Applying Cloud Principles to Transforming Metro Networks

Introduction

0:03 hi everybody and welcome to our Cloud

0:07 Metro session now and uh probably going

0:09 to talk about a little boring thing old

0:11 school networking no more applications

0:13 and services but though those networking

0:16 are pretty much driven by requirements

0:19 which are coming from those new class of

0:22 applications and yeah I'll try to stick

0:24 with the with the title of this

0:27 presentation we'll talk about the cloud

0:28 principles which can be applied to the

0:31 metro to transform it which is

0:33 um under pressure of those 5G

0:37 Razer 5G and actually

0:40 um the influence of

0:42 different types of new class of

0:44 applications and content which are

0:46 placed in the

0:47 locations in the matter where it never

0:49 have been before and

Cloud Principles

0:53 in general that requires to transform

0:56 the traditional Metro where it is just a

0:58 maybe dump pipe

1:00 created in the first place to transfer

1:03 traffic north south so into something

1:05 more sophisticated more organism which

1:08 allow us to serve more complex traffic

1:11 Partners importantly sustain the huge

1:14 traffic growth which is uh

1:18 um driven by the 5G to large extent and

1:21 also which is another important thing to

1:23 connect applications or maybe okay we

1:26 can have Services which provide

1:27 applications at any literally location

1:30 of those Metro infrastructure so that's

1:33 one Trend the second trend is actually

1:35 virtualization of the network functions

1:39 of the

1:41 transport infrastructure itself of the

1:43 network itself and this is uh something

1:45 we see that happening all across the

1:47 place and 5G is the brilliant example

1:50 with all the virtualized she used to use

1:53 virtualized evolve packet core which in

1:56 the context of the x-hole will be placed

1:58 all across the network from access to

2:00 the more Central Central locations and

2:03 in fact it happening in other places uh

2:06 cable is another example so those guys

2:09 are trying to digitize their R5 function

2:12 also virtualize cmts

2:15 traditional Broadband aggregation so

2:18 that's where the former monolithic BNG

2:22 transforms slowly towards the newer caps

2:25 models with disaggregated BNG control

2:28 and BNG

2:31 um data plane right and now we can take

2:33 it and drop it in any place of the

2:35 matter at least that's the that's the

2:37 intent

2:38 and

2:40 um

2:41 the point here that many of those

2:43 virtual functions already now coming in

2:46 the form of the cloud native

2:47 applications and of course they're

2:49 coming with the with the thought behind

2:52 that okay so we can move them freely and

2:54 orchestrate them as flexible as we do

2:56 orchestration of the services and

2:58 applications in the cloud and even so

3:00 where are we moving application from One

3:02 Cloud to another

3:04 around and and if we think about the why

3:08 the concept works so well for the cloud

3:10 is it's pretty much for the fact that

3:13 the cloud is based on the very well

3:15 thought

3:16 um data center fabric whereas well

3:19 separation between the undelay and

3:21 overlay tunneling which in its turn

3:24 based on the closed-like topology which

3:26 is relatively simple or at least well

3:27 understood right and that's what creates

3:30 the foundation for freely moving those

3:33 applications all across my

3:35 infrastructure so the first concept

3:38 which we can actually Implement for the

3:40 Metro is about this converting the

3:43 highly segmented and solid Metra

3:47 infrastructures towards the end-to-end

3:50 IP service fabric right so here we

3:53 probably not talking about the

3:55 converging the metro into gigantic close

3:59 topology so the material will have still

4:02 lots of inertia and there's some

4:05 economical reasons why all types of

4:08 topology will be will contain to stay in

4:11 mattress or rings partial mesh Etc but

4:14 the key building block of the mattress

4:16 of the routing system

4:18 which independently or what size

4:21 in an amount of terabytes it's it's

4:24 capable to uh to put through it so it

4:28 should provide the same set of

4:29 networking capabilities which allows to

4:32 enable the function of the

4:36 IP service Fabric in any at any location

Example

4:40 and if we double click on probably on

4:44 more or less common example of how the

4:47 metro metro Network looks like for the

4:50 service provider so we see this kind of

4:53 layers access aggregation Metro services

4:56 and services on the top so there's a

4:59 multi-service edge of the talk with

5:01 access which interconnects our terminals

5:04 that might be cell towers business

5:07 customers some residential customers and

5:10 at the very top that's where the Metro

5:12 means the other part of the network

5:14 infrastructure of the service provider

5:16 would be when internet public and

5:18 essentially Telco Cloud so that's that's

5:21 the type of network which we can see

5:24 today and the transformation

5:27 would happen with the first step on

5:29 Telco Cloud moves closer to the to the

5:33 access right so with that we're also

5:36 transforming

5:37 the aggregation layering to something we

5:40 call the lean Edge so the lean Edge

5:42 function will be provide all the same L2

5:45 L3 vpns which we know for now so it will

5:49 enable immediately the East-West traffic

5:51 pardons it will enable the low latency

5:54 Services the application can be placed

5:56 on this Telco Cloud infrastructure

6:00 with respect to the existed multi-store

6:02 research we believe it it stays where it

6:05 is for quite some time it allows a

6:08 smooth transformation from the older to

6:10 the new network architectures

6:12 and it will serve for some Corner cases

6:15 where we need to teach some VPN Services

6:18 which are not or not stitchable just

6:21 with the use of the lean Edge platform

6:23 because lean Edge is a lineage so it

6:25 still comes with some restrictions right

6:27 and on the next step what will happen is

6:30 just we'll need to scale out the Telco

6:32 Cloud infrastructure by placing more

6:35 applications by terminating more traffic

6:37 streams at this location

6:39 and also to scale to scale the part of

6:44 the aggregation Network and here as a

6:46 matter of fact we'll see the Rays of the

6:48 type of a close approaches because they

6:51 say scale out approach is very much

6:54 applicable for this for this one but

6:56 still will be restricted by the by the

6:59 location of a particular Central officer

7:01 or maybe a point of presence in the

7:03 network

7:04 and the very last step

7:06 is moving the compute Edge to the very

7:09 down in the axis right and and here we

7:12 see a couple of interesting applications

7:14 specifically coming from the area of the

7:17 5G networking and specifically a

7:20 virtualization of the salesite router

7:23 function of the network infrastructure

7:25 itself and placing it to the compute

7:27 node sitting next to the next to the vdu

7:30 for example so we believe there are

7:32 really advantages from the toll cost of

7:35 ownership in some dense 5G 5G

7:41 uh

7:42 deployments

7:44 and

7:45 the statement here is

7:49 we can see that the Metro meets the

7:51 element of the Telco Cloud practically

7:53 in every point of this network

7:54 infrastructure and here what what where

7:56 are we talking about that the Metro

7:59 becomes the new service edge of the

8:01 service provider Network

8:04 and so far so good so we've built our IP

8:08 service fabric uh the only the only

8:11 drawback of this thing is the relatively

8:13 High complexity of the network which we

8:15 facing over here it is higher than what

8:18 we would expect some simplification of

8:21 this of this designs will come through

8:23 decreasing amount of protocols so we're

8:26 migrating towards the sar from mpls and

8:28 then to this rv6 there will be some

8:31 hierarchical designs which also simplify

8:33 the implementation and troubleshooting

8:35 of this network infrastructure but

8:38 another thing which I want to talk here

8:40 which relates to this enabling

8:42 simplification is actually the network

8:43 slicing is the another concept which

8:46 we're proposing within the cloud Metra

8:49 vision

8:50 and probably in the 5G world the network

8:53 slicing raises relatively naturally so

8:56 for example there is a case we need to

8:58 stitch the 3gpp slices to the slices and

9:00 network infrastructure in the broader in

9:03 the broader use cases may not so obvious

9:06 but if we take some retrospective view

9:09 back into the some dark times when the

9:12 operator may may

9:14 be in possession of the multiple

9:17 physical infrastructure one

9:19 infrastructure to use to provide some

9:21 particular use case a particular type of

9:23 connectivity so this thing

9:26 used to be

9:28 simple but very costly and at those

9:31 times the industry passed through

9:33 transformation towards the single single

9:36 physical infrastructure which where the

9:38 old type of connectivity Converge on top

9:41 of the same packet switch Network and

9:43 well that's pretty much the stage where

9:45 we are right now so the total cost of

9:47 ownership or the the high total

9:49 consumption problem have been solved but

9:52 that's where the complexity uh was

9:55 introduced and from this perspective

9:57 they're introducing

9:59 the set of abstract layers might be a

10:03 very natural natural step

10:06 to be implemented to be implemented

10:08 deployed in the network so it preserves

10:10 the all the concept of of having this

10:12 single physical infrastructure uh with

10:16 the virtualized layers of networking

10:19 which also introduces the ability in a

10:22 simple weight or in a simpler way to

10:25 provide a differentiated class of

10:26 service to different types of traffic

10:29 flows and traffic flows belongs to the

10:31 engineerative by different applications

10:32 of course right

