EPG Architecture
EPG Architecture
Disclaimer
The contents of this document are subject to revision without notice due to
continued progress in methodology, design and manufacturing. Ericsson shall
have no liability for any error or damage of any kind resulting from the use of this
document.
Contents
1 Platform Architecture 1
1.1 Deployment Types 2
1.2 Virtual Machines 2
1.3 Cloud Environment 3
2 Networking 4
2.1 Logical View of VNs 4
2.2 Admin Networking 8
2.3 Internal Networking 8
2.4 External Networking 9
2.5 Virtual Networks 9
3 Storage 11
5 Application Architecture 13
5.1 Startup Procedure 13
5.2 Control Plane Applications 13
5.3 User Plane Applications 15
5.4 EPG Deployments 16
1 Platform Architecture
Virtual EPG
CEE
VMs VMs
VRP VRP
VSFO VSFO
Hypervisor Hypervisor
Generic
OpenStack
HW (x86) HW (x86)
Switch Switch
VRP
Control Plane VSFO
Router Router
External PDN
The virtual EPG can be deployed in the cloud system using Heat. The virtual EPG
supports any dynamic deployment that follows the rules specified in Deploying
Virtual EPG.
For more information on how to configure the VMs and deploy the EPG, refer to
Deploying Virtual EPG. For more information on scaling the EPG, refer to Scaling
The EPG.
1.2.1 VM Types
The virtual EPG uses the following VM types:
— VRP: The Virtual Route-Processor (VRP) serves as a node manager for the
virtual EPG application that performs cluster supervision, software
distribution and provides O&M. There is one active VRP and one passive VRP
for redundancy.
1.3.1 Hypervisor
Kernel-based Virtual Machine (KVM) hypervisor is an open-source software
running on top of physical hardware. It is used to provide virtualization layer in
Ericsson Cloud System (ECS). It supports live Motion, that is, moving a VM from
one physical node to another with almost no disruption in service. In the KVM
hypervisor, there is a software-based switch called Virtual Switch (vSwitch). A
vSwitch allows one VM to communicate with other VMs. For more information on
Hypervisor and vSwitch, refer to CEE documentation.
1.3.2 Hardware
Intel® x86 Compute hosts are used as hardware. For full details on the type of
hardware used, refer to CEE documentation.
2 Networking
The VNs used in the virtual EPG depend on how the virtual EPG is deployed, as
shown in Table 1.
For more information about software versions, see Deploying Virtual EPG.
NM NM Control
RP RP Plane
MATE
BP
vFAB
MGMT VMname_EXT-1
O&M Ext
Figure 3 shows the logical view of the VNs if the virtual EPG is deployed in CEE or
generic OpenStack using software version 1 in a standalone Control Plane
deployment.
NM NM Control
RP RP Plane
MATE-1, MATE-2
DBG
BP-1, BP-2
vFAB
MGMT VMname_EXT-1
O&M Ext
NM NM User
RP RP Plane
MATE
BP
vFAB
MGMT VMname_EXT-1
O&M Ext
Figure 4 Standalone User Plane Deployment: Logical View of the VNs Deployed
in CEE or Generic OpenStack using Software Version 2 (or later)
Figure 5 shows the logical view of the VNs if the virtual EPG is deployed in CEE or
generic OpenStack using software version 1 in a standalone User Plane
deployment.
NM NM User
RP RP Plane
MATE-1, MATE-2
DBG
BP-1, BP-2
vFAB
MGMT VMname_EXT-1
O&M Ext
Figure 5 Standalone User Plane Deployment: Logical View of the VNs Deployed
in CEE or Generic OpenStack using Software Version 1
NM NM Control User
RP RP Plane Plane
MATE
BP
vFAB
MGMT VMname_EXT-1
O&M Ext
Figure 7shows the logical view of the VNs if the virtual EPG is deployed in CEE or
generic OpenStack using software version 1 in a single node deployment.
NM NM Control User
RP RP Plane Plane
MATE-1, MATE-2
DBG
BP-1, BP-2
vFAB
MGMT VMname_EXT-1
O&M Ext
The Backup and Restore Management (BRM) function uses the Admin interface
to back up the normal O&M service and to restore backup files.
The BP VN or BP-1 and BP-2 VNs are used for internal signalling between all
virtual EPG VMs.
The vFAB VN is used for forwarding external control signalling and user data
traffic between VSFOs.
The DBG VN is reserved to use for internal debugging. The DBG VN is not
reserved if the virtual EPG is deployed in CEE or generic OpenStack with Heat
using software version 2 (or later).
