Network Function Virtualization and Network Service Insertion and Chaining
This is a modal window.
The media could not be loaded, either because the server or network failed or because the format is not supported.
Formal Metadata
Title |
| |
Subtitle |
| |
Title of Series | ||
Number of Parts | 199 | |
Author | ||
License | CC Attribution 2.0 Belgium: You are free to use, adapt and copy, distribute and transmit the work or content in adapted or unchanged form for any legal purpose as long as the work is attributed to the author in the manner specified by the author or licensor. | |
Identifiers | 10.5446/32558 (DOI) | |
Publisher | ||
Release Date | ||
Language |
Content Metadata
Subject Area | ||
Genre | ||
Abstract |
|
00:00
Principal idealStaff (military)Thermodynamisches SystemLecture/Conference
01:08
HierarchyThermodynamisches SystemComputer architectureWave packetComputer networkService (economics)Insertion lossMultitier architecturePoint cloudCartesian coordinate systemDatabaseCASE <Informatik>Web serviceCloud computingSoftware developerElasticity (physics)MultiplicationÜbertragungsfunktionVirtual realityWeb 2.0Multiplication signDataflowVelocityData centerScalabilityComputer hardwareGame controllerOpen setÜberlastkontrolleChainDifferent (Kate Ryan album)Server (computing)Connectivity (graph theory)Projective planeVariety (linguistics)Entire functionOrder (biology)DeterminismFunctional (mathematics)Data managementScaling (geometry)Computer networkRight angleLevel (video gaming)Frobenius methodXMLUMLLecture/ConferenceComputer animation
04:28
DivisorMultiplicationConnectivity (graph theory)Revision controlNeuroinformatikFunctional (mathematics)Virtual realityMultiplication signQueue (abstract data type)Server (computing)Connected spaceService (economics)Computer networkMedical imagingINTEGRALObject (grammar)InternetworkingHypermediaComputer networkPlanningVirtual machineData storage deviceÜbertragungsfunktionHorizonAuthenticationLecture/ConferenceProgram flowchart
05:29
SoftwareNatural numberProxy serverChemical equationComputer networkOpen sourceFunctional (mathematics)Virtual machineLastteilungTable (information)Computer networkOperator (mathematics)Computer hardwareWordAbstractionFirewall (computing)Virtual realityCartesian coordinate systemIPSecÜbertragungsfunktionService (economics)Virtueller ServerThermodynamisches SystemLatent heatLecture/ConferenceComputer animation
07:00
Channel capacityComputer networkCartesian coordinate systemMedical imagingSet (mathematics)Multiplication signVirtual realityÜbertragungsfunktionVector spaceFunctional (mathematics)Computer fileComputer animationLecture/Conference
07:57
Condition numberWorkstation <Musikinstrument>Library (computing)Functional (mathematics)Computer hardwareBitVirtual machineDigital libraryÜbertragungsfunktionComputer animation
08:28
ÜbertragungsfunktionCartesian coordinate systemHuman migrationPoint cloudMathematicsComputer networkService (economics)Front and back endsPropagatorFunctional (mathematics)Multiplication signMultilaterationLecture/Conference
09:16
ÜbertragungsfunktionStructural loadComputer networkComputer networkScaling (geometry)Virtual realityService (economics)Macro (computer science)Functional (mathematics)Scalar fieldVideo gameArithmetic meanThermodynamisches SystemPoint cloudSoftwareIntelligent NetworkElasticity (physics)Pattern languageFirewall (computing)Computer animationLecture/Conference
10:06
Event horizonWordPower (physics)Video gameComputer networkPoint cloudVirtual machineOpen setDataflowCommunications protocolSoftwareComputer animation
10:46
Order (biology)View (database)Computer networkComputer programmingService (economics)Category of beingGroup actionMereologyArithmetic meanBuildingSoftwareLecture/ConferenceComputer animation
11:35
Event horizonComputer networkVirtual machineDifferent (Kate Ryan album)Data centerBefehlsprozessorService (economics)Data storage devicePoint cloudChannel capacityNeuroinformatikSource codeOrder (biology)Lecture/Conference
12:30
TelecommunicationDataflowMessage passingDifferent (Kate Ryan album)Type theoryChainoutputComputer networkComputer networkService (economics)Open setGame controllerProduct (business)Thermodynamisches SystemPoint cloudEvent horizonVarianceServer (computing)Hand fanBitMotion captureOrder (biology)Connectivity (graph theory)FinitismusQueue (abstract data type)Data storage deviceDerivation (linguistics)MultiplicationLevel (video gaming)Computer animation
