Big 5G Event: Cut through the commotion of network slicing

00:00 good day everyone uh welcome to the

00:02 panel at the big 5g conference on the

00:05 commotion

00:06 and cutting through the commotion of

00:08 network slicing and we have five great

00:11 speakers today

00:12 uh but before we get started i'd just

00:14 like to put a little bit of comment on

00:16 the state of network slicing

00:19 network slicing as many of you know was

00:21 long touted as the

00:23 large-scale revenue generator for 5g

00:26 and in many cases viewed as the anchor

00:29 business case

00:31 as we all know it's been a little slow

00:33 to develop

00:34 but i think we're all optimistic that

00:36 with 5gsa and with some extra things

00:38 we're going to be talking about today

00:41 slicing is finally going to be

00:42 manageable cost effective and maybe in

00:45 the next couple of years become dynamic

00:47 so that it's actually cheaper to use 5g

00:50 for a private virtual vpn rather than

00:54 even sd-wan

00:55 so

00:56 let me introduce our speakers and ask

00:58 them to say a little bit about

01:00 themselves

01:01 let's start with anand shah

01:04 hey sue thanks a lot my name is anan

01:06 shah i work for verizon i'm a director

01:08 of technology and product management

01:10 there and as most jobs entail in the

01:12 telco world a lot of 5g under my belt

01:15 but also we concentrate heavily on

01:17 private networks and network slicing

01:22 great

01:23 i'm ray doland

01:25 thanks sue my name is ray dolan i am the

01:27 ceo at cohere technologies

01:30 um cohere has developed a suite of

01:33 x-apps that allows the intelligence that

01:35 currently lives in base stations to be

01:38 left there or moved to the edge data

01:40 center uh a lot of that bears on what

01:43 openran is looking for

01:45 and it directly relates to being an

01:48 enabling technology for network slicing

01:50 so we're looking forward to the

01:51 discussion

01:52 cool uh constantine

01:57 oh thank you sue i'm constantine

01:59 polygonopolis i'm vice president of 5g

02:01 and telco cloud at juniper networks

02:04 and i'm focusing i'm responsible for 5g

02:07 products and solutions we're focusing at

02:09 jury pair particularly on the oren

02:11 architecture

02:13 the rig x-apps are apps

02:15 and the end-to-end smo that stitches

02:17 everything together to deliver the

02:19 promise of network slicing

02:22 great

02:22 franco

02:24 hey it's franco and franco missouri from

02:26 infovista i'm the chief product and

02:29 strategic officer for the company

02:32 improvista is a is a leader company in

02:35 what we call the network life cycle

02:37 automation that means driving from

02:41 the planning to the testing through

02:42 deployment

02:44 everything about the 5g life including

02:47 the network

02:48 license thank you and troy

02:52 hi troy sonya here from bell canada i

02:54 have the pleasure of supporting a team

02:56 that does our strategy for what we call

02:58 our smart core which is the intelligence

03:01 that helps us

03:02 manage and uh and uh monetize

03:05 our best-in-class fiber and 5g networks

03:09 and i'm really looking forward to our

03:11 dialogue today around network slicing

03:13 great

03:14 so let's start off with our first topic

03:17 and i'm going to ask annan to lead it

03:19 off so and how are the operators setting

03:22 up the foundation for really making

03:24 network slicing a big part of revenue

03:26 generation in the future

03:28 absolutely and again i'm super excited

03:30 to be here and happy to talk about

03:32 network slicing it's as it's become a

03:35 serious buzzword in the telecom world

03:37 but hopefully we can dissect the issues

03:39 and provide you with the insights you

03:41 need today um so just some background

03:44 sue i mean the the main thing here is

03:46 that we have built a network especially

03:49 verizon and other mmos across the world

03:52 that feeds 150 million customers uh

03:56 today and it's a

03:58 it's a one single network it's a best

04:00 effort network

04:02 and we're slowly decomposing the network

04:05 to see if we can build 150 million

04:08 networks for each user that's out there

04:12 now of course this flexibility comes

04:14 with price and it's going to be hard to

04:17 properly manage the flexibility won't be

04:19 there so it's going to be of essence to

04:21 have the right management tools in place

04:24 to to have a network

