advanced/upf.rst

P4-based User Plane Function (P4-UPF)
=====================================

Overview
--------

SD-Fabric supports running a 4G/5G mobile core User Plane Function (UPF) as part
of the switches packet processing pipeline. Like the rest of the pipeline, this
is realized using P4 and for this reason we call this P4-UPF.

P4-UPF is integrated with the ONF's SD-Core project. By default, SD-Core ships
with BESS-UPF, a containerized UPF implementation, based on the Berkeley
Software Switch (BESS).

SD-Fabric can be used with BESS-UPF or any other UPF implementation that runs on
servers. In this case, the fabric switches can provide routing of GTP-U packets
to and from radio base station and servers. When P4-UPF is enabled, the same
fabric switches perform GTP-U tunnel termination.

.. image:: ../images/bess-p4-upf.png
    :width: 700px

**Supported Features**

SD-Fabric's P4-UPF implements a core set of features capable of supporting
requirements for a broad range of enterprise use cases:

* GTP-U tunnel encap/decap: including support for 5G extensions such as PDU
  Session Container carrying QoS Flow Information.
* Accounting: we use switch counters to collect per-flow stats and support usage
  reporting and volume-based triggers.
* Downlink buffering: when a user device radio goes idle (power-save mode) or
  during a handover, switches are updated to forward all downlink traffic for
  the specific device (UE) to DBUF, a K8s-managed buffering service running on
  servers. Then, when the device radio becomes ready to receive traffic,
  packets are drained from the software buffers back to the switch to be
  delivered to base stations.
* QoS: support for enforcement of maximum bitrate (MBR), minimum guaranteed
  bitrate (GBR, via admission control), and prioritization using switch
  queues and scheduling policy.
* Slicing: multiple logical UPFs can be instantiated on the same switch, each
  one with its own QoS model and isolation guarantees enforced at the hardware
  level using separate queues.

**Distributed UPF**

.. image:: ../images/upf-distributed.png
    :width: 700px

In SD-Fabric we support different topologies to meet the requirements of
different deployment sizes: from a single rack with just one leaf
switch, or a paired-leaves for redundancy, to N x M leaf-spine fabric for multi-rack
deployments. For this reason, P4-UPF is realized with a "distributed" data plane
implementation where all leaf switches are programmed with the same UPF
rules, such that any leaf can terminate any GTP-U tunnel. This provides several
benefits:

* Simplified deployment: base stations can be connected via any leaf switch.
* Minimum latency: the UPF function is applied as soon as packets enter the
  fabric, without going through additional devices before reaching their final
  destination.
* Fast failover: when using paired-leaves, if one switch fails, the other can
  immediately take over as it is already programmed with the same UPF state.
* Fabric-wide slicing & QoS guarantees: packets are classified as soon as they
  hit the first leaf. We then use a custom DSCP-based marking to enforce the
  same QoS rules on all hops. In case of congestion, flows deemed high priority
  are treated as such by all switches.

**Control Architecture and Integration with SD-Core**

SD-Fabric's P4-UPF is integrated with the ONF SD-Core project to provide a
high-performance 3GPP-compliant mobile core solution.

The integration with SD-Core is achieved via an ONOS application called UP4,
which is in charge of populating the UPF tables of the switch pipeline.

.. image:: ../images/up4-arch.png
    :width: 600px

The interface between the mobile core control plane and the UPF is defined by
the 3GPP standard Packet Forwarding Control Protocol (PFCP). This is a complex
protocol that can be difficult to understand, even though at its essence the
rules that it installs are simple match-action rules. The implementation of such
protocol, such as message parsing, state machines, and other bookkeeping can be
common to many different UPF realizations. For this reason, SD-Fabric relies on
an implementation of the PFCP protocol realized as an external microservice
named “PFCP Agent”, which is provided by the SD-Core project.

The UP4 App abstracts the whole fabric as one virtual big switch with UPF
capabilities, we call this the One-Big-UPF abstraction. Such abstraction allows
the upper layers to be independent of the underlying physical topology.
Communication between the PFCP Agent and the UP4 App is done via P4Runtime. This
is the same API that ONOS uses to communicate with the actual switches. However,
in the former case, it is used between two control planes, the mobile core, and
the SDN controller. By doing this, the deployment can be scaled up and down,
adding or removing racks and switches, without changing the mobile core control
plane, which instead is provided with the illusion of controlling just one
switch.

The One-Big-UPF abstraction abstraction is realized with a ``virtual-upf.p4``
program that formalizes the forwarding model described by PFCP as a series of
match-action tables. This program doesn't run on switches, but it's used as the
schema to define the content of the P4Runtime messages between PFCP Agent and
the UP4 App. On switches, we use a different program, fabric.p4, which
implements tables similar to the virtual UPF but optimized to satisfy the
resource constraints of Tofino, as well as tables for basic bridging, IP
routing, ECMP, and more. The UP4 App implements a P4Runtime server, like if it
were a switch, but instead it internally takes care of translating P4Runtime
rules from ``virtual-upf.p4`` to rules for the multiple physical switches running
fabric.p4, based on an up-to-date global view of the topology.

