indeed - I'm not wrong (I hpe) in saying tha the starting point for edge compute in mobile data nets was
to virtualise the radio fucntions so that 4G and 5G could support earlier protocols without having to
support legacy hardware so the whole NFV thing originated there i guess
but now we're taking about application space, but it has to be prtty tightly coupled to information
about location (not just for latency, but also security and energy reasons)
the problem for the "two" communities is that while cloud (including new offerings on edge clould from
just about all the major players, and things like lamda) is open source (even some application
platforms like tensorflow, which includes wide area machine learning tools - as do lots of research
projects), the 5G protocol space is almost completely opaque to researchers:-)
> Speaking personally, there remain a number of architectural challenges if
> you look at this in the 5G landscape because the standards coming out of
> 3GPP really still see the network and compute/application landscapes as
> separate.
>
> MEC has not been widely adopted because the notion of the compute edge in
> 4G is beyond the interface to the non-mobile network (SGi in 3GPP speak)
> and is likely to be in the same place in 5G (N6 interface). Main
> transport protocol in both 4G and 5G remains GTP with IPSec between
> eNB/gNB and SGW/UPF.
>
> There are not-insignificant issues with Lawful Intercept in MEC as well -
> the only solid proposal I've seen for a commercial MEC solution suffers
> two problems - one, a "bump-in-the-wire" for GTP and IPSec (yuck!) and
> two, a lack of LI which typically happens at the SGI point in current 4G.
>
> The issue is that the current mobile network is not really that mobile in
> architecture - extensive use is made of fast-moving tunnels in both 4G
> and 5G and there are (currently) no plans to change as it works.
>
> In 5G, the functional elements of the data-plane (UPF) can be moved
> closer to the edge but this creates issues in concurrency of data - in
> other words, I can place portions of the 5G network near the cell-tower
> and put an edge cache there but what happens when I move between cells
> because I now have two problems - firstly, I've broken the mobility which
> works VERY well in todays' networks (less than 20ms of lost data on the
> most disruptive move in a cellular network, typically more like 1ms
> without breaking/reallocating any IP/TCP/TLS) and secondly I now need a
> scheme to distribute data to those caches in the first place.
>
> The problem we seem to be facing is one where there is no obvious
> standards group to move this forward - IETF will look at it from a
> bits-and-bytes level at the network (and really are not engaged in 5G at
> all), 3GPP only see the mobile network as a transport between handset and
> the outside world.
>
> Yes, there is LOTS to consider here...
>
> Where do you want to start???!!!
>
> David
> -----Original Message-----
> From: Next Generation Networking <[log in to unmask]> On Behalf Of Jon
> Crowcroft
> Sent: 16 March 2019 12:45
> To: [log in to unmask]
> Subject: Re: Architectural developments for 5G?
>
> there's a bunch of projects on "edge" computing - huawei & BT also are
> both pursuing what they call compute first networking (or in-network
> processing) - back in the day, telefonica did a bunch of work assuming
> you'd have a rack at every cell tower....
>
> you are absolutely right that the architecture for scaling the
> application (what in cloud terms is the orcestration or coordination
> system) is crucial, and not at all a mapped out design space, so I think
> there's lots of interesting research to do on what systems like
> kubernetes for edge cloud would look like (that's just one piece of the
> problem - security architecture is also not obvious in that world, and
> mobile/migrating computation for mobile users makes life even more of a
> challenge)
>
> great thought!
>
> > Dear all,
> > Just a thought for today. An interesting talk last year at Coseners on
> > the realities of 5G.
> > Is the research community looking at local processing needs (or,
> > rather, future needs) for applications that develop on the back of 5G?
> By 'local'
> > I mean within the decision time frame required by the application and
> > hosted somewhere on the network and replicated probably at or near the
> > edge. The questions I have in mind are practical realisations that
> scale.
> > There are several views on what that level of scale might evolve
> > towards and what latency requirements of applications might drive the
> > preferred solution and I don't wish to comment on every scenario.
> > There are some open questions (I would guess) on an architecture for
> best scaling.
> > Regards,
> > John
> >
> > ######################################################################
> > ##
> >
> > To unsubscribe from the NGN list, click the following link:
> > https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.
> > jiscmail.ac.uk%2Fcgi-bin%2Fwebadmin%3FSUBED1%3DNGN%26A%3D1&data=02
> > %7C01%7Cd.lake%40SURREY.AC.UK%7C427fafdd81c24456d7c108d6aa0d73e8%7C6b9
> > 02693107440aa9e21d89446a2ebb5%7C0%7C0%7C636883372086240451&sdata=s
> > S5r15cq7wYSVmU0kHdvaOAaIgk9x4%2B5%2Fx5XkX1InXY%3D&reserved=0
> >
> ########################################################################
>
> To unsubscribe from the NGN list, click the following link:
> https://eur02.safelinks.protection.outlook.com/?url=https%3A%2F%2Fwww.jiscmail.ac.uk%2Fcgi-bin%2Fwebadmin%3FSUBED1%3DNGN%26A%3D1&data=02%7C01%7Cd.lake%40SURREY.AC.UK%7C427fafdd81c24456d7c108d6aa0d73e8%7C6b902693107440aa9e21d89446a2ebb5%7C0%7C0%7C636883372086240451&sdata=sS5r15cq7wYSVmU0kHdvaOAaIgk9x4%2B5%2Fx5XkX1InXY%3D&reserved=0
>
########################################################################
To unsubscribe from the NGN list, click the following link:
https://www.jiscmail.ac.uk/cgi-bin/webadmin?SUBED1=NGN&A=1
|
