[b2g] question regarding realtime performance guarantees

[debamitro]

Hello,

Not sure if this is the right forum, but asking anyway as this is the next 
smartphone OS after Android, as I see it. Is there anyway to ensure guaranteed 
realtime performance for some parts of the phone OS? I want this for ensuring 
that basic telephony services (making/receiving a call/text) work with a 
realtime guarantee regardless of the state of the OS (i.e., even if it is 
sluggish or hanging due to some heavy apps). I feel this is an important design 
consideration for a mobile OS. Any inputs regarding something done on these 
lines in b2g, or something that can be done?

Regards,
Debamitro Chakraborti
___
dev-b2g mailing list
dev-b2g@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-b2g

Re: [b2g] question regarding realtime performance guarantees

[Dave Hylands]

Since Android and FxOS are both based on linux, and linux can't guarantee 
realtime performance, neither can guarantee realtime performance.

Both Android and FxOS use garbage collected languages (Java for Android, 
JavaScript for FxOS) which don't use realtime garbage collectors, so even if 
the OS were realtime, it still wouldn't help unless the portions of interest 
were written in languages which don't need GC (like C or C++).

Dave Hylands

- Original Message -
> From: debami...@gmail.com
> To: dev-b2g@lists.mozilla.org
> Sent: Thursday, March 27, 2014 9:29:31 PM
> Subject: [b2g] question regarding realtime performance guarantees
> 
> Hello,
> 
> Not sure if this is the right forum, but asking anyway as this is the next
> smartphone OS after Android, as I see it. Is there anyway to ensure
> guaranteed realtime performance for some parts of the phone OS? I want this
> for ensuring that basic telephony services (making/receiving a call/text)
> work with a realtime guarantee regardless of the state of the OS (i.e., even
> if it is sluggish or hanging due to some heavy apps). I feel this is an
> important design consideration for a mobile OS. Any inputs regarding
> something done on these lines in b2g, or something that can be done?
> 
> Regards,
> Debamitro Chakraborti
> ___
> dev-b2g mailing list
> dev-b2g@lists.mozilla.org
> https://lists.mozilla.org/listinfo/dev-b2g
> 
___
dev-b2g mailing list
dev-b2g@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-b2g

Re: [b2g] question regarding realtime performance guarantees

[Debamitro Chakraborti]

Hi Dave,

On Fri, Mar 28, 2014 at 12:36 PM, Dave Hylands  wrote:

> Both Android and FxOS use garbage collected languages (Java for Android,
> JavaScript for FxOS) which don't use realtime garbage collectors, so even
> if the OS were realtime, it still wouldn't help unless the portions of
> interest were written in languages which don't need GC (like C or C++).
>

This is exactly what I had in mind. Possible approaches might be:
1. Allow some apps to be written at a lower level and not use dynamic
memory management and GC
2. Handle telephony operations by an interrupt, so that calling/texting
always overrides the rest of the app framework
I think we'll build a better mobile OS this way. My humble opinion only.

Regards,
Debamitro
-- 
http://about.me/debamitro
*don't polish your ignorance, it will shine*
___
dev-b2g mailing list
dev-b2g@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-b2g

Re: [b2g] question regarding realtime performance guarantees

[Thomas Zimmermann]

Hi,

Just out of interest, is there any actual use case here?

>From a technical POV, phones are short on resources and the software is
mostly driven by user input. RT, in contrast, is often bad for resource
usage and interactive workloads, because you have to expect WCET for
many components. So you have to allocate resources for RT tasks,
although you probably won't actually need them. In the end you'll
achieve the opposite of what you intended.

And please note that telephony runs on a separate CPU, which comes with
an RT-enabled OS.

Best regards
Thomas

On 28.03.2014 08:35, Debamitro Chakraborti wrote:
> Hi Dave,
> 
> On Fri, Mar 28, 2014 at 12:36 PM, Dave Hylands  > wrote:
> 
> Both Android and FxOS use garbage collected languages (Java for
> Android, JavaScript for FxOS) which don't use realtime garbage
> collectors, so even if the OS were realtime, it still wouldn't help
> unless the portions of interest were written in languages which
> don't need GC (like C or C++).
> 
> 
> This is exactly what I had in mind. Possible approaches might be:
> 1. Allow some apps to be written at a lower level and not use dynamic
> memory management and GC
> 2. Handle telephony operations by an interrupt, so that calling/texting
> always overrides the rest of the app framework
> I think we'll build a better mobile OS this way. My humble opinion only.
> 
> Regards,
> Debamitro
> -- 
> http://about.me/debamitro
> /don't polish your ignorance, it will shine/
> 
> 
> 
> ___
> dev-b2g mailing list
> dev-b2g@lists.mozilla.org
> https://lists.mozilla.org/listinfo/dev-b2g
> 

___
dev-b2g mailing list
dev-b2g@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-b2g

Re: [b2g] question regarding realtime performance guarantees

[Debamitro Chakraborti]

Hi Thomas,

Just out of interest, is there any actual use case here?
>
> Nope, so far purely hypothetical. I started the conversation inspired by
research like 
this,
and from my personal experience of my smartphone being unable to allow fast
calling when I needed it.

>From a technical POV, phones are short on resources and the software is
> mostly driven by user input. RT, in contrast, is often bad for resource
> usage and interactive workloads, because you have to expect WCET for
> many components. So you have to allocate resources for RT tasks,
> although you probably won't actually need them. In the end you'll
> achieve the opposite of what you intended.
>
> okay, I had no idea of this aspect!


> And please note that telephony runs on a separate CPU, which comes with
> an RT-enabled OS.
>
> Yes I am aware of this -- you mean the baseband processor, right? I was
thinking about possible modifications to how the telephony app works, to
ensure better performance. Anyway I am not much knowledgeable in OS
architecture so might need to dig around a little more to see if what I am
thinking makes sense!
Thanks for the replies

Regards,
Debamitro

> Best regards
> Thomas
>
> On 28.03.2014 08:35, Debamitro Chakraborti wrote:
> > Hi Dave,
> >
> > On Fri, Mar 28, 2014 at 12:36 PM, Dave Hylands  > > wrote:
> >
> > Both Android and FxOS use garbage collected languages (Java for
> > Android, JavaScript for FxOS) which don't use realtime garbage
> > collectors, so even if the OS were realtime, it still wouldn't help
> > unless the portions of interest were written in languages which
> > don't need GC (like C or C++).
> >
> >
> > This is exactly what I had in mind. Possible approaches might be:
> > 1. Allow some apps to be written at a lower level and not use dynamic
> > memory management and GC
> > 2. Handle telephony operations by an interrupt, so that calling/texting
> > always overrides the rest of the app framework
> > I think we'll build a better mobile OS this way. My humble opinion only.
> >
> > Regards,
> > Debamitro
> > --
> > http://about.me/debamitro
> > /don't polish your ignorance, it will shine/
> > 
> >
> >
> > ___
> > dev-b2g mailing list
> > dev-b2g@lists.mozilla.org
> > https://lists.mozilla.org/listinfo/dev-b2g
> >
>
>


-- 
http://about.me/debamitro
*don't polish your ignorance, it will shine*
___
dev-b2g mailing list
dev-b2g@lists.mozilla.org
https://lists.mozilla.org/listinfo/dev-b2g