15:22
Data managementMessage passingQueue (abstract data type)Game controllerOpen setDataflowSoftware developerService (economics)Interface (computing)Point cloudEvent horizonStack (abstract data type)Lecture/ConferenceComputer animation
16:01
Focus (optics)Message passingGame controllerINTEGRALConstructor (object-oriented programming)3 (number)Right angleReading (process)Open setDataflowCustomer relationship managementDatabaseDaylight saving timeDevice driverLecture/Conference
16:50
Open setGame controllerInterface (computing)Message passingMaxima and minimaComputer networkToken ringDataflowRepresentational state transferDatabaseService (economics)INTEGRALPoint cloudCartesian coordinate systemDerivation (linguistics)InformationDimensional analysisSelf-organizationComputer animationLecture/Conference
17:26
Self-organizationWordCartesian coordinate systemDataflowGame controllerMechanism designConfiguration spaceRevision controlOpen setDaylight saving timeDevice driverContent (media)INTEGRALInternet forumOpen sourceConnectivity (graph theory)Latent heatDatabaseÜberlastkontrolleComputer animationLecture/Conference
18:21
Water vaporData managementService (economics)Partition (number theory)Buffer overflowCartesian coordinate systemInterface (computing)Game controllerPoint cloudDataflowOpen setConfiguration spaceÜberlastkontrolleRepresentational state transferDatabaseComputer animation
18:53
Execution unitEvent horizonData managementSelf-organizationOpen setDataflowFigurate numberSpeech synthesisÜberlastkontrolleGame controllerNeuroinformatikVirtual machineStack (abstract data type)DatabaseComputer networkComputer networkService (economics)Insertion lossChainConfiguration spaceLecture/Conference
20:01
MultiplicationMathematicsDifferent (Kate Ryan album)Service (economics)Computer networkComputer networkNetwork topologyComputer programmingEndliche ModelltheorieVirtual realityPoint cloudComputer animation
20:39
Water vaporRight angleInternetworkingUniform resource locatorType theoryGame controllerCorrespondence (mathematics)Computer networkWeb applicationCartesian coordinate systemOpen setFunctional (mathematics)ÜbertragungsfunktionDataflowVirtual realityTable (information)ChainService (economics)LastteilungConfiguration spaceDifferent (Kate Ryan album)Endliche ModelltheorieVirtual machinePoint (geometry)ConcentricAreaConservation lawComputer networkChemical equationProxy serverLecture/ConferenceComputer animation
21:41
Set (mathematics)Different (Kate Ryan album)Type theoryRight angleFunctional (mathematics)Computer networkDataflowVirtual realityComputer networkGame controllerCartesian coordinate systemOpen setÜbertragungsfunktionVirtual machineLecture/Conference
22:21
Representational state transferInterface (computing)Computer networkPhase transitionStack (abstract data type)Game controllerConfiguration spaceOpen setData managementDataflowCartesian coordinate systemÜbertragungsfunktionMotion captureConservation lawFunctional (mathematics)Virtual realityComputer animationLecture/Conference
22:55
ÜbertragungsfunktionDemonFigurate numberExecution unitFunctional (mathematics)NeuroinformatikGame controllerConfiguration spaceMessage passingService (economics)Chemical equationWritingState observerComputer networkDataflowComputer networkOpen setPeripheralFirewall (computing)Similarity (geometry)Speech synthesisFocus (optics)Web serviceData managementPhysical lawDatabaseÜberlastkontrolleWordDifferent (Kate Ryan album)Type theoryPlanningCartesian coordinate systemEvent horizonMotion captureVirtual machinePoint cloudLastteilungWeb 2.0Server (computing)Open sourceWeb applicationLatent heatLevel (video gaming)Communications protocolLecture/ConferenceXML
26:38
Computer networkComputer networkService (economics)ÜbertragungsfunktionChainFunctional (mathematics)Human migrationDynamical systemConfiguration spaceGeometric quantizationNeuroinformatikEvoluteWave packetOffice suiteLecture/Conference
27:35
Insertion lossStatisticsElasticity (physics)Computer networkService (economics)Functional (mathematics)Computer networkÜbertragungsfunktionComputer animationLecture/Conference
28:07
ÜbertragungsfunktionComputer networkOpen sourceService (economics)StatisticsBimodal distributionVirtual realityDynamical systemElasticity (physics)Game controllerCartesian coordinate systemSubsetFunctional (mathematics)Computer networkINTEGRALLecture/Conference