04:26 slicing or to have network slicing in a

04:28 network this big

04:30 so

04:31 as uh as a company right now you know we

04:34 started in the 4g days and in the 4g

04:36 days we went through the same thing when

04:39 we had apns we were thinking of

04:41 20 30 40 50 apns and and where each apn

04:46 will lead to in the network and we were

04:48 so excited about it and at the end we

04:50 found out that hey maybe five 10 apns

04:53 are good enough the same thing can

04:55 happen today

04:56 in network slicing we're looking at

04:58 three big slices right now as you know

05:00 embb which has high throughputs high

05:02 mobility

05:04 we have mission critical as well which

05:06 is maybe low data

05:08 uh no mobility and then you are llc

05:11 which is extremely high latency or low

05:14 latency that you need for mission

05:15 critical applications as well and low

05:17 latency applications

05:19 and then um we can keep expanding from

05:22 that so i think what's very important in

05:25 network slicing if i want to name two

05:27 things are truly understanding the um

05:31 the customer needs that

05:33 digital technologies are means to an end

05:36 not an end themselves okay so when we

05:40 talk about that it's truly understanding

05:42 the throughputs from a

05:44 downlink throughput and uplink

05:46 throughput the jitter that's required

05:48 the latency that's required and maybe

05:51 putting them into categories that equate

05:53 to slices right so

05:56 first we need to understand what the

05:57 customer wants rather than getting into

06:00 the tech hype and the second thing is

06:02 ensuring that the right knobs are in

06:04 place for an mno the right knobs means

06:07 from the device to the ram to the

06:09 transport to the core there are so many

06:12 elements with so many different vendors

06:14 that are involved

06:16 having the right pieces and the right

06:18 knobs in place so we could actually

06:20 create a slice

06:21 and sustain that slice throughout the

06:24 life cycle

06:26 is going to be very important for the

06:27 mnos sue

06:30 excellent point

06:31 troy do you see things the same way or

06:34 does bell canada have a slightly

06:35 different take

06:37 now i'd have to say we're fairly aligned

06:40 um

06:41 you know some of the things that we need

06:43 to really consider is the fact that

06:45 slicing itself is still in its early

06:48 days of course uh we've got different

06:51 levels of maturity depending on the

06:52 vendor or partner that's bringing the

06:54 equipment and the software to the

06:56 network and the early days of course are

06:58 going to be entirely

07:00 manual driven so as we start to deploy

07:03 slicing some of the foundational

07:05 elements that we need to get out there

07:06 are okay obviously if we when we launch

07:08 our 5 gsa core there'll be at least one

07:10 slice you need one for embb and and

07:13 there may be a few others but the true

07:16 promise and i really liked anne's uh

07:19 comment around understanding customer

07:21 need because there are customers who

07:23 have requirements around

07:24 uh uh you know a lot of devices low

07:28 bandwidth

07:29 and so that you know kind of uh speaks

07:31 to the mmtc slice that will be coming

07:33 right as as it gets ratified into hpp

07:35 and gets released with our vendors and

07:37 then there are others who have high

07:39 bandwidth requirements so those three

07:41 natural big slice types are great

07:43 but then inside those the sub slices are

07:46 what becomes really interesting

07:48 and

07:49 there are a lot of customers that beyond

07:51 the specific performance requirements

07:54 for a given slice are also looking at

07:56 slicing as a privacy play as another way

07:59 and you mentioned it early on sue as

08:01 another mechanism beyond sd-wan to uh to

08:04 look at replacing the old mpls kind of

08:07 network right so uh so lots of uh work

08:11 is going on to understand in parallel to

08:13 the work that we're doing to build all

08:16 of the orchestration and automation

08:18 which we'll talk about a little bit

08:19 later um but in parallel we need to

08:22 really understand the customer need and

08:24 what use cases are going to leverage

08:26 that foundational capability of slicing

08:29 now let me ask you troy do we have to

08:32 wait for 5g or are there some things we

08:33 can do on slicing with lte before we get

08:37 to 5g and do with release 15.

