Reviews: Cinder

[Rodrigo Fonseca]

Hi,

Please post your reviews to Cinder here. Congratulations on submitting
your last round of reviews!

Thanks,
Rodrigo

[Matt Mallozzi]

Matt Mallozzi

12/2/10

Title:

Energy Management in Mobile Devices with the Cinder Operating System

Authors:

Roy, Rumble, Stutsman, Levis, Mazieres

Date:

2010

Novel Idea:

An operating system that treats energy as a resource given to threads and

processors, similar to memory and CPU time. The system should provide

isolation, subdivision, and delegation of energy resources among principles.

Main Results:

A mobile operating system based on HiStar that provides isolation,

subdivision, and delegation of energy resources among principles. It

accomplishes this by adding two more kernel object types for HiStar to keep

track of: reserves, which represent quantities of energy resources, and

taps, which represent rates of energy consumption, that also serve as

connections between taps.

Impact:

As it is, Cinder itself is unlikely to break into the mobile operating

system space: this is already pretty well filled by Android, iOS, Windows

Mobile, etc. However, if Cinder could be integrated into Android (abandoning

the "clean slate" that the authors valued), then it has real potential to

turn Android into the first mobile OS that actually acts like it is running

on a battery-constrained system.

Evidence:

Sample usages of each of the desired properties of the system (isolation,

subdivision, and delegation) via some common scenarios. These scenarios for

the most part are very simple.

Prior Work:

Cinder builds most heavily on HiStar, which is an operating system that

focuses on security policies. Cinder adds taps and reserves as kernel

objects to be managed by HiStar. The idea of taps and reserves is based on

resource containers.

Competitive Work:

ECOSystem allows for energy to be divided by containers, but fails hard in

that the container layout is flat, not hierarchical - either a parent

process shares a container with its child, or it puts it in an entirely new

and unrelated container.

Reproducibility:

I can't tell if Cinder itself is open source, but HiStar is. With HiStar,

the ideas presented in the paper should be relatively straightforward to

implement.

Criticism:

As with most other papers involving mobile phones, Cinder is only tested on

one device. Cinder at least has a better excuse for it, since running on

other devices would have involved more fiddling with drivers. However, it is

possible that their "shim" idea to interface with the black-box Android

drivers might have made running on another device easy enough to be worth

testing.

Ideas For Further Work:

Integrate the developments of Cinder with Android.

[Abhiram Natarajan]

Paper Title: Energy Management in Mobile Devices with the Cinder
Operating System

Author(s): Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis,
David Mazieres, Nickolai Zeldovich

Date: 2010

Novel Idea: Development of an operating system for mobile phones which
allows users and applications to control and manage limited device
resources such as energy.

Main Result(s): Cinder accurately tracks principals responsible for
resource consumption even across interprocess communication, and
allows applications to delegate their resources either in terms of
rates or quantities.

Impact: Cinder will help users monitor and analyse energy usage for
various applications. It helps model energy usage, while going beyond
the capabilities of current operating systems.

Evidence: The authors show how Cinder maintains system lifetime in the
presence of malicious applications, reserves energy for critical
functions such as 911, supports energy-aware applications, easily
augments existing Unix applications with energy polices, properly
amortizes costs across multiple principals, and allows applications to
sandbox untrusted subcomponents (such as browser plugins).

Prior Work: There exists the following work in the corresponding
areas:
(a) Resource Management - Abstraction of resource containers, cgroups,
ECOSystem
(b) Measurement, Modelling and Accounting - PowerScope, ECOSystem,
Koala-enabled system, Quanto
(c) Energy Efficiency - Spinning down disks, Odyssey system

Competitive Work: The authors verify whether:
(1) Cinder can control power through subdivision, Delegation and
isolation
(2) It provides visibility into the energy and power of a running
system
(3) Cinder can improve a system's energy efficiency by managing
complex devices with non-linear power consumption

Reproducibility: The authors are moderately transparent about the
implementation details, hence it would probably be possible to
reproduce cinder in some measure.

[Duy Nguyen]

Paper Title 

Energy Management in Mobile Devices with the Cinder OS

Authors 

Arjun Roy, Stephen M. Rumble,...

Date 

Unknown

Novel Idea 

Unlike other works in energy which mostly deal with tracking and giving the

visibility of energy usage, Cinder is an Operating System which allows 

controlling and managing energy in mobile phones.

