APPN provides peer networking for SNA. APPN is based on SNA node type 2.1, and implements the following new functions.
- It overcomes the single hop limitation.
- Resources (LUs) can be located dynamically.
- The route that a session will follow across a network is calculated dynamically.
- Required system definition is significantly reduced.
- Flow control and congestion management have been improved.
An APPN network is composed of three types of APPN node.
- Low Entry Networking (LEN) Node
- End Node
- Network Node
A simple APPN network containing all three types of nodes is illustrated in the diagram below.
A Simple APPN Network
LEN Node
An APPN LEN node is equivalent to a pre-APPN type 2.1 node. It provides peer to peer connectivity with all other APPN nodes. A LEN node:
- perceives the connected APPN network as having all the network resources located on its adjacent network node
- does not participate in the APPN topology or directory services
- must be statically configured for its target resources.
End Node
An End Node is similar to a LEN node in that it participates at the periphery of an APPN network. An End Node includes a Control Point (CP) which allows it to exchange network control information with an adjacent network node (its network node server).
An End Node:
- may register its local resources with its network node server
- may participate in the dynamic location of resources
- does not require static configuration
- does not participate in APPN network topology interchanges.
Network Node
The backbone of an APPN network is composed of one or more Network Nodes which provide network services to attached LEN and End Nodes.
A Network Node:
- maintains network topology and local directory databases (including a cache of recently located remote LUs)
- is responsible for locating remote LUs and generating appropriate session routes across the network according to the required class of service
- provides intermediate session routing for sessions that span multiple hops in the network.
Connectivity
The first phase of operation in an APPN network is to establish a physical link between two nodes. When it has been established, the capabilities of the two attached nodes are exchanged using XIDs. At this point, the newly attached node is integrated into the network, which means:
- Network Nodes may establish CP-CP sessions to one or more adjacent Network Nodes
- End Nodes may
- establish CP-CP sessions to their server Network Node
- register their resources at their Network Node server.
Network Nodes exchange topology information by propagating topology updates that reflect the newly attached node through the network. The newly attached node is now ready to participate in the steady-state network operation.
Location of a Target LU
Information about the resources (currently only LUs) within the network is maintained in a database which is distributed across the End and Network Nodes in the network. End Nodes hold a directory of their local LUs.
Network Nodes hold a directory of:
- their local LUs
- LUs in attached End Nodes that have registered their resources
- a cache of recently located remote LUs.
When an originating LU wants to establish a session to a remote LU, it must first locate the remote LU across the network. First, the originating node searches its local directory to determine whether this LU is already known. It would be known if it was:
- a local LU
- a registered LU in an attached End Node
- a remote LU which has been cached from a previous locate.
If the remote LU is found in the directory, a directed search message is sent across the network to the remote machine to ensure that the LU has not moved since it was last used or registered. If the local search is unsuccessful, a broadcast search is initiated across the network. When the node containing the remote LU receives a directed or broadcast search message, it sends back a positive response. A negative response is sent back if a directed or broadcast search fails to find the remote LU.
Route Selection
When a remote LU has been located, the originating Network Node server calculates the best route across the network for a session between the two LUs. Every Network Node in the APPN network backbone maintains a replicated topology database. This is used to calculate the best route for a particular session, based on the required class of service for that session. The class of service specifies acceptable values for session parameters such as propagation delay, throughput, cost and security. The route chosen by the originating Network Node server is encoded in a route selection control vector (RSCV).
Session Initiation
A BIND is used to establish the session. The RSCV describing the session route is appended to the BIND. The BIND traverses the network following this route. Each intermediate node puts a session connector for that session in place, which links the incoming and outgoing paths for data on the session.
Data Transfer
Session data follows the path of the session connectors set up by the initial BIND. Adaptive pacing is used between each node on the route. The session connectors on each intermediate node are also responsible for segmentation and segment assembly when the incoming and outgoing links support different segment sizes.
Dependent LU Requester
Dependent LUs require a host based System Services Control Point (SSCP) for LU-LU session initiation and management. This means that dependent LUs must be directly attached to a host via a single data link. The Dependent LU Requester (DLUR) and Dependent LU Server (DLUS) extension to the APPN architecture:
- provide an APPN network with the Mainframe connectivity to support legacy applications and networking systems
- removes any restrictions on the location of dependent LUs within an APPN network. APPN dependent LU support is provided through the paired functionality of Dependent LU Requester (DLUR) and Dependent LU Server (DLUS).
DLUS resides on the host and DLUR resides on nodes within the APPN network. DLUS/R establish a connecting pipe (a pair of LU 6.2 sessions) across the APPN network which encapsulates SSCP-PU and SSCP-LU sessions. Subsequent BINDs and LU-LU sessions are not encapsulated, but routed across the APPN network in the normal way.
High Performance Routing
High Performance Routing (HPR) is an extension to the APPN architecture which:
- improves the performance of APPN routing by taking advantage of high-speed, reliable links
- improves data throughput by using a new rate-based congestion control mechanism
- supports nondisruptive re-routing of sessions around failed links or nodes
- reduces the storage and buffering required in intermediate nodes.
APPN Architecture Documentation
The IBM Redbook Inside APPN - The Essential Guide to the Next-Generation SNA (SG24-3669-03) is a good introduction to the APPN architecture. This book was previously called IBM APPN Architecture and Product Implementations Tutorial.
A model for the internal architecture for APPN End and Network Nodes is described in IBM SNA APPN Architecture Reference (SC30-3422-04).