08:41 that's always the the hard question

08:42 because of course the value prop that

08:44 you're trying to generate in the

08:45 marketplace around 5g

08:47 also carries weight and although you may

08:49 be able to do what i would call slice

08:52 like capability you notice my air quotes

08:54 slice-like capability either in 5-g nsa

08:57 or even lte i think the use cases are

08:59 going to be very particular where you

09:01 want to do that and as as operators

09:05 i know typically there's a huge

09:07 competition to you know sell the best

09:10 network and there's a lot of effort that

09:12 goes into ensuring that we win the best

09:14 network claims i know that in canada we

09:17 do that in verizon that's their value

09:18 prop uh south of the border and so when

09:21 you think about that whether it's 4g or

09:23 5g

09:24 selling better than best is really

09:26 challenging

09:28 yeah

09:29 yeah and then do you have any

09:32 extra comments on whether we can do

09:33 something today with lte or whether we

09:35 can even do something today with 5g

09:38 again

09:39 like troy said it's a big debate right

09:41 you have apns you have qos's you have

09:44 qcis is there a combination between

09:46 those two that can get close to slicing

09:49 but remember slicing is much more than

09:51 that slicing is guaranteeing the sla

09:54 slicing is having the orchestration

09:57 to determine whether that kpi or that

10:00 sla is being delivered and if it's not

10:02 spinning up the right resources in the

10:04 right compute to make sure that slice is

10:06 there which is maybe not always there

10:09 for qcis or for apn so slicing takes it

10:13 one step above but we can get maybe 75

10:15 percent there with what we have today

10:18 so franco uh you're coming at it from

10:20 the vendor who actually tries to deliver

10:22 some of that service assurance uh how do

10:25 you see that

10:26 oh he had this great question so i think

10:28 that i'm really following what uh troy

10:30 and anand are saying and i hundred

10:32 percent agree with them in my opinion

10:34 there are five steps that we have to do

10:37 to reach the real net lies in terms of

10:40 automation right now because we have a

10:43 nyberg situation with 4g 5g combined

10:46 together fundamentally i really believe

10:48 that what the operator have to do is

10:50 exactly what anand and tori said just

10:53 select three four major slides

10:55 and fit the current traffic there the

10:58 second level is fundamental automation

11:01 without automation without orchestration

11:03 we cannot reason the next step next step

11:06 is the automation from the service point

11:08 to in the network point of view looking

11:10 the service provider at reality there is

11:13 another automation linked to this that

11:15 is from the user controller the

11:16 capability of the user to select

11:19 automatically what they need what they

11:21 want and from the maybe that is the

11:23 first step the fourth step fundamentally

11:26 has to be in an in a complete

11:28 homogeneous type of network so hybrid

11:31 has to disappear because otherwise you

11:33 cannot really expand everything and the

11:35 fifth step to the mentor is a full

11:38 uh

11:39 artificial intelligence driven type of

11:41 network that fundamentally it's probably

11:44 six seven year the uh far from from now

11:48 but

11:49 the census insurance story has been

11:50 always one of the major discussions sla

11:53 is nothing new it's uh it's on the

11:55 tables in 30 years for different

11:57 productivity but these uh the

11:59 netherlands is bringing a new challenge

12:01 that was not part of the of the past

12:04 fundamentally has to be has to be

12:06 something that changed

12:08 dynamically following the dynamicity of

12:11 the network dynamics of the service

12:13 before sla was a static site the

12:16 customer signed an agreement with the

12:18 with the with the service provider that

12:20 the agreement has specific performance

12:22 figure quality figure and you have to

12:25 verify that these the figure are true

12:27 and are applied and that control is

12:29 delivered but here this figure can

12:32 change and can change the penny by the

12:33 day and the hour and the situation the

12:36 mobility speed and there are a lot of

12:38 parameter so the sla

12:41 insurance system has to really become

12:44 part of the network itself before it was

12:46 a a parallel tools that you can never

12:49 you cannot have but fundamentally was a

12:52 separate element in the new vision of

12:54 the real 5gsa network that's sva network

12:58 with real orchestration this insurance

13:01 has to be built in

13:03 as one of the components of the network

13:05 and this to be a sort of first-career

13:08 for everybody for the customer for the

13:10 operator also because operator will uh

13:14 fundamentally use and leverage other

13:16 type of network think about the public

13:18 network combined with the private

13:19 network so at this point you need to

13:21 have something that really take a look

13:24 to the sla in a much more holistic way

13:26 and provide the vision that really fit