Main Result 

Cinder is built on top of the HiStar operating system. It introduces the

abstractions of "reserves" and "taps" to model resource usage. "Reserves"

is used to describe a right to use resource and acts as an allotment from

which applications can draw resources. "Taps" places rate limits on the

consumption of applications and provides fine grained accounting.

Impact 

Unknown 

Evidence 

A prototype has been implemented on Google G1. Some applications are

also implemented to experiment the "reserves" and "taps" concepts provided

by Cinder: a task manager controlling energy consumption of background

processes, an energy constrained web browser,...

Prior Work 

Cinder "taps" & "reserves" are built based on Resource Container concept.

ECOSystem and Quanto are inspiration of Cinder accounting system.

Reproducibility 

Yes. Source code is available 

Questions

Why they chose HiStar as the underlying infrastructure? Does exokernel

architecture of HiStar makes it easier to implement Cinder? Android is

also opensource and is much more popular than HiStar, so this feature

can reach larger user population.

On Wed, Dec 1, 2010 at 7:57 PM, Rodrigo Fonseca <rodrigo...@gmail.com> wrote:

[Shah]

Title:

Energy Management in Mobile Devices with the Cinder Operating System

Authors:

[1] Arjun Roy
[2] Stephen M. Rumble
[3] Ryan Stutsman
[4] Philip Levis
[5] David Mazieres
[6] Nickolai Zeldovich

Source and Date:

N/A (couldn’t find a reference online)

Novel Idea:

The authors develop a novel, built-from scratch operating system
called Cinder for mobile phones and devices.

Main Result:

The scientists present Cinder and further introduce two new
abstractions - reserves and taps. The former describes a right to use
a given resource. The latter is a special purpose thread whose sole
job is to transfer energy between reserves.

Impact:

Since I’m not clear on when this paper was launched, it’s hard for me
to say what impact it’s had to date.

Evidence:

The authors provide ample evidence to support their claims.
Specifically, they not only build special, custom applications, indeed
they have a detailed evaluation section.

Prior Work:

The authors don’t list prior work per se - they lump’Prior Work’ and
‘Competitive Work’ together as ‘Related Work’ in Section 7.

Competitive Work:

The authors divide all related work into three categories:

[1] Resource Management

[2] Energy Accounting

[3] Energy Efficiency

In the first group they mention ECOSystem. In the second, they list
PowerScope, Koala and Mantis. Finally, they mention Odyssey.

Reproducibility

The results shown in this paper are very reproducible. The operating
system is open source.

Question:

This is a cool idea. I especially like the section that detailed
saving 5 minutes for an emergency call. Will this kind of catch on?

Criticism:

None. This was an interesting paper.

Ideas for Further Work:

As the authors state, in Section 8, they could explore more areas on
imposing reserves on say, text messages sent by certain number of
applications.

[Dimitar]

Energy Management in Mobile Devices with Cinder
Operating System

Authors: Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis,
David Mazieres
Nickolai Zeldovich

Novel: idea: The paper presents Cinder, and operating for mobile
phones and devices, which allows
users and applications to control limited resources such as energy.
The key difference between Cinder
and other operating systems for mobile phone such as Android is that
Cinder offers visibility and
control over energy usage while the others only offer visibility. In
order to achieve this Cinders uses three
mechanisms: isolation , subdivision , and delegation.

Results: Working operating system based on HiStar OS. HiStar's
flexible security model allows Cinder
to provide fine-grained application policies. Cinder uses the
abstraction reserve to measure energy use.
Every application has a fine number of resources which are consumed
when application takes an action,
an action will not be allow to take place if there are not enough
resources. Cinder uses another
abstraction call tap which is a thread that can transfer resources
between applications.

Impact:Since a lot of Android applications can be easily ported to
Cinder. I think the operating system
has a great potential.

Evaluation: The cases prove that Cinder can prevent unintended battery
drains , and it can save
energy by managing complex devices.

Prior Work: Cinder complements Quanto on modeling and accounting for
energy use.

Competitive Work: Android

Reproducibility: The test cases and implementation of the operating
system are reproducible because
Cinder is open source.

On Dec 1, 7:57 pm, Rodrigo Fonseca <rodrigo.fons...@gmail.com> wrote:

[Zikai]

Paper Title: Energy Management in Mobile Devices with the Cinder
Operating System

Author(s): Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis,
David Mazières, Nickolai Zeldovich

Date/Conference: Technical Report CSTR 2010-02, Computer Science
Department, Stanford University, 3. June 2010

Novel Idea: Introduce two new abstractions: reserves and taps. Based
on them, form an IPC system that fundamentally accounts for resource
usage on behalf of principals and implement isolation, subdivision and
delegation mechanisms to effectively control power.

