Image Processing Reference
In-Depth Information
congestion. The probability of complement can be seen as the reliability of the network path. The
reliability defines how reliable packet transmission should be on the network. A requirement can
thus be that
P
(
packet received
)≥
W
min
(.)
with
W
min
as the minimum probability for packet reception.
Packet Ordering
Some services rely on order preservation of a sequence of packets. Obviously, the order can change
only if the network topology provides different parallel paths from sender to receiver. In fieldbus sys-
tems,thisistypicallynotanissue;however,itcanberelevantforgeneral-purposeIP-basednetworks
used in an automation context.
A different possibility to assess the performance of a network is to look at it from the application
point of view. Depending on the application's needs, the network must meet certain requirements.
These requirements may of course differ for individual data types. With respect to the packet delay as
the most important parameter, it has become customary to distinguish several “performance classes”
that can be used to define characteristic features for the delay density
f
(
D
)
.
Hard Real-Time
If an application has hard real-time requirements, any given deadline must be met under all circum-
stances. Missing a deadline will inevitably result in a catastrophic failure of the system. Although
the definition of what actually constitutes a catastrophic failure is not always clear, there is at least a
concise requirement for the delay, which must be upper-bounded, i.e.,
f
(
D
)≡
r
D
≥
D
max
(.)
Soft Real-Time
If missing of a deadline does not lead to a catastrophic failure, the application imposes only soft
real-time requirements. his definition is rather loose and can be further split in two subsets:
Guaranteed Delivery
For this requirement, the actual delay of a packet is irrelevant. What counts is that the packet is
definitely received by the addressed node. here are no particular constraints for the delay density, it
may not even be bounded (if it was, we had effectively a hard real-time situation).
Best Efort
This scenario is the weakest one. Here, the network only attempts to deliver the packets as fast as
possible, without giving guarantees concerning deadlines or successful delivery at all. Formally, this
comes down to treating the delay density
f
(
D
)
as conditional distribution,
f
(
D
∣
X
)=
f
(
D
)
p
(
X
)
(.)
where
X
denotes the successful delivery of the packet, which is independent of the delay distribution.
The conditional distribution has the convenient property
∞
f
(
D
∣
X
)
d
D
=
p
(
X
=
true
)<
(.)
∫
−∞
so that the residual probability that a packet is lost in the best effort scenario is implicitly included.
