Image Processing Reference
In-Depth Information
Server
Client
Skeleton
Stub
4
8
Dispatcher
100 Hz
2 KHz
HI
HI
HI
HI
7
nORB
Logical clock
1
3
5
nORB
LO
2
Logical clock
LO
6
ME
Reactor
port:10000
RequestId = 23
Logical priority = LO
…….
IOR
ObjectKeyA
…….
hostA:10000
FIGURE .
Processing with a logical clock.
priority levels. his is different from the previous approach discussed in Section .. and
isappliedonlywhenusingtheclocksimulation.helanethreadsaregiventhehighest
actual OS priority and the clock thread itself is assigned the lowest actual OS priority. his
configuration of actual OS thread priorities reduces simulation times while still ensuring
synchronization between nodes.
. As on the client side, the clock thread then chooses the most eligible item from its queues
and enqueues the item on the appropriate lane thread's queue.
. The lane thread dispatches the enqueued item.
2.4 Design Recommendations and Trade-Offs
In this section, we present recommendations and trade-offs that we encountered in designing and
developing the nORB special-purpose middleware to meet the challenges described in Section ...
While we focus specifically on our work on nORB, the same guidelines can be applied to produce
middleware tailored to other networked embedded systems.
2.4.1 Use the Application to Guide Data-Types to Be Supported
We used the application domain to guide our choice of data types that the ORB middleware sup-
ports. In the damage detection application, e.g., sequences of simple structures are exchanged