For more information on external networking, see Configure Virtual EPG External
Network Connectivity.
For more information about SGi-LAN service chaining networking, see Service
Chaining.
The virtual EPG requires that the VM virtual Network Interface Cards (VNICs)
and VNs are connected in a specified order. The order depends on how the virtual
EPG is deployed.
Table 3 specifies the required connections between VM VNICs and VNs if the
virtual EPG is deployed in CEE or generic OpenStack using software version 2 (or
a later software version).
Table 4 specifies the required connections between VM VNICs and VNs if the
virtual EPG is deployed in CEE or generic OpenStack with software version 1.
3 Storage
The virtual EPG supports local storage by using the physical disk on the compute
host where the VM is running. For information on storage, refer to Managing
Files.
The virtual EPG guest OS is NUMA-aware and the virtual EPG application uses
resources accordingly. However, there is a significant capacity penalty and a
complicated integration procedure when user-plane VMs (VSFO as user plane)
are deployed across NUMA boundaries. Control plane VSFO can span two or
more NUMA nodes.
It is recommended to ensure that all VCPUs and the memory of user plane VSFOs
are mapped to physical CPUs and memory on a single NUMA node. In addition,
for maximum capacity of the vSwitch, the vSwitch threads and the physical NIC
need to be on the same NUMA node as the user plane VSFOs. A reference
deployment for user plane VSFOs following this recommendation is called the
compact deployment, described in the virtual EPC documentation.
5 Application Architecture
The EPG application software consists of control plane and user plane
applications running on the VSFOs. An allocation of the application types on the
VSFOs takes place during application startup. Based on configuration, each
VSFO is assigned as control plane or user plane. For a description of the startup
procedure, see Startup Procedure on page 13.
1. The EPG application software is loaded on the VSFOs during the deployment
of EPG. For more information on deploying EPG, see Deploying Virtual EPG.
3. The retrieved configuration data is used to determine which VSFOs run each
of the application types. Roles of VSFOs are determined and applications are
started with the needed configuration data.
For more information on the EPG control plane functionality, see PGW-C
Overview and SGW-C Overview.
5.2.1 GSC
The following are functions of the GSC:
— Admission control
The GSC can run on any control plane VSFO, but only on one control plane VSFO
at a time.
5.2.2 PSC
The PSC establishes and controls user sessions towards the APN networks by
processing GTP-C messages received on the Gn/Gp, S5/S8-C, GTP-based S2a,
and S2b interfaces. The PSC also establishes and controls connections towards
external supporting nodes, such as the Online Charging System (OCS) and the
Event-Based Monitoring (EBM) server.
For information on the port range used by the multicore architecture on the
PGW-C interfaces, refer to Routing.
5.2.3 SSC
The SSC establishes and controls user sessions by processing GTP-C messages
received on the S4-C, S5/S8-C, and S11-C interfaces. The SSC also establishes
and controls connections towards external supporting nodes, such as the
Charging Data Function (CDF) and the EBM server.
The SSC handles charging data generation for SGW-CDRs and Rf ACRs.
each logical interface IP address to the surrounding nodes in the same network
through signalling messages. For information on SGW-C networks with single IP
addresses, see EPG Deployments on page 16.
For information on the port range used by the multicore architecture on the SGW-
C interfaces, refer to Routing.
For more information on the user plane functionality, see User Plane Overview.
— Packet inspection
— Usage measurements
— Policy enforcement
— Packet enrichment
— Redirection
The Data Plane also serves as the Service Function Forwarder (SFF) when
service chaining is used.
The PFCP End-Point is stateless. When handling signaling for a user session, the
user session information is firs retrieved from the Database. Once retrieved, the
signaling is then processed and the Database updated. Finally, the user session
information is written back to the DB.
5.3.4 Database
The user session information is stored in a distributed replicated Database to
provide fault tolerance if there is user plane VSFO failure.
The control plane node can be connected to one or several user plane nodes. For
more information, see PGW-U Selection Guideline in the PGW-C and SGW-U
Selection Guideline in the SGW-C.
The minimum configuration for a standalone control plane node is two VRPs and
two control plane VSFOs. All control plane VSFOs run the PSC application when
the PGW-C is configured and the SSC application when the SGW-C is configured.
If both the PGW-C and the SGW-C are configured, the control plane VFSOs run
the PSC and the SSC applications simultaneously. In addition to running the PSC
application, the SSC application, or both, one control plane VSFO also runs the
GSC application.