13:29 in each of the expectations that the

13:31 different

13:32 participants will have

13:36 so it's getting much more complex but it

13:38 needs to still be end to end and one of

13:40 the other factors that's really changing

13:42 how we look at end to end slicing is the

13:46 new capabilities for including the ran

13:48 as part of the slice and monitoring the

13:50 rent

13:51 so ray maybe you would like to comment

13:53 on how does the complexity that openran

13:56 is bringing to network slicing but also

13:58 the opportunity

14:00 yeah thanks sue and um i appreciate the

14:03 comments from everybody up until now i i

14:05 would say if i could just chime in on

14:06 that question you asked can you do this

14:08 on lte the answer is yes and we'll we'll

14:11 show that very soon we actually have

14:12 already um

14:14 whether the operators want to market it

14:15 as a 5g capability or not it's the same

14:18 radio so why can't you do the same thing

14:20 guys it's an ofdm waveform um

14:22 now having said that uh for openran uh

14:25 it's i actually think openran removes

14:28 the complexity if you do it right as

14:30 opposed to adds complexity

14:32 the vision of opening iran

14:34 in many people's minds is an economic

14:37 issue of trying to disaggregate the

14:39 supply chain it's not an economic issue

14:41 it's an innovation and architectural

14:43 issue all right so

14:45 uh you'll get some better economics too

14:47 you'll get cloud economics if you do it

14:49 right but the primary thing that we've

14:51 discovered is that you can move the

14:53 intelligence to wherever you want now

14:56 think about when you do network slicing

14:57 you're trying to match up what a user

14:59 wants

15:00 including what that user is doing right

15:02 now

15:03 and what they've committed to

15:04 contractually with the ability to

15:06 enforce that across a network and if the

15:08 user plane functions in the core and the

15:10 enforcement and resource allocation

15:11 mechanisms are at the edge it's a very

15:13 complicated thing to do

15:15 if you move the intelligence and all the

15:17 resource allocation into x-apps and you

15:20 put it on a rick closer to the core you

15:22 literally have a natural marriage of the

15:24 slicing elements which is what the user

15:26 wants to do with the ability to enforce

15:28 it across a network

15:29 so what we were able to do a cohere

15:31 using delay doppler

15:33 is to take the resource allocation

15:35 functions we announced this with

15:36 vodafone recently actually vodafone

15:38 announced it we did it in partnership

15:40 with intel and vmware and they also what

15:42 they really didn't focus on slicing they

15:44 focused on spectral efficiency because

15:47 they doubled the spectral efficiency of

15:48 their network which by the way is a

15:50 pretty big thing for most operators

15:52 but what you'll see us announcing in the

15:54 next set of trials that we do is all

15:56 about network slicing because i actually

15:58 think that as big a deal as it is to

16:00 double or quadruple the network capacity

16:03 slicing is the ultimate monetization

16:05 engine it's if you can take a portion of

16:07 your network and on a per bid basis sell

16:09 it for 10x because of the productivity

16:12 benefits and the guarantees that you can

16:14 associate with it that's where

16:16 the real business models come out so

16:18 i actually think openran

16:20 is the answer to the complexity sue it

16:23 doesn't add to it

16:24 but it does add to it if you leave all

16:27 of that compute see there's a certain

16:29 set of cloud economics which says bring

16:31 everything to the core

16:33 but unfortunately in wireless people are

16:34 bringing everything to the edge because

16:36 there's this notion

16:38 that

16:39 the edge is so complicated you need to

16:40 bring the cloud to the base station

16:43 you actually can bring the base station

16:45 to the cloud and that's that's actually

16:46 what we're doing and i'm very excited

16:49 that we're now in partnership with two

16:50 other operators in the u.s one in

16:52 australia and one in in europe doing

16:55 some trials we should be able to show

16:57 this within 12 months possibly within

16:59 even six months

17:00 that's really cool but i would say the

17:02 edge is wherever it needs to be

17:04 hopefully efficiently and dynamically um

17:07 constantine um you juniper's getting

17:10 into this business um

17:12 uh from the from the rick point of view

17:14 and adding intelligence uh do you see uh

17:18 the opportunity the same way or is there

17:20 a lot of complexity to be handled to

17:22 make it work

17:24 thank you sue uh no i see i think uh

17:27 ray said it said it very well

17:30 um it's really a paradigm shift right uh

17:33 i i i think the better question is uh to

17:36 look at um you know what are the new

17:38 opportunities that oren brings to the

17:40 picture right uh the analogy i like to

17:43 uh to use is like asking whether the

17:45 electric motor adds more complexity to

17:48 the av

17:49 right it's a paradigm shift it's a

17:50 different thing and

