Hardware Reference
In-Depth Information
In all the three figures, we can see that the points that guarantee minimum average
latency and minimum memory (translated in number of cells used in packet and FEC
memories), while maintaining the pre-set UDP throughput, are reached at around
1,064 cells for PKT and FEC memories. However, a more in depth study with
modeFRONTIER analysis tools is needed for deriving the exact figures.
7.2.3.3
Objective Assessment
From an objective point of view, after the assessment of the procedure proposed
in MULTICUBE project for system optimization, the overall conclusion is that the
proposed flow can save up to a 80% of designer time while achieving better results
in terms of performance since much more simulations can be run and analyzed
following this flow obtaining 10 times more possible combinations than with the
semi-automatic design flow. Also, with just one run we can already identify the
optimum for the design parameters. A second or third iteration would fine tune the
design in order to obtain the final parameters.
As an example, we have run a test case where we have tried to optimize the
parameters described in previous sections and we have measured the values presented
Fig. 7.17 (an estimation based on designer's inputs has been made).
As we can clearly see from Fig. 7.17 , the automated procedure provides:
￿
Higher number of points: The number of points that can be analyzed by follow-
ing MULTICUBE methodology is much higher than the ones obtained by using
the semi-automated procedure
￿
Better quality of the results: The quality of the results obtained through the use of
the automated procedure is higher since modeFRONTIER guarantees the correct
Fig. 7.17 Assessment procedures comparison (semi-automated vs automated)
 
Search WWH ::




Custom Search