For a standalone control plane deployment, the Sx routing instances need to use
externally routable addresses.
EBM IP Address
Gom IP Address
(Ga/Gx/Gy/Rf(PGW)/S6b)
S5/S8-C IP Address
VSFO (CP)
PSC SSC S4-C/S11 IP Address
Table 5 summarizes the applications that can run on a standalone control plane
node.
Rf network IP address
For more information on control plane networks, see Routing. For information on
configuring IP addresses and address ranges, see the following documents:
— Event-Based Monitoring for the PGW-C and Event-Based Monitoring for the
SGW-C for the EBM network.
— Configure Diameter for the PGW-C and Configure Diameter for the SGW-C
for the Rf network.
The user plane can be connected to one or several control plane nodes.
The minimum configuration for a standalone user plane node is two VRPs and
two user plane VSFOs running the user plane application.
For a standalone user plane deployment, the Sx routing instances need to use
externally routable addresses.
S1-U/S4-U/S12/N3 IP Address
Gn/Gp/S5/S8/S2a-U/S2b
IP Address
VSFO (UP)
Sxa/Sxb/N4 IP Address
Payload
Signaling
Table 6 summarizes the network IP addresses and ranges per network function
that the user plane can be used as.
N4 network IP address
N9 network IP address
(1) The Gn/Gp-U and S5/S8-U interfaces belong to the same network, but are configured
separately. For more information, see Configure Interfaces for the User Plane.
For more information on user plane networks, refer to Routing. For information
on configuring IP addresses and address ranges, see the following documents:
— Configure Interfaces for the User Plane for the GTP-based networks and the
SGi network.
— Event-Based Monitoring for the User Plane for the EBM network.
5.4.3 Single Node with Collocated Control Plane and User Plane
The EPG can be run with the user plane node functionality and the control plane
node functionality deployed on the same COTS servers. The control plane node
and the user plane node are logically two separate nodes communicating
through Sx, but they share the same O&M admin login address.
From the O&M perspective, there is one active RP and one stand-by RP in a
single node. From functionality perspective, the control plane node or the user
plane node does not know that it is in the same COTS HW server as the other
node.
— Two control plane VSFOs: the PGW-C runs the PSC application and the
SGW-C runs the SGW Session Controller application. If the node acts as a
PGW-C, one control plane VSFO also runs the GSC application.
— Two user plane VSFOs running the user plane application. The user plane
application can be used as an SGW-U, a PGW-U or as a combined SGW-U
and PGW-U.
— Two VRPs.
For the single node deployment, the Sx routing instances can use private
addressing. Example 1 shows the basic configuration for a single node
deployment with internal routing context. For information on creating a new
context, see Contexts and Interfaces.
co-located
!
For expansion of the single node with external user planes, the single node can
be deployed by putting the Sx interfaces on external routing contexts. The
procedure is the same for both external and internal configurations except for the
IP addresses that are different for external routing. The IP addresses can vary
based on the configuration and network planning. For more information on the
IP addresses for external routing, see Configure Sxa Interface for the SGW-C,
Configure Sxb Interface for the PGW-C, and Configure Interfaces for the User
Plane.
Figure 10 shows an example of an EPG running as a single node with all its
applications and network IP addresses and ranges.
Single Node
SG
SGi IP Address Range VSFO (CP)
S5
PSC SSC
SGi-LAN service chaining
network IP Address Range (Ga/Gx/G
S5/S8-U IP Address S
VSFO (UP)
VSFO (CP)
S1/S4/S12-U IP Address S4/
PSC SSC
Gn/Gp/S5/S8-U/S2a/S2b
IP Address or IP Address Range GSC
Payload
Signaling
In a single node deployment, the control plane can be connected to the user
plane in the same node as well as external user planes (by deploying Sx
interfaces on the external routing context). The user plane can only be connected
to the control plane in the same node.
For more information about the control plane and user plane networks, see
Routing.
For information on configuring IP addresses and address ranges for the control
plane, see the following documents:
— Event-Based Monitoring for the PGW-C and Event-Based Monitoring for the
SGW-C for the EBM network.
— Configure Diameter for the PGW-C and Configure Diameter for the SGW-C
for the Rf network.
For information on configuring IP addresses and address ranges for the user
plane, see the following documents:
— Configure Interfaces for the User Plane for the GTP-based networks and the
SGi network.
— Event-Based Monitoring for the User Plane for the EBM network.