17:52 oren i agree with ray oran takes away

17:55 complexity shifts the complexity out of

17:58 the infrastructure

18:00 and we have a truly intelligent software

18:02 driven now radio infrastructure right

18:05 can you imagine

18:07 how simple how powerful it is to be able

18:09 to upgrade the capabilities the features

18:11 of the radio

18:13 through a few key strokes as opposed to

18:15 you know uh uplifting entire uh you know

18:19 uh

18:20 boxes and uh

18:21 and bbqs and and uh radio units etc so

18:25 it's a very powerful um

18:28 model new model

18:30 and i think eventually it will take

18:33 some time for us to get it right

18:36 but then it will bring huge improvements

18:39 in tco for operators but most

18:42 importantly new capabilities that you'll

18:44 be able to uh to deliver to the

18:46 customers to the end customer right that

18:48 is not possible today the other thing

18:50 that i want to bring uh to the attention

18:52 here regarding complexity is that we

18:54 keep thinking about aura only

18:57 that's one big part there are three what

18:59 i call three

19:00 major semantic domains in the 5g

19:03 uh the run

19:05 the transport

19:06 and the core and you have many

19:09 incarnations of this right even within

19:11 the run you have transport you have the

19:12 mid hall you have

19:15 several islands of transport

19:17 you may have the functions of the core

19:20 distributed across different clouds etc

19:22 so how do you stitch all of that

19:23 together to deliver them to an sla

19:27 that goes back to what anand was saying

19:30 the importance of the of the

19:32 orchestration automation visibility

19:35 and uh dynamic control right

19:38 so uh as part of of

19:41 delivering that capability we have to

19:42 start thinking about the concept of

19:45 multiplexing right you've got one

19:47 physical infrastructure that needs to

19:48 have the ability to

19:51 other scale onto virtual resources and

19:54 on top of that we need to have to figure

19:55 out exactly how we manage

19:58 the concept of multiplexing

20:00 multiple disparate use cases onto the

20:02 same physical infrastructure to deliver

20:05 them to an sla

20:11 you're muted sue thank you i'm sorry

20:14 thank you um that's very interesting um

20:17 in the interest of time i think we have

20:19 to push on to the next topic but i think

20:21 it's related and that how does this all

20:23 fit into the private networking world so

20:26 maybe you can take that thought you were

20:28 just giving constantine and look at how

20:30 that means what the enterprise is going

20:32 to do with this kind of slicing

20:34 capability

20:36 uh that's the 1 billion or 1 trillion

20:39 dollar question right how the private

20:42 5g and the public clouds the public

20:47 5g networks are going to come together

20:48 right and who is going to do what

20:50 i believe that

20:53 we're going to see uh three buckets in

20:55 the industry one is the standalone

20:58 private

21:00 5g and 4g for that matter cbrs in the

21:03 u.s is really taking off

21:05 uh we're going to see the so-called

21:07 neutral neutral cost which is a you know

21:10 mix of private with in combination with

21:13 public

21:14 5g and then the big question is how

21:18 successful operators are going to be to

21:20 deliver the end-to-end network slicing

21:22 as a service

21:24 whereby an enterprise can really get

21:26 into a portal

21:27 set up their own use case their own

21:29 network

21:31 dimension it for a number of users maybe

21:34 have different slices multiple slices

21:37 and deliver all of that value through

21:39 the operator i think

21:40 that would be if the operators managed

21:42 to unlock that that would be a huge new

21:45 revenue stream for operators right and

21:48 that's probably the more uh challenging

21:50 use case to to realize

21:52 uh but if we you know if we look at the

21:55 industry the private enterprise

21:58 we cannot really fit them into one model

22:00 there are many verticals that have

22:02 different requirements and some of them

22:05 are aligned more with a private

22:07 standalone use case some of them

22:10 will have to rely on the operator to uh

22:12 to provide you know the capability of

22:14 network slicing and the private network

22:17 as as a as an operator-led

22:20 uh service

22:23 so

22:24 you know i don't think we can we can fit

22:26 everything into one model

22:27 but um

22:30 you know most likely the standalone is

22:32 going to um

22:34 rear its end first and then we're going

22:36 to see operators really unlock the um

22:40 you know the complexity of offering

22:42 end-to-end network slicing standalone uh

22:45 private 5g through network slicing which

22:47 i think is the most interesting the most

22:49 challenging use case but the most

22:51 powerful one

22:52 so and i saw you nodding um

22:55 is that how verizon is going to launch

22:57 it or do you see 5g

22:59 private networking being sort of islands

23:01 or do you see a hybrid multi-domain

23:04 public-private network

23:06 so

23:07 both um the way i see it is that if you

23:09 look all the way on the left right you