Main Results: (1) Design and implement Cinder, an operation system
that is designed around security and fine-grained resource accounting
and control.
(2) Design and implement several applications that use reserves and
taps to achieve power control: a task manager that limits energy
consumption by background applications; energywrap which augments
existing applications with energy policies even when they are entirely
energy-unaware; a mobile browser which has fine-grained control over
energy consumption of plugins; energy reservation for emergency phone
calls.
(3) Evaluate the above applications to see whether they have achieved
design goals.

Impact: Cinder is a template for mobile operation systems on how to
provide good energy management and accounting support. With Cinder, we
are having a big step towards solving one of the largest problems for
current smart phones, i.e. energy shortage.

Evidence: In Part6, authors evaluate whether the applications that use
reserves and taps mentioned in Main Results (2) have achieved their
design goals. Also they test whether Cinder achieves isolation,
subdivision and delegation when handling buggy and malicious
applications.

Prior Work: In Part7, authors discuss related work in resource
management, measurement/modeling/accounting and energy efficiency.

Reproducibility: Cinder is available through Git on the HiStar page.
One can install it on smart phones, write test applications that use
reserves and taps as described in the paper and reproduce the tests.

Question: In Part3.3, the authors claim that performing fine-grained
resource transfers and accounting with special-purpose threads will
consume a large fraction of energy to be transferred. Is the tap
abstraction a more energy-efficient way? If so, why?

Criticism: Though in Part4, authors cover their implementation
details, they do not cover details that are related to reserve and
tap. It will be good to know software/hardware modifications to
implement reserve and tap model and the cost of doing so.

[James Chin]

Paper Title: “Energy Management in Mobile Devices with the Cinder

Operating System”

Authors(s): Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis,
David Mazieres, and Nickolai Zeldovich

Date: 2010

Novel Idea: This paper presents Cinder, an operating system for mobile
phones and devices, which allows users and applications to control and
manage limited device resources such as energy. Cinder introduces two
new abstractions, reserves and taps. Unlike prior approaches, Cinder

accurately tracks principals responsible for resource consumption even
across interprocess communication, and allows applications to delegate
their resources either in terms of rates or quantities.

Main Result(s): The authors showed how Cinder maintains system

lifetime in the presence of malicious applications, reserves energy
for critical functions such as 911, supports energy-aware
applications, easily augments existing Unix applications with energy
polices, properly amortizes costs across multiple principals, and
allows applications to sandbox untrusted subcomponents (such as
browser plugins).

Impact: As applications are bound to shift from simply being buggy to
being outright malicious it will be critical that mobile operating
systems provide mechanisms to protect the user’s data and resources.
Also, while at least one mobile operating system, Android, provides
improved visibility into system power use power use, it does not
provide control. Cinder takes care of both of these issues, as it’s
an operating system designed around security as well as fine-grained
resource accounting and control.

Evidence: In order to gain experience with Cinder’s abstractions, the
authors developed a number of applications using reserves and taps.
Using those applications, the authors evaluated whether Cinder can
control power through subdivision, delegation, and isolation as well
as whether it provides visibility into the energy and power of a
running system. Furthermore, by examining how applications use the
phone datapath, they evaluated whether Cinder can improve a system’s

energy efficiency by managing complex

devices with non-linear power consumption. All of the experiments
were evaluated using Cinder running on an HTC Dream.

Prior Work: Related work comprises projects dealing with resource
management, energy accounting, and energy efficiency.

Competitive Work: [Same as prior work.]

Reproducibility: The findings appear to be reproducible if one follows
the testing procedures outlined in this paper and has access to the
code for Cinder.

Question: What’s the next step for Cinder in terms of development or
deployment?

Criticism: Testing was only done on an Android-based phone.

Ideas for further work: Apply reserve and tap abstractions to other
resource allocation problems beyond energy consumption.

[Zikai]

Paper Title: Energy Management in Mobile Devices with the Cinder
Operating System

Author(s): Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis,
David Mazières, Nickolai Zeldovich

Date/Conference: Technical Report CSTR 2010-02, Computer Science
Department, Stanford University, 3. June 2010

Novel Idea: Introduce two new abstractions: reserves and taps. Based
on them, form an IPC system that fundamentally accounts for resource
usage on behalf of principals and implement isolation, subdivision and
delegation mechanisms to effectively control power.