Downlink Buffering (DBUF)
-------------------------

A UPF is required to buffer packets when UEs are in idle-mode or during
handovers, this is usually called *downlink buffering*, as it applies only to
the downlink direction of traffic. Most switches provide buffering capabilities
to handle congestion, they cannot hold packets indefinitely. For this reason, we
provide DBUF, a microservice
responsible for providing the downlink buffering capabilities to P4-UPF.

.. image:: ../images/dbuf.png
    :width: 400px

When a UE goes idle and turns off its radio, or during handovers, the mobile
core control plane uses PFCP to update the Forwarding Action Rules (FARs) for
that UE to enter buffering* mode. When this happens, UP4 updates the switch rules to
steer packets to DBUF using GTP-U tunnels.

UP4 uses gRPC to communicate with DBUF. DBUF notifies UP4 about buffering
events, which are relayed to the mobile core control plane as Downlink Data
Notifications (DDN). When a UE becomes available again, UP4 triggers a buffer
drain on DBUF and updates the switch rules to start sending traffic to the UE again.

Deploying DBUF is optional (can be enabled in the SD-Fabric Helm Chart).
DBUF feature requires SR-IOV and DHCP support on NICs and Kubernetes CNIs.

ONOS Configuration
------------------

The UPF configuration is split in two configurations, that can be provided
independently to ONOS. Th first is used to configure the UP4 ONOS application
and defines UPF-related information such as S1U Address, network devices
implementing UPF etc. The second one, instead, is used to configure parameters
related to the DBUF functionality.

Here's a list of fields that you can configure via the UPF Network Configuration
for UP4:

* ``devices``: A list of devices IDs that implements the UPF data plane. This
  list must include all the leaf switches in the topology. The UPF state is
  replicated on all devices specified in this configuration field. The devices
  specified in this list must use a P4 pipeline implementing the UPF
  functionality. *Required*

* ``s1uAddr``: The IP address of the S1-U interface (equivalent to N3 for 5G).
  It can be an arbitrary IP address. *Required*

* ``uePools``: A list of subnets that are in use by the UEs. *Required*

* ``dbufDrainAddr``: The IP address of the UPF data plane interface that the
  DBUF service will drain packets towards. *Optional*

* ``pscEncapEnabled``: Set whether the UPF should use GTP-U extension PDU
  Session Container when doing encapsulation of downlink packets. *Optional*

* ``defaultQfi``: The default QoS Flow Identifier to use when the PDU Session
  Container encapsulation is enabled. *Optional*

Here is an example of netcfg JSON for UP4:

.. code-block:: json

    {
        "apps": {
            "org.omecproject.up4": {
                "up4": {
                  "devices": [
                    "device:leaf1",
                    "device:leaf2"
                  ],
                  "s1uAddr": "10.32.11.126",
                  "uePools": [
                    "10.240.0.0/16"
                  ],
                "dbufDrainAddr": "10.32.11.126",
                "pscEncapEnabled": false,
                "defaultQfi": 0
                }
            }
        }
    }

The DBUF configuration block is all *optional*, we can use UP4 without the
downlink buffering functionality. Here's a list of fields that you can
configure:

* ``serviceAddr``: The address of the DBUF service management interface in the
  form IP:port. This address is used to communicate with the DBUF service via
  gRPC (for example, to trigger the drain operation, or receive notification for
  buffered packets).

* ``dataplaneAddr``: The address of the DBUF service data plane interface in the
  form IP:port. Packets sent to this address by the UPF switches will be
  buffered by DBUF. The IP address must be a routable fabric address.

Here is an example of netcfg for DBUF:

.. code-block:: json

    {
        "apps": {
            "org.omecproject.up4": {
                "dbuf": {
                    "serviceAddr": "10.76.28.72:10000",
                    "dataplaneAddr": "10.32.11.3:2152"
                }
          }
        }
    }

.. note::
    When deploying DBUF using the SD-Fabric Helm Chart you do **NOT** need to
    provide the ``"dbuf"`` part of the UP4 config. That will be pushed
    automatically by the DBUF Kubernetes pod.

PFCP Agent Configuration
------------------------

PFCP Agent can be deployed as part of the SD-Fabric Helm Chart.

See the Helm Chart documentation for more information on the configuration
parameters. Once deployed, use ``kubectl get services -n sdfabric`` to find out
the exact UDP endpoint used to listen for PFCP connection requests.

UP4 Troubleshooting
-------------------

See :ref:`troubleshooting_guide`.