Hardware Reference
In-Depth Information
70
60
50
40
30
20
10
0
No error
One error
in mixing
One error
in splitting
One error in mixing
and on in splitting
With CCD camera-based
sensing system
Without error recovery
mechanism
Fig. 2.17
Completion time for biochip with CCD camera-based sensing system, and the biochip
without error-recovery mechanism when errors are injected in the sample preparation of plas-
mid DNA
outcome of the entire experiment will be incorrect if an error occurs during the
bioassay. As a result, the biochip has to be discarded, and the experiment must
be repeated on a new biochip in order to correct the error. If we assume that re-
execution of the experiment will be successful, the bioassay completion time will
be twice as the completion time in the fault-free case. Based on these results, we note
that error-recovery can reduce the bioassay completion time, and the consumption
of biochemical reagents/samples can be reduced.
In reliability-driven error-recovery, the electrodes where an error is deemed to
occur, will not be used in other operations. On the contrary, for the reliability-
oblivious error-recovery process in [
17
], when an error occurs during execution,
the region where error occurs will continue to be used in subsequent operations.
As discussed in Sect.
2.3.2
, these electrodes with defects may further lead to more
errors.
To compare the completion times derived from reliability-oblivious and
reliability-driven error-recovery procedures, the following simulation is set up.
In the reliability-oblivious error-recovery, we randomly select one operation
opt
fe
as the first instance of error in the execution of bioassay. The electrodes that are
used to perform
opt
fe
are referred to “electrodes with defects”. When another
operation is implemented again on these electrodes with defects, we assume that
there exists a probability P
fail
that this operation will also fail. For a fixed value
of P
fail
, we simulate reliability-oblivious error-recovery 15 times, and determine
average completion time.
Figure
2.18
compares the completion time of reliability-driven error-recovery
and average completion time of reliability-oblivious error-recovery for different
values of P
fail
. Here the randomly selected
opt
ef
is a mixing operation implemented
on a 4
1 electrode array. As expected, Fig.
2.18
shows that the reliability-driven
error-recovery leads to shorter assay completion time in the presence of defects.
