[cs615asa] HWN

[Peter F]

Hello,

Here is my write up for HW #N:

I went to the “Using In-Memory Data-Parallel Computing for Operational
Intelligence” event hosted by the SQL NYC group on meetup.com. This
event was a presentation on in-memory data-parallel computations. The
speaker was the CEO of a company that does work for hundreds of
clients relating to this topic. The main point that the speaker was
trying to make is that computations that run on data that is kept in
memory rather than in a database is much faster that computations that
run on data that is kept in a database. This is because data that is
already in memory does not require disk access which is one of the
biggest bottlenecks in modern computing. This enables the processing
of terabytes of data in seconds and allows the use of big data in real
time computations. One of the examples was using real time updating
data from a stock market and using it to produce events that alert the
trader when rules he/she has created are triggered. Another topic of
discussion was map/reduce which is also used in this computation. They
had to adapt their system to map/reduce which generally uses smaller
objects whereas their system is generally intended for large objects
such as an object representing a patient in a hospital. Using objects
allows their system to easily adapt to use of object oriented
languages such as C++, Java, and C#. I mainly chose this event because
I am interested in distributed computing and databases. What I learned
in this event are that moving data between machines is slow and
destroys scalability. Having data in memory rather than in a database
reduces latency greatly and enables real time systems such as for a
cable box server which serves recommendations to a user based on their
viewing history in real time.

Overall the presentation was great and fairly easy to understand even
for someone with little technical expertise on the subject.

Thanks,
Peter Fatyga