Main Results: (1) Design and implement Cinder, an operation system

that is designed around security and fine-grained resource accounting
and control.

[Visawee]

Paper Title :
Energy Management in Mobile Devices with the Cinder
Operating System


Author(s) :
Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis, David

Mazi`eres, Nickolai Zeldovich


Date :
Year 2010


Novel Idea :
Design mobile phone operating system kernel from scratch using HiStar
and two new kernel object types: reserves and taps. This new OS
provides security and fine-grained resource accounting and control.


Main Result(s) :
(1) Cinder is able to isolate malicious or buggy application. The
experiment in section 6.1 shows that Process B is free to subdivide
and delegate energy from its reserve without affecting Process A
energy consumption.
(2) Cinder is able to support energy-aware application. The authors
implemented an energy aware networked picture gallery viewer to
support this claim. The result shows that when an energy become
scarce, the application adapts to lower the quality of images
downloaded.
(3) Cinder is able to amortize costs across multiple principals. The
authors use the experiment in section 6.3 to support this claim. The
cooperative networking stack doubles quality of service while spending
less energy. 12.5% less energy is used in the same time interval for
an equivalent amount of work.


Impact :
A mobile phone operating system that allows users with better security
and fine-grained energy control.


Prior Work :
This paper is related to work in the following areas
(1) Resource Management - ECOSystem.
(2) Measurement, Modeling, and Accounting - PowerScope, EcoSystem,
Quanto.
(3) Energy Efficiency - Odyssey system.


Reproducibility :
The results are reproducible if given Cinder code because the
experiments are mentioned in detail in the paper.

[Siddhartha Jain]

Novel Idea:
An OS, Cinder that allows users to monitor and control total energy usage and rate of energy usage
for different applications. Cinder gives one actual control over energy usage. Introduces concepts
of reserves of energy from which energy can be allocated to enforce total energy usage and taps to
control the rate of energy usage.

Main Results:
The mechanism by which Cinder controls energy usage and its rate is described. The system has been
implemented and an evaluation based on common scenarios is given. An energywrap utility to sandbox
malicious or buggy apps is described.

Evidence:
The OS is evaluated based on case studies. The authors demonstrate that Cinder meets its goals of
being energy aware and allocating energy properly and sandboxing buggy/malicious apps properly.

Reproducibility:
Not easily reproducible as the number of factors involved in an entire OS are huge and may need to
be described to reproduce results.

Prior Work:
Lot of prior work - borrows a lot from ECOsystem in terms of the abstractions.

Ideas for future work:
Implementing a system of priorities for the energy consumption would be interesting. For instance
if you're downloading something while taking a call, then taking a call should take priority.

On Wed, Dec 1, 2010 at 7:57 PM, Rodrigo Fonseca <rodrigo...@gmail.com> wrote:

[Joost]

Title: Energy Management in Mobile Devices with the Cinder Operating
System

Authors: Arjun Roy, Stephen M. Rumble, Ryan Stutsman, Philip Levis,
David Mazieres, Nickolai Zeldovich
Date: 2010

Novel Idea: This paper proposes a new OS for mobile devices based on
the idea of placing limits on processes power consumption. In
particular each process, one creation is allocated a certain amount of
power throughput that it can manage as it desires.
Main Results: The authors successfully implemented this new platform
which rations the use of power to different processes in order to
limit energy consumption on the device.
Impact: While the idea of creating a mobile platform that uses this
type of battery saving technology seems cool, it presents issues both
in terms of integrating the OS into existing mobile devices, creating
intuitive ways for app designers to work with these added constraints,
and dealing with a populace that may not few eeking out another hour
or two of battery live worth some of the drawbacks that this platform
has in usage.
Evidence: The authors present data on several different use cases
including data transference, camera image rendering, and network
request activation that demonstrate that their device consumes
radically lower amounts of energy in usage than traditional platforms.
Prior Work: This paper builds both on previous work to decrease power
usage on small devices (like Quanto) as well as existing research in
voltage and frequency scaling of the CPU.
Criticism/Question: What kind of noticeable performance hits does a
system operating under these constraints take, when performing an
action like checking one's mail or other common device usages?

On Dec 1, 7:57 pm, Rodrigo Fonseca <rodrigo.fons...@gmail.com> wrote: