2. USAGE CONSIDERATIONS › 2.2 General Analysis Considerations › 2.2.2 SNA Overview › 2.2.2.2 SNA Network Design Fundamentals
2.2.2.2 SNA Network Design Fundamentals
There are three objectives which must be satisfied in any
network design:
o Service
o Availability
o Efficiency
The relative importance of each of these factors determines
(or at least should determine) the design of the network.
Service is usually measured by response time. Response time
may be loosely defined as the interval between an end-user
dispatching a request and the arrival of the corresponding
response from the session partner. Minor differences in this
definition revolve around the interpretation of when the
response has actually arrived.
Availability is the state of a given network facility being
functional at a given time. Availability is generally
measured as the ratio of the actual functional amount of time
to the committed functional amount of time, expressed as a
percentage. Availability is the most difficult network
attribute to measure because the availability of a given
network element is dependent on the availability of other
network elements in the path to it. Thus, the availability
of a given network element may vary with the point of access
to the network.
Efficiency is the quality of meeting service and availability
objectives for the minimum cost. Because there are so many
differences in the resource requirements, efficiency is
generally measured on an application-by-application basis
and frequently by user group as well.
NETWORK TOPOLOGY
Network topology is the general name given to the physical
layout of network nodes and links most often represented as
"ball-and-stick" diagrams on network managers' office walls.
As complex as these diagrams often appear at first glance,
virtually all networks use one of three basic designs or
topologies:
o Bus
o Ring
o Star
The majority of network traffic is routed to or from
applications that reside in the network's mainframe
processors (host subarea nodes). The communications
processor subarea nodes which are channel-attached to these
processors and the links which interconnect them are subject
to the highest traffic loads and are of the highest concern
in network design. These nodes and links together are
referred to as the network backbone. Defining the structure
of the network backbone in terms of the three basic
architectures clarify much of any network structure being
examined.
Bus Topology
Bus topologies are the simplest form of network design. Bus
structures result when there are only two nodes in the
network backbone. In practice this usually results when an
organization has two operational centers (one or both of
which may have mainframe processors) which are fairly widely
separated geographically. Figure 2-9 is a schematic
representation of a network employing bus topology.
+--------------+
| HOST |
| PROCESSOR |
| B |
| |
+--------------+
||
+-------------------------------------------------||--------+
| || |
| +--------------+ +--------------+ |
| | |--------------- | | |
| | CCU |------------- / | CCU | |
| | NODE A | / ------------| NODE B | |
| | | --------------| | |
| +--------------+ TRANSMISSION GROUP +--------------+ |
| || |
+-------||--------------------------------------------------+
|| BUS BACKBONE
+--------------+
| |
| HOST |
| PROCESSOR |
| A |
| |
+--------------+
Figure 2-9. Bus Topology
Ring Topology
Ring topologies result when each node in the backbone is
connected to two adjacent backbone nodes. This is typically
done to reduce the dependence on the availability of an
intermediate or "through" node by providing an alternate
path. Ring structures are common in organizations with
decentralized operational centers. Figure 2-10 is a
schematic representation of a network employing ring
topology.
| PROCESSOR |
| A |
+--------------------+
|| RING BACKBONE
+--------------------------||-----------------------------+
| || |
| +---------+ |
| | CCU | |
| | | |
| | NODE A | |
| +---------+ |
| / \ |
| TG AB / \ TG AC |
| / \ |
| /\ / \ /\ |
| / \ / \ / \ |
| / \/ \/ \ |
| / \ |
| / \ |
| / \ |
| +---------+ +---------+ |
| | CCU |--------------- | CCU | |
| | | / | | |
| | NODE B | ----------------| NODE C | |
| +---------+ TG BC +---------+ |
| || || |
+------||---------------------------------------||--------+
|| ||
+---------------+ +---------------+
| PROCESSOR | | PROCESSOR |
| B | | C |
Figure 2-10. Ring Topology
Star Topology
Star topologies are used when the operational centers of an
organization are geographically separated but the mainframe
processing capabilities are centralized. The star design
reduces communications line costs by multiplexing multiple
sessions across the transmission group(s) linking the remote
node to the central node. Figure 2-11 is a schematic
representation of a network employing star topology.
STAR BACKBONE
+---------------------------------------------------------+
| |
| +---------+ |
| | CCU | |
| | | |
| | NODE A | |
| +---------+ |
| | |
| | /| |
| +-----------------+ |/ | TG DA |
| | | | |
| | | +---------+ |
| | | | CCU | |
| | PROCESSOR D |==| | |
| | | | NODE D | |
| | | +---------+ |
| | | / \ |
| +-----------------+ / \ |
| TG DB / \ TG DC |
| / \ |
| /\ / \ /\ |
| / \ / \ / \ |
| / \/ \/ \ |
| / \ |
| / \ |
| +---------+ +---------+ |
| | CCU | | CCU | |
| | | | | |
| | NODE B | | NODE C | |
| +---------+ +---------+ |
| |
+---------------------------------------------------------+
Figure 2-11. Star Topology
While the network backbone is not an SNA concept, it is
nonetheless a functional entity. Its utility is twofold: it
makes it easier to visualize the network structure and it
identifies the nodes and links that are most vital to network
performance and availability. Once the backbone structure
has been designed, it must be defined using the SNA routing
techniques, which are briefly outlined below.
NETWORK LINK CONFIGURATION AND DATA FLOW
SNA network links are organized into several different
logical and physical configurations based on the complexity
of the network and the type of sessions supported.
Adjacent communications controllers can be connected by
multiple links operating concurrently. These are called
parallel links. Each parallel link must be assigned to a
logical entity called a transmission group (TG). A
transmission group appears as a single link to the path
control network. Software within the CCU acts to evenly
distribute the load across the links which make up a TG.
To allow data to flow in the network, one or more paths must
be defined between each pair of network addressable units
(NAUs). Paths are defined by specifying an explicit route
(ER) and, if necessary, a peripheral link. The explicit
route is the portion of the path between two subarea nodes.
The peripheral link is the portion of the path between a
subarea node and a peripheral node.
Path definitions are distributed among all the nodes along
the path and stored in routing tables. A routing table entry
exists for each path and consists of the destination node,
explicit route number, next subarea node, and applicable
transmission group number. Thus each node retains only
enough information to route the message to the next node.
Session traffic is assigned to a logical connection called a
virtual route. Virtual routes are then mapped to an explicit
route. Virtual routes can have up to three different
transmission priorities. Messages are queued and selected
for transmission on a TG based on these transmission
priorities.
Virtual routes are assigned to a session when it is
initiated. A session-initiation request contains a class of
service. The SSCP looks up this class of service in its
class of service table, which contains pairs of virtual
routes and transmission priorities, to assign a particular
virtual route to the session.
FLOW CONTROL
Because of the relatively autonomous and asynchronous nature
of session traffic throughout the network, the amount of data
to be sent to a given node at any time may exceed that node's
ability to accept data. Since any network element's ability
to buffer data is limited, some method of preventing overload
is required; this technique is called flow control. SNA flow
control is implemented through pacing. Pacing allows a
receiving node to limit the rate at which traffic is sent to
it. SNA defines two types of pacing:
o Session-level pacing
o Virtual-route pacing
Both session-level and virtual-route pacing are based on a
pacing window whose size is the number of message units that
can be outstanding between origin and destination. When this
limit is reached, the sender stops sending until the receiver
acknowledges readiness by sending a pacing response.
Specifying explicit routes defines the network topology in an
SNA environment. The specification of virtual routes,
transmission priorities, and pacing parameters controls
service levels and throughput for network services.