28:52
Game controllerDataflowClient (computing)Event horizonService (economics)Computer networkOpen setComputer networkSoftwarePoint cloudChainComputer animation
29:30
Service (economics)Computer networkInsertion lossWave packetSoftware frameworkEnvelope (mathematics)Functional (mathematics)ÜbertragungsfunktionAsynchronous Transfer ModeGame controllerConfiguration spaceVirtual realityCartesian coordinate systemChainLecture/Conference
30:17
String (computer science)Data managementSoftwareComputer networkMechanism designNeuroinformatikNumbering schemeConservation lawMedical imagingWordMessage passingDialectConfidence intervalState of matterÜbertragungsfunktionConfiguration spaceLecture/Conference
31:30
Functional (mathematics)Connectivity (graph theory)MeasurementComputer networkElectronic program guideState of matterMedical imagingCASE <Informatik>Lecture/Conference
32:19
Musical ensembleThermodynamisches SystemGroup actionComputer networkChainAsynchronous Transfer ModeSoftware testingÜbertragungsfunktionProof theoryFunctional (mathematics)MathematicsPolarization (waves)Lecture/Conference
33:02
Software developerLecture/Conference
Transcript: English(auto-generated)
00:58
Hi, my name is Balaji Padnala. I am from Freescale working as a principal staff systems
01:25
and architecture. So, today I am going to take you through network service insertion and chaining using OpenStack. So, this is the agenda we have. First I will take you through the overview and then network
01:42
function virtualization and SDN. And we will go through the cloud resource discovery and open flow controller and we have a couple of components in open flow controller, CRM consumer, flow manager and then we will see network service insertion and chaining and
02:04
the future work summary and then we have questions. So, let us see the overview like the need of network services in the cloud environment like we have the scalability, elasticity and service velocity and enabling a wide variety
02:25
of ecosystems. So, right now we have all the commercial hardware where you have to buy with lot of money. So, that they will be deployed in your data centers using cloud
02:43
and when you want to install a new service, the time to required to get the new service into the deployment takes more than you know switching the hardware and then making it suitable for the cloud deployment. And as if you want to increase the have the
03:01
scalability and elasticity and again the same the time for the project it takes more when compared to the hardware. So, the network services using network functions using virtualization ok. So, there we can have we can reduce the time for you know service
03:24
velocity and we can have scalability and elasticity as well. So, multiple deployment use cases for a tenant like application development and the entire application architecture and we have a hybrid cloud deployment. So, based on the tenant
03:42
requirements we can deploy these network services for example, we have a three tier web services tier and application tier and database tier and we have a different network services required to handle the traffic coming on to the web tier and then the traffic generating from web tier to the application tier and again the from application tier to the database
04:06
tier that entire application architecture. And also we have the hybrid cloud deployment where we have multiple you know R&D applications running in the cloud and we have the web servers running for general purpose use for the customers like that. So, the need
04:26
of network services for these different tiers are different as such. So, just I will give a glance at the OpenStack like we have multiple components of OpenStack like Horizon, Keystone and we have Nova. Nova we have Nova scheduler,
04:46
Rabbit time queue server and we have Nova API and Nova compute and we have Neutron for network function, network virtualization and connectivity services and we have Glance for image storage and retrieving the virtual machine images and we have other components
05:03
like Cinder, Horizon and Keystone for authentication and Horizon is for the dashboard where we can configure the services and networks and we have a Cinder for object storage.