23:11 have us a purpose-built public network

23:14 that is best effort as you move to the

23:17 right um you might get uh uh some

23:20 network slicing capabilities here on the

23:23 public network as you move further to

23:25 the right to get more ownership of those

23:27 resources you may have a private network

23:30 in a sense as a network slice from the

23:32 public network right

23:34 there might be a private core that's a

23:35 hosted private core that could be

23:37 delivered as a network slice to your

23:40 on-premise solution and then if you go

23:42 all the way to the right we're talking

23:44 about an on-site private network and

23:46 where you have

23:47 all the rand resources all the core

23:50 resources transport resources dedicated

23:53 to your enterprise whether it's a

23:55 walmart a coke whatever it is all those

23:57 resources dedicated to you now the cool

24:00 thing about the private network is that

24:02 it'll be a handful of vendors it'll be a

24:05 handful of platforms not as complex and

24:08 not as much of a variable as we have in

24:11 the public network so what you'll see is

24:14 you'll see 5g standalone come out much

24:17 earlier on private networks you'll see

24:20 network slicing coming out much earlier

24:22 on private networks and public networks

24:24 just because it's easier to deliver a

24:27 static slice on

24:28 a box that is delivered by let's say one

24:31 vendor that's out there right so you'll

24:33 see those economics come

24:36 to fruition because

24:37 it's a

24:39 it's a now it's purpose-built for the

24:41 enterprise you could do a lot of

24:43 experimentation on that private box

24:45 itself so you'll see network slicing

24:47 standalone all those new features come

24:49 out on private networks quicker

24:52 cool

24:54 so

24:54 let's turn to how do i actually operate

24:57 and monitor these kind of networks

24:59 because we've said it can be simpler but

25:02 that's by hiding the complexity because

25:04 there are more choices and more dynamics

25:07 so uh let me ask franco

25:10 from info vista's perspective how can

25:12 operators really monitor manage and

25:15 deliver those capabilities with a an

25:17 easy-to-use user interface to both

25:19 private networks and their own network

25:22 operations center

25:24 yeah thanks to this great question

25:26 remember that they regard the complexity

25:28 the step the logical step

25:31 to ensure that the private network works

25:33 uh like uh the owner is expecting are

25:36 always the same from the planning time

25:39 when you plan the network that's going

25:41 to be a dedicated network and hybrid the

25:44 network as a network as a service from

25:46 an operator from the planning time to

25:49 the time in which you define the

25:51 services that you want to have over the

25:53 network to the time in which you ensure

25:57 that everything is there and when the

25:59 step are always the same and the type of

26:01 tools fundamentally and approach is

26:03 exactly the same of course the scale is

26:05 different skill is different

26:07 and that will make the difference and i

26:10 agree with that and then that is much

26:12 more simple than today because the scope

26:14 and the powerful number is easier subset

26:17 of the scope of the problem that you

26:18 have in a public network of course

26:20 because because you have a variety of

26:22 type of customer instead when you talk

26:23 about the enterprise the the the set of

26:26 customer or profiles there is a it's a

26:30 definitely less but fundamentally the

26:33 product we have is always the same the

26:35 planning it's linked to test and defined

26:38 sla that you can have within the private

26:40 republic network doesn't matter of

26:42 course it's just a subset of this

26:44 then

26:45 they provide everything about the

26:47 performances the inventory and the

26:49 planning and fundamental insurance of

26:51 different services and the fourth one is

26:54 become predictive in terms of uh

26:57 diagnostic root cause analysis

26:58 troubleshooting these are the four major

27:00 steps toward achieve the final target

27:03 that is optimization and then is a it's

27:06 a complete circle it's a loop that

27:08 different than the past has to run

27:10 continuously it's not just something

27:12 that you do today and then you do one

27:14 year from now

27:16 the optimization the planning the

27:18 insurance is something that you'd have

27:19 to do 24x7 and of course in the private

27:22 network that is less critical

27:25 than fundamentally in the public network

27:27 where as i said the type of profile of

27:29 the customer has so many so various that

27:32 the network has to adapt in a much more

27:36 dynamic way than the product but

27:38 definitely i really believe and we

27:40 really believe that private network will

27:42 be probably one of the major

27:44 marketing business for the next 10 years

27:47 so ray let me ask you um as you try to

27:50 link your new technology into all of

27:52 these automation and operations tools is

27:54 that something that 5g makes it easier

27:57 to integrate than it's been in the past