10:35 so this is about the network slicing and

10:39 we need to jump to the automation

10:41 and actually the network slicing by

10:43 itself is not the feature or function of

10:45 the transport infrastructure only and

10:47 it's probably not possible to deploy it

10:50 without

10:52 the automation complex and here we

10:55 prefer to talk about the closed loop

10:57 automation which is another another

11:00 principle inherited heavily from from

11:03 the cloud networking right which in most

11:06 cases are highly automated who automated

11:08 is a different question actually so for

11:12 now we're able to demonstrate very

11:13 practical examples of closed loop

11:16 automation which is based on collecting

11:18 and analyzing data from the from the

11:21 network comes which comes from the whole

11:22 different source and then Loop it back

11:25 through the actions of changing some

11:26 traffic engineering policies like in

11:28 this example but um

11:31 maybe the the actual the actual value of

11:36 this closed loop automation concept will

11:39 come when we introduce also as a I

11:44 algorithm as working as part of those

11:46 closed loop automation story right and

11:49 the whole Loop starts with the

11:51 provisioning or intent based

11:53 provisioning which is just the beginning

11:55 the real value is actually of AI comes

11:58 as per our opinion when it comes to

12:00 troubleshooting right so the formula for

12:04 the future networking in this concept

12:06 has continuously measure the user

12:09 experience continuously measure the

12:11 network service quality through the

12:13 embedded service Assurance mechanism and

12:16 compare it with the other patterns and

12:20 anomalies which the EI can derive

12:23 through analyzing of big data and

12:25 finally come up with the recommendations

12:27 about the root cause problem yes and

12:31 we believe that it will happen a way

12:34 faster than in certain cases we have to

12:37 spend on troubleshoot our networks as of

12:40 today and that would be the true and

12:42 very powerful step towards the as a more

12:44 sustainable operations in this space

12:48 all in all this is part of the concept

12:51 and part of the vision for this uh

12:53 metric Cloud Metro

12:56 um all in all we are pretty much in the

13:00 stage of the honeymoon with the eye

13:02 when the world is Rosy and everything is

13:05 possible at some point in time we

13:07 probably will see how the closed loop

13:09 automation will try to will allow to fix

13:12 us directly the problem in the network

13:15 but

13:16 yeah we'll see so the AI is powerful in

13:19 fixing things as creating harmful to

13:21 your network definitely uh but

13:24 definitely we can level we can leverage

13:26 for it's for some analysis in the

13:29 network sorry

Conclusion

13:31 okay and just the last slide to conclude

13:36 on

13:37 um a little bit expanding the concept of

13:40 the cloud Mentor Vision uh Beyond only

13:43 those Concepts which can be inherited

13:45 from the cloud so it comes to us to

13:47 three major areas operations systems and

13:51 architecture automated and programmable

13:54 operations and they P service fabric as

13:56 I said were inherited from the cloud and

13:58 when it comes to the systems or routing

14:00 system which actually consists you the

14:02 network so that's obviously one of the

14:05 most important component here and to

14:08 transform to the newer Metra infra with

14:12 new and new set of platforms which are

14:14 cost optimized which are feature Rich

14:15 which are power optimized and which can

14:19 sustain the traffic growth in this

14:21 segment of the network for a very very

14:23 long time so the longevity all those new

14:25 platforms being placed in the network

14:27 should spend up to the 7 on 12 years

14:30 according our idea and this is way

14:33 higher than what we saw in the past in

14:35 the same segment

14:37 and probably once again to repeat the

14:41 the major goals here for this Cloud

14:45 Mentor is to implement and simplify the

14:49 network deployment and network

14:50 operations because the level of

14:52 complexity is really really high and

14:54 that's something that will allow us to

14:56 enable new types of a better experience

14:59 for two major stakeholders in this

15:02 picture which is the end user and

15:04 service provider

15:06 for sure and just by the way since their

15:09 sustainability goes as a red line

15:12 through the through the all they all the

15:15 event I should say that the steps in

15:18 this diagram represent the

15:21 every one step brings you closer to the

15:25 uh to the cloud matter and also

15:27 contributes in into one

15:30 into one aspect of the sustainable

15:32 operational sustainable sustainable

15:34 Network which would be either a business

15:36 growth or people or trying to build a

15:40 more Greener company or more Korean

15:42 Network

15:44 so all in all

15:45 this is everything I have for for you

15:48 today

15:49 thank you very much great attention

15:56 [Music]