05:22
So, what is the network function virtualization like we are how do we virtualize the network function? So, running any network function within a virtual machine for example, let us take an open source softwares like HAProxy, it is a server load balancer
05:42
which we can run it inside the virtual machine and we can run you know for IP tables firewall we can use it as a virtual firewall network function virtualization firewall. Similarly, we can use any VPN applications as such like IPsec etcetera. So, basically it reduces the expenditure and operational expenditure there is a primary
06:06
advantage and we can replace the custom hardware replacement like we do not need any specific branded networking gear where we can use the existing virtualized environment, virtual server environment to deploy these network services into the tenant you know networks
06:27
and we have SDN networks and the non-SDN networks like network function virtualization does not need as such SDN networks like software defined networks it can work on the non-SDN
06:40
networks as well and NFV appliances deployment is very much complementary to SDN networks and because the SDN networks give the flexibility of network abstraction where we can use this network abstraction to deploy these NFVs and they are very well together you know complement
07:02
each other the tech these two technologies. So, the network and application deployment without the need for upgrades of network devices like we do not need to upgrade you know if you want to increase the capacity like you do not need to upgrade a new device as such but you just need to change the image itself that is enough. So, we have tremendous advantages on
07:26
having the NFV and you know the coming up the SDN networks are complementing this NFV deployments where we can save a lot of money and we can definitely showcase a
07:42
better performance. So, the NFV challenges we have is like the performance first of all. So, when we say that network function virtualization the network function running on a specific hardware versus the network function running inside a virtual machine.
08:06
So, obviously we have you know the lot of bottlenecks where the pack is going through the hypervisor and then the virtualization library and then to the virtual machine the function inside the virtual machine running. And again due to this the deploying of this
08:28
network services the backend applications which the end user is going to access. So, the impact of performance of network function directly gives impact on the access
08:41
performance of the application as such. Performance is the major challenge and we have the another challenge like migration. So, we have the application migration where you know whenever the application migrates from one cloud to another cloud or you know and the policy changes
09:07
which required you know for these network functions those policy changes as well has to migrate along with the application. So, migration is another biggest challenge to implement these network function virtualization. And we have non-SDN networks for scale up and
09:27
scale down and automatic initialization of network services based on traffic load. Like the elasticity where the traffic is low we do not need the network services the
09:42
many network services. So, we can have a few firewalls running and we have a heavy load running into the system we can scale up these network services based on the traffic patterns to the cloud of the tenant. So, what is the this is the cloud resource data.
10:05
So, we will see the software defined networking. So, what is the cloud resource data? The cloud resource data is the events generated when we create when we launch a virtual when we spin a virtual machine or when we create our network and you know
10:27
any other like when you add a network when you create a port when you delete a port. So, these are the different events and the data generated along with these events are very useful like it is. So, we have the again the open flow protocol for SDN software defined
10:45
networking like we can define then the networking like how do we want to do with the data. So, like when we create a network we will store that network data into a database and
11:01
we can take the appropriate action based on that data. The intelligence build surround the data is very much important for defining the networking. So, the real meaning of the software defined networking like we have to define the networking like we have to program the network and how do we program the network. We have we need the data which is required
11:24
and we want to use the data in a specific manner you know based on our requirement based on the tenant requirements. For example, scaling up the network services scaling down the network services similar to that and we have the compute events and data.
11:42
For example, like whenever we create a virtual machine the data will be generated like what is the compute host on which the VM is getting created and whatever the host node details what is capacity of the host node and what is the storage of the what is the RAM what is the CPUs and all
12:00
the host node details and whenever a new host is added into the network into the data center and that details as well required to store in the cloud resource data. And we have network events and data for example, like creating a network creating a subnet and
12:21
creating a port and creating the vxlan ports on OES in a compute node. So, these are the different data events and data which we are going to collect together and build an intelligence and we have to we will define the networking for launching these network services
12:41
and chain. So, this is the high level picture of cloud resource data. We have the cloud resource NOA and Neutron. NOA is for the spinning up the VMs and we have Neutron for networking
13:02
and we have the cloud resource data which is required where we are going to capture all the events generated by these two components and we are going to store that data into the cloud resource derivation service database. And then we have if you see in the picture
13:24
we have the open flow controller and we have the open flow controller and we have the another component called CRD consumer which is required for capturing this data generated from the open flow control open stack controller to the open flow controller. And we have the different compute
13:44
nodes connected to the open stack controller using open v switch. So, all this data communication the communication between the cloud resource data and open flow controller is through the
14:03
message queues. We have the RPCs, JSON messages and we have different notification finite exchanges where we will create the queues to capture this data from the cloud resource data to the open flow controller. So, we have the network service infrastructure where we can
14:22
create the chains network service chains and we can launch the network service chains integrated into the cloud resource data as a plugin. And we have two different features like layer 2 chains and layer 3 chains and we have the CRD consumer where we will consume the data
14:43
generated by this NOA Neutron through the cloud resource derivation service to the CRD consumer. So, the cloud resource data relay like the data which is captured from these events
15:06
these are relayed to the cloud resource data consumer where through the notification finite exchange the open flow queue for tapped NOA Neutron messages and fan out queue for each OF
15:22
controller like we can have multiple open flow controllers also deployed in your network. So, there is a common queue for the management messages from open flow controller to the cloud resource discovery and we have the cloud resource data consumer where there is a direct
15:41
interface between this cloud resource data consumer to the open flow controller. So, that the data collected from CRD cloud resource data of open stack controller will be passed on to the open flow controller. The CRD consumer will receive JSON messages
16:00
from CRD service all the events are through the JSON messages where that JSON message can be converted into the messages which are required for the open flow controller. Like we can have open flow controller like open daylight or any third party or any commercial vendor as such just we have to write the corresponding driver where this JSON messages
16:25
are converted into the messages which can understand by the open flow controller. And the CRM consumer configures this open flow controller database.
