable research towards a better trade-off between performance and energy consumption will have a wide economic and societal impact.

We expect the impact of our work to come directly or indirectly from our project outputs. SUMMER will have these outputs:

1. Open source framework for mobile workload recording, replay, and analysis; 2. Open source prerecorded workloads;
3. Open source interaction lag prediction heuristics;
4. Open source heterogeneous scheduler based on user experience.

The following is a list of the major impacts we want our work to achieve:

1. A change in the way optimisations are evaluated. We want design decisions, from compiler and operating systems heuristics, even down to choice of heterogeneity in multi-core processors, to be properly driven by quality of experience. We want for this to pervade multiple industries.
2. Standardised mobile benchmarks. We want the prerecorded workloads to become a standard benchmark suite, replacing the old suites that are not appropriate for performance-to- energy trade-offs.
3. Longer battery life or more powerful applications. Lowering the energy consumption of mobile devices will allow either more power hungry code to be used or for the devices to last longer between charges. This will be hugely beneficial for both the general public and business people. We expect this to come partly from our new scheduler, but also from derived works that use our new evaluation methodology or that embed the interaction lag predictors.
Description Image you are trying to make mobile phones' batteries last longer. You need to make sure the phone's software uses less energy. There are a few ways to do this, the most notable of which is to slow the phone's CPU (it's processor) down. This saves energy but also makes the phone's software run slower. There are many times when this won't be noticed by the user, but we need to know when those times are so that we only slow down then. Otherwise we risk annoying the user.
We found that we can create repeatable workloads that match what real users do with their phones. We can replay those on a phone and determine the impact that any change might have on performance, energy, and, crucially, user annoyance. Prior to our work, that last bit was not possible. This allows us to design better operating system components that improve energy consumption without the user noticing any slowdowns.
Exploitation Route Techniques derived from ours are being used by a large chip and OS company to improve the performance and power consumption of its systems.
Sectors Digital/Communication/Information Technologies (including Software)

Description We showed how to measure the performance and power consumption of mobile applications in a repeatable way. This is essential to improving those aspects of mobile phones and interactive devices. Anecdotally, techniques derived from ours are currently being used in a mobile chip and OS company to improve their testing. A different, large mobile device manufacturer is working with us to build an energy efficient heterogeneous OS scheduler.
First Year Of Impact 2018
Sector Digital/Communication/Information Technologies (including Software)
Impact Types Economic

Title CLGen, Deep learning based program generator for OpenCL 
Description CLGen uses deep learning to generate human like programs for further machine learning and compiler fuzzing. 
Type Of Material Improvements to research infrastructure 
Year Produced 2017 
Provided To Others? Yes  
Impact Ongoing.