28:00 absolutely sue and i i think the

28:02 comments that have been made here are

28:04 just spot-on uh private network

28:06 opportunity is massive

28:08 uh i i think from franco's point of view

28:10 agree anan said it's going to come first

28:13 initially it's going to become it's

28:15 going to come first because it appears

28:16 easier but then there'll be a question

28:19 in this continuum that anon described

28:21 where its best efforts public wide area

28:24 highly

28:25 you know individualized but localized in

28:28 an enterprise those are two different

28:30 supply chains

28:32 question now for the industry is are the

28:35 wide area cellular guys going to become

28:36 enterprise savvy

28:39 or are the enterprise guys going to

28:40 become part of the supply chain for the

28:42 traditional cellular guys i will tell

28:44 you that i'm heavily biased towards the

28:46 latter

28:48 this is the time of massive disruption

28:50 the enterprise guys intuitively know and

28:53 the cloud guys intuitively know

28:55 communication scale and i t

28:58 the question is can they crack the last

29:00 black box which is the radio access

29:02 network and that's what cohere is

29:04 helping them do

29:06 i i call this the pc revolution for

29:08 radio when you can put your own

29:10 components together and make it make a

29:12 good solution

29:14 but you may still need a little tuning

29:16 to make it uh optimal uh so uh we're

29:19 we're getting close to the end of our

29:21 session but we have one major topic

29:23 which i'm going to ask everyone to

29:24 comment on but have troy kick off so

29:27 when can we really expect all of this to

29:30 come together

29:31 what's a realistic time frame

29:34 so as we said thanks very much sue and i

29:37 have to say that the comments in general

29:38 have been really really good

29:40 um

29:41 what i would say is we talked at the

29:44 outset of laying the foundation and

29:46 that's kind of the stage that we're at

29:47 now there's changes uh that have been

29:50 mentioned in terms of the radioaccess

29:52 network and adding intelligence there

29:53 ultimately slicing is is not just about

29:55 the ram though as we've mentioned it's

29:57 it's really you know ue ran transport

30:00 various types of transport out into the

30:02 core and some even think about slicing

30:05 including the apps that sit on the cloud

30:07 at the edge of the network right and for

30:09 the experience for the end customer

30:11 whether it be consumer or whether it be

30:13 enterprise to truly be assured

30:16 you need to have some visibility across

30:18 all those elements and that to me is

30:20 what slicing is really bringing to the

30:22 table that's new we've been doing

30:24 prioritization in qs on both wireline

30:26 and wireline networks for you know

30:28 dozens of years right but what's

30:30 different here is we're looking at the

30:32 resources in real time and saying can i

30:35 assure this new slice that's being

30:37 requested and following the life cycle

30:39 of that slice so that's why we're saying

30:42 starting out of the gate it's going to

30:43 be a very low number of slices it's

30:45 going to be manual while we learn about

30:46 it learning about the interaction the

30:48 dependencies between those high level

30:50 resources that we need to manage and

30:52 also we'll start plugging in our big

30:55 data machine learning capability so that

30:58 we can actually understand the patterns

31:00 that happen in the network and start to

31:02 build algorithms for closed loop

31:04 capability and so that kind of ai

31:08 leading into orchestration and

31:10 automation

31:11 i see as as really starting to

31:14 hit mainstream probably

31:17 very late next year and early in 2023 uh

31:21 in terms of operators being able to

31:23 start to offer more automation in their

31:26 5g slicing services and you know i see

31:29 that you know initially it will be

31:32 putting on new slice types more easily

31:34 being able to manage the network more

31:35 easily being able to hit the demands

31:37 that our customers are going to have and

31:39 and new

31:41 uh services and mapping those service

31:43 requirements to the slice

31:45 performances that are going to be given

31:47 that is key and that i think there's not

31:49 enough effort uh being talked about and

31:52 put into it's not just about the network

31:54 it's about what sits on top of the

31:55 network that's what matters at the end

31:57 of the day and that's what we need to

31:59 you know start putting our eye towards

32:02 and then we i start seeing you know 2023

32:05 2024 when now we're getting into dynamic

32:07 slicing even as was mentioned earlier

32:10 multiple slices per large enterprise

32:12 customer and it becomes really

32:14 interesting when you start mixing the

32:16 private and public network because i

32:19 didn't comment on it earlier but i truly

32:21 believe that this is going to be the

32:23 largest swath of this is probably going

32:25 to be hybrid where we've got private

32:27 network resources that need to be on

32:29 site for survivability and performance