16:40
It configures all the data received generated from the open stack controller using the integration API. So, we will see the CRD consumer on the open flow controller where there is a REST API interface between the CRD and the open flow controller.
17:04
The initial messages like whenever the open flow controller comes up in the network, the initial messages will be sent to the open stack controller cloud resource derivation service where it will store all the open flow controller information in its database
17:24
and integration of open flow controller application service. The open flow controller can have multiple applications as such like it can have L4, L7 applications also in open flow controller and also we have
17:41
the version control mechanism where the configuration versioning also is maintained in the database. The integration drivers for vendors specific or open source open flow controller like open daylight. So, where it is not only a specific open flow controller,
18:04
we can have it for the any commercial vendor as such a vendor open flow controller where we have to write our own drivers to integrate to capture this data and configure the open flow controller. So, this is another component like flow manager.
18:24
Flow manager application open flow controller application service where this open flow controller application service is a REST API interface to capture the data and configure the open flow controller. The configuration of cloud resources like whatever
18:42
data captured from Novanutron through the cloud resource data can be configured into the open flow controller and we have another database running in the open flow controller where we store all this data and we have versioning support. The flow manager like the configuration of these events from open stack controller and
19:06
putting it in the database of open flow controller. The events generated from those configuration of database in the open flow controller will be given to the flow manager
19:23
where the flow manager will build the intelligence like creating the flows on the OES switch on the compute nodes. So, based on this the virtual machine data, VM data from Nova and the networks and ports data from Neutron will be stored in the
19:47
open flow controller and open flow controller the flow manager in the open flow controller will capture the data and create the flows required and configures the open V switch on the compute nodes. So, we have the network service insertion chain I think yeah so we have the
20:18
virtual network topologies that layer 2 and layer 7 networks. The tenant network isolation like
20:25
how do we generate create the isolation of traffic between different tenants in the cloud deployment and defining the networks through programming. The network services in multiple deployment model like we have edge node where the network node
20:45
in the open stack where the traffic comes from the internet falls onto the network node and then goes to the corresponding web application in the VM. And we have the another type of network application like OF controller the layer 2 layer 7 support of open
21:04
flow controller and we have a third type called virtual network functions like running a network function inside a virtual machine. So, these are the three different models of network services where we can create the services we can insert the services and
21:22
chain the services together. And the configuration of these network services running inside a virtual machine is another important point where for example we have the server load balancer which is HAProxy it is having its own configuration like you
21:41
have to configure all the pools of data and VIP etc. And similarly for IP tables we have all the policies required this can be different in different types of deployment like for example policies are different in the network node policies are different for OF controller network applications deployed using open flow controller networks or the virtual network functions
22:07
based on the tenant networks. So, how do we how do we relay the configuration from open stack controller to these virtual functions running inside a virtual machine.
22:22
So, for edge network node we have the JSON interface the just a rest API interface where we can configure relay this configuration data from open stack controller to the edge network node. And we have a resource manager for open flow controller network applications
22:42
where we capture all the data from the configuration data of network functions to the open flow controller network applications. And for virtual network functions we have the relay agent. The relay agent is just another daemon which is running on the compute node
23:02
where the network functions are running. So, when you configure the network function from open stack controller the configuration will get relayed through this JSON message RPC messages and it will capture by the relay agent on the compute node and the relay agent will again use the libword what I was serial to configure the network function inside the virtual machine.
