My relative
Miguel Octavio is quoted in today’s “Expediente”
in El Universal talking about the capabilities of the fingerprint machines. Here is the detail of his calculations. Combine this with the error of the system calculated by Bruni and the system is simply a big joke and it is claer it does not do what they claim it does:
In the
absence of the details of the network, we can make the following simple
calculation: A press release by Cogent indicates they sell a
system that allows for the comparison of up
to 500,000 fingerprints per second, the highest number quoted on the
Internet that we have been able to find. But this refers only to the time one
server takes to make such a comparison. The CNE system would require, in order
to preserve the one voter, one vote principle to compares it to all previous
voters and add the new fingerprint to the database.
Let’s do a
simple calculation. We assume that on December 3d. 10 million people will vote
in more or less uniform fashion during ten hours. That is one million per hour
or 277 voters per second. That means, that the system has to evaluate those 277
fingerprints almost in instantaneous fashion, since any delay would make that the
next 277 fingerprints, would have to wait in the queue if the response is not
within one second. Let’s assume that the polls open at 8 AM and let’s imagine
that it is already noon that is 4 million people have voted.
Each
fingerprint needs to be compared to that of the 4 million voters that have already
cast their vote. That means that if there was only one server, it would take 8
seconds per fingerprint. Obviously this is unacceptable. A way to do it in one
second is to have half a million fingerprints per server and have eight
servers. The problem is that at noon we would need 277×8=2216 servers to be
able to complete the task in one second.
Nevertheless,
this simple calculation has a problem, the system can not be designed to handle
averages, but peaks, that is, moments in which you may have twice that number
of voters (There are 16 million registered too), since Venezuelans tend to go
and vote in the morning more than in the afternoon. That means we could be
talking about 4432 servers to satisfy the needs of the system.
But what
happens in the afternoon, let’s say at 4 PM, when we would have twice as many
votes and thus fingerprints and we would need once again twice the number of
servers , that is, 8864 to be able to compare the fingerprints and allow the
voter to vote or not?
The problem
is that this simple calculation does not take into account the design of the
network. Each fingerprint has to be sent to the CNE by satellite. There, there
has to be a server that acts as a traffic manager, deciding which server or
servers will handle the fingerprint that just arrived, which fingerprint is
processes first, which one is ready, which votes already voted and add the new fingerprint
to the appropriate server. You need to have some redundancy too, that is, there
has to be intelligent communication to be able to detect if a fingerprint
arrived or not, whether it had a conflict with another and the like.
All of
this adds time to the servers and as far as we know, the CNE does not have a
huge number of them. It would not help either to have a huge and fast server
since what takes the longest time in the processing is the transfer of the
fingerprints from disk to memory and it would be difficult to have all of them
in memory. On top of that this would not allow the simultaneous processing of
more than one fingerprint.
In the
absence of more information from the CNE or Cogent there isn’t much more that
can be said. It would seem that Venezuela has the fastest AFIS system in the
world if you compare it to the leading crime institutions of the world, or
could it be that it does not do what it is supposed to do, that is guarantee
the principle of one voter, one vote as the CNE Directors say all the time, but
it objective is really something else?