32:31 reasons think control on a factory floor

32:34 think medical you know surgeries

32:37 you can't there is no room for error in

32:38 those cases but

32:40 customers walking onto those campuses

32:43 are still going to need to have public

32:44 network access right so understanding

32:47 that and separating those two accesses

32:49 via slicing i think is going to be a

32:51 really cool use case and that's going to

32:53 be

32:54 that's going to be what our future looks

32:55 like

32:56 pass it off to the rest of the panelists

32:58 yeah we've only got about three minutes

33:00 left so i think what i'll do is ask each

33:02 of you in in responding to the time

33:04 frame for what when we'll see network

33:06 slicing to also add your own comments

33:08 about what are the two or three most

33:10 important things you want everybody to

33:12 take away so anna why don't you go next

33:15 okay

33:16 troy was spot on when it comes to the

33:18 management automation orchestration of

33:20 the network it's going to be tough their

33:23 physical network functions their virtual

33:25 network functions their containerized

33:26 network functions some are on arm based

33:29 hardware some are on x86 we have a slew

33:32 of vendors from wind river to red hat to

33:35 vmware that are out there it's it's a

33:38 little bit messy so having the right

33:40 knobs in place to orchestrate a slice

33:42 end to end is going to be a little tough

33:45 and that's why maybe troy is saying

33:47 end of 20 22 early 23 which seems

33:50 accurate right and the other point is

33:52 that right now it's a push technology

33:54 right we're creating the slice we're

33:56 trying to make it fit into some um some

33:59 customer need but if it becomes a pull

34:02 technology if there's some value that

34:04 could be created if there's a business

34:05 justification behind it where customers

34:08 really need the slice i guarantee you

34:11 network operators vendors will work

34:12 twice as hard to getting this out the

34:14 door so that's going to be the the

34:16 science that we have to figure out

34:18 behind network sizing constantine how do

34:21 you see it

34:22 yes i i completely agree with uh what

34:25 was said and uh i think troy hit the key

34:28 notes here

34:30 um

34:31 i also believe that it's going to be 20

34:33 late 2023 before we see end to end

34:36 slicing in all of its glory and i

34:38 believe we'll see dynamic

34:40 network slicing a bit more optimistic on

34:43 that

34:44 the takeaway that i want to underscore

34:47 here is that network slicing is not just

34:49 about 5g or 4g or 6g

34:52 network slicing is going to affect

34:55 everything in the networking industry

34:57 okay it's fundamental new way of doing

35:00 networking um and uh we're going to work

35:04 before uh we run in 5g right so

35:08 um that's the first takeaway the other

35:11 is the oran right the oran is i call it

35:13 the operating system of the radio right

35:14 now it opens new capabilities we're

35:17 going to see smart radios uh introducing

35:21 new features in the radio if we unlock

35:24 the secret of how we bring that into the

35:27 end-to-end

35:28 sla and ecosystem i think we're going to

35:31 to see new use cases that um

35:35 will be really disruptive

35:37 in the industry

35:38 um

35:39 so franco we've got about one minute so

35:43 we need to be ver be curious and brief

35:45 but what are your key takeaways

35:48 yeah they make make uh key takeaways

35:50 very similar i think that 2023 is a year

35:53 in which we start to see something

35:54 serious intentional dynamic fundamental

35:57 network slicing as info vista we have

35:59 launched these nla the network life

36:01 cycle automation that exactly the design

36:04 and from the designing time to the

36:06 testing the monitoring to their uh the

36:08 optimization exactly to be sure that we

36:11 can provide the

36:13 monitoring the capability to ensure that

36:16 the operator will deliver the network's

36:17 license in a dynamic way as the customer

36:20 the respect

36:21 and ray you get the last word um

36:24 you seem to have a vision for how this

36:26 all comes together but there are a lot

36:27 of pieces

36:28 well 12 to 18 months sounds right to me

36:31 i would focus on anon's comment this is

36:33 more likely to be pulled by the

36:35 enterprise than pushed by the operators

36:37 i don't say that because the operators

36:39 don't get it i say it because the money

36:40 is in the poll

36:42 with a large enterprise on the other

36:44 side

36:44 everyone will line up around that and it

36:47 will happen almost overnight think about

36:50 how fast the iphone changed this world

36:51 from a 3g to a 4g world a large

36:55 enterprise use case will pull slicing

36:57 into this network

36:58 faster than you can say network slicing

37:02 it's probably years ago

37:04 great with that i have to bring this

37:06 wonderful session to a close uh thank

37:08 you all it's been a great session and

37:10 let's hope we all get together to debate

37:12 this again in the next few months

37:15 have a great day thanks very much the

37:17 sheriff's license 23

37:19 yes yes