23:26
So, these are the three different types of network services which we are required and the configuration are used for these network services.
23:40
So, this is the high level picture for network services. So, we have the open stack controller, we have open flow controller and open flow controller we have the cloud resource discovery consumer and open flow controller application service and we have a firewall web application firewall and the server load balancer. We have the db where we configure all the data
24:06
captured from this neutron nova through the cloud resource discovery and the events from the database will be captured by the flow manager and flow manager using open flow protocol is going to configure the open v switch of compute nodes.
24:25
And we have the network services like firewall web application firewall and server load balancer they are running on the compute they are in a virtual machines inside the virtual machine running on a compute node and we have a web servers running on another compute node.
24:45
So, the corresponding flows which are required based on the events generated will be captured and the flows will be pushed from the flow manager to the open v switch of compute nodes where the network function and network services are running and as well on the plane applications
25:07
like web applications or applications, web servers or applications or database type the three type. So, this is how the events generated by the nova and neutron will be captured
25:24
and stored in a database of open flow controller and the intelligence built with this data from the database by the flow manager and pushes the flows required from to the open v switch on the
25:42
compute nodes. So, the flows will be configured from the flow manager whenever there is a miss packet generated by the open v switch connected to the open flow controller.
26:02
The network services here we have those are all vendor neutral like we have a network services can be from different vendors together like or any open source applications as like we do not need to have use a specific vendor for network services and the accelerators
26:26
to for virtual functions. So, we need accelerators for these virtual functions to compare the performance of these virtual functions from the physical devices. So, we have the network service chaining infrastructure where we can configure the
26:47
network functions and we can create the chains and we have dynamic networks creation for chaining and we have the configuration the network function configuration where we will map the
27:01
configuration and map these configuration to the network service chain and then we will deploy the chain and during this network service launch chain launch we have different issues like we have the same host or the network functions inside a chain are two functions in one compute
27:22
node and another function in another compute node and also we have the migration support where we will the network functions will be migrating from one compute to the another compute yeah the future work which we are planning to have is the layer 3 network service insertion
27:45
and chaining support and then network service statistics and elasticity the statistics of
28:01
networks network functions based on the statistics generated by these network functions the network services can be scaled up or scaled down and elasticity of these network services and dynamic launching of these network services based on the network function statistics and
28:21
multimode network services support edge networks like network node and the OF controller application support for l4 l7 applications and network services using a network function virtualization and also to improve the performance of network functions the PCA through support for network service VMs yeah a wide acceptance of OpenStack from commercial
28:52
and open source communities so that's why we started using OpenStack and there is a need for cloud resource data like to capture these events and use it for you know SDN kind of networks
29:08
to deploy the network services chain and launching and bringing in the OpenFlow controller into the OpenStack deployments is very much needed for
29:24
so that the software defined networking will be of you know better use and the network services insertion and chaining framework where we can create the network network chains and insert the network services and configure these network services
29:44
and washing off this network services configuration as such and multi-deployment modes like deploying these network services in a network node or deploying these network services inside OpenFlow controller applications
30:04
or deploying these network services as a network function virtualization yeah these are the guys behind this work like we have Srini Adeepalli from
30:21
he's a fellow and chief software architect and we have the manager which are Sarath Bhavu and we have the team Trinidad, Veera, Srikanth, Sairam and Ram Prasad together to make this happen I'll take up questions
31:05
We have the configuration relay mechanism where we have the relay agent which is running on a compute node so the configuration done from OpenStack dashboard is relayed through the JSON messages
31:22
to the compute nodes and the relay agent is going to capture these messages and configure the network function so it's a pretty good like we don't have any how many milliseconds we didn't count as such we didn't see but it's well we didn't see a big delay
31:48
I think it's a few milliseconds yeah if that is the case then it is true for
32:03
all the components in the OpenStack as well right we have nova and neutron nova which will you know will launch the images and neutron which will create the networks they are coexisting together and we don't see anything
32:40
we are in the the proof of concept mode right now yeah so we are not in the deployment just
32:46
we made a couple of network functions as a chain and we deployed in our test networks okay thank
33:05
you