Open Tsunami Alerting System
                             OTAS
                       Charles R Martin
INTRODUCTION
=================================================================
The recent extraordinary events in the Indian Ocean have awakened
the world to the hazards presented by tsunami.  While there are
extensive tsunami warning systems available for the Pacific
Ocean[1], the tsunami warnings from the International Tsunami
Information Center[2] were not disseminated quickly enough to
prevent massive loss of life, because no warning communications
infrastructure was available in the eleven countries affected by
the tsunami of 2004 December 26.
The Open Tsunami Alerting System (OTAS) is an open-source project
to supplement the existing tsunami warning systems in a
decentralized, non-governmental fashion.  OTAS *does not*
attempt to supplant any existing warning systems, and will be
designed to integrate easily into other tsunami warning systems.
GOALS
=================================================================
This sort of system presents some very interesting requirements.
First of all, the need for reliability and availability of the
network as a whole is high, but the reliability of the
individual components is low compared to the frequency of real
events.   This implies high redundancy.
Second, because of the economic limitations of the countries
where these systems are most vital, the hardware and network
required cannot be very expensive; in fact, ideally the entire
network should be buildable with the sort of small PC machines
that are "cast offs" in first-world countries.
Third, the system should be built to allow many different means
for delivering the actual alerts: for example, email, telephone
alerts, and SMS messages.
Fourth, the implementation should be simple, and ideally should
not depend on complicated numerical models --- but it should be
able to make use of those models if available.
ARCHITECTURAL PRINCIPLES
=================================================================
Basic guiding principles for OTAS are:
  1.  OTAS will be platform independent and open source under
      the GNU Public License (GPL).
  2. OTAS will be built to be _simple_, _conservative_, and
     _correct_. That is to say:
     * _simple_ --- the design will be intentionally limited;
       rather than including complicated geophysical codes, we
       will concentrate on clarity and transparency.
     * _conservative_ --- if OTAS makes an error, we will design
       to fail safe and fail soft: we prefer issuing a false
       alert to failing to issue an alert.
     * _correct_ --- we will apply all available techniques to
       ensure that the released packages are as nearly error
       free as we can make them.
  3. OTAS will be decentralized. Geophysical or seismological
     data will be accepted from multiple sources (a later post
     will discuss some open sources of seismo data), and no
     notification site will be a "master". I currently am
     thinking about a peer-peer component of the architecture to
     provide for multiply redundant data paths.
IMPLEMENTATION APPROACH
=================================================================
We are still early in conceptualizing and designing OTAS.  Our
current thoughts suggest the architecture will be based on a
service-oriented, web-service based, distributed network
architecture composed of two classes of nodes:
(1) prediction nodes, which receive data feeds from
    seismological and geophysical data sources and produce alert
    messages
(2) service nodes that display or propagate the predictions to
    subscribers.
The actual throughput requirements are quite low, and probably
dominated by the necessity of maintaining some network traffic
to assure network integrity, rather than by the load of actual
predictions.  As such, we don't need to worry about highly
efficient implementations (with the possible exception of
numerical prediction codes), but we want to be able to implement
quickly in a platform independent fashion.
For these reasons, we currently expect to implement the basic
network in Python, with possible FORTRAN and Java codes as
required by the situation.
LICENSING
=================================================================
We plan to release OTAS under the the GPL.
=================================================================
Footnotes:
[1]  ??
[2]  http://www.prh.noaa.gov/itic/