Linux based LISS systems. The ISEV and SISLook examples
A. Llobet1,2, J. Vila1,2, O.Fors1,4 and R.Macià2,3
1Departament d'Astronomia i Meteorologia, Universitat de Barcelona
2Laboratori d'Estudis Geofísics "Eduard Fontserè",
Institut d'Estudis Catalans
3Departament de Matemàtica Aplicada II, Universitat Politècnica de Catalunya
4Observatori Fabra, Reial Acadèmia de Ciències i Arts de Barcelona
Introduction -
General scheme -
Actual implementation. The ISEV system
Real Time Analysis -
Further developments. The SISLook system -
Acknowledgements
Introduction
Recent advances in technology have provided seismic stations with many useful new capabilities, one of the most important being the continuous real-time (RT) data transmission. However, many instruments developed a few years ago were designed without these advantages, although they work excellent as data acquisition system. Many seismic acquisition systems designed with technology from the beginning of the nineties are PC-based systems. For these old systems advances in technology and operating systems can be used to provide them with some of the useful capabilities of modern systems.In this contribution we present a general scheme for the development of Live Internet Seismic Server (LISS) class servers based on Linux and TCP/IP connections using Secure File Transfer Protocol (SFTP). The objective is to develop a cheap and easy way to implement a very stable system with minimum hardware requirements. At the same time we require good performance and robust data loss prevention. Moreover, we will show how real-time analysis of the recorded data can be easily implemented in this system. Real-time analysis is a good tool for data quality control, for quick detection of station malfunctioning or for specific experimental needs.
A system of this kind has been implemented as a Live Internet Seismic Server (LISS) at the Observatori Fabra of the Reial Acadèmia de Ciènces i Arts de Barcelona in collaboration with the Departament d'Astronomia i Meteorologia of the Universitat de Barcelona. It can be accessed at the d'Iinformació Sísmica En Viu (ISEV) website.
In the first section we will explain how the system works and the choice of the software. After that we will explain the actual implementation of the ISEV system. In the third section the real-time capabilities of the system are explained and its use in the ISEV system is presented. To finish we present further developments and a complementary service SISLook (Seismogram Information Server) that provides an easy way to do a preliminary visual inspection of the seismograms in the datafarm. This service allows you to view selected segments or full days.
General scheme
This LISS system is formed by two subsystems, the STATION and the DATA CENTER, connected through the Internet.
STATION
Figure 1. Data flux diagram of the Station subsystem.
In the STATION we have a PC-based system that periodically extracts the data from the ring buffer and converts it to fully compatible SEED-volumes. In this process the headers are edited, checked and changed when necessary. After the volumes have been generated they are stored into a machine running under Linux and configured as SFTP server. Two copies of each volume are made, one in the local directory where all the data is stored for backup, the second in a temporary directory used for data transmission via SFTP. The temporary directory contains all the data that has not been yet successfully requested by the DATA CENTER.
The DATA CENTER, acting as client, requests periodically all the data currently stored in the temporary directory of the STATION, the server. Once a SEED-volume is successfully received and stored in the DATA CENTER its temporary copy is deleted. If a connection fails the data remain in the temporary directory and will be requested in the following connection. This is a very simple way that prevents any loss of data during the transmission.
DATA CENTER
Figure 2. Data flux diagram of the DATA CENTER and services offered to the public.
When new data is received the DATA CENTER organizes it into de datafarm, updates the 24h plots and does the RT analysis. The data stored in the datafarm, the 24h plots and the results of the RT analysis are made visible to the public by means of a web page with script executing capabilities. Moreover, an AutoDRM system makes the datafarm accessible. Both the processing and serving tasks can be done using computers running Linux.
Up to now we have explained the general working scheme of
the system. Now we want to underline the reasons to choose Linux as the base
system and SFTP as transmission protocol.
Linux is a very stable operating system and has good networking
services. Besides it does not need very powerful machines to offer performance
good enough for the tasks needed in a system like this one. Its cron services
and scriptable shell are very handy for any data management process
repeated periodically. Also important are the tools used to plot the data
into web compatible figures. There are many of them that work under Linux
and are easily scriptable. And last and very important, all this is free
so the cost for software is minimum. Although direct FTP from the PC-based
acquisition system has been tested, there are two main reasons to use SFTP. First
of all, as it uses TCP/IP, you can know the state of the connection, what
is very important for the simple system we use to avoid data loss. Security
is also an issue and SFTP guarantees the privacy of the passwords.
Actual implementation. The ISEV system
The above presented scheme has been implemented by the Observatori Fabra and the Department d'Astronomia i Meteorologia. The data used is obtained from the broadband stations of Fabra (FBR) and Fontmartina (FONT). The LISS can be accessed at the ISEV website.
ISEV System
Figure 3. ISEV system scheme.
The STATION subsystem is located in the Observatori Fabra.
The data in the Ring Buffer is extracted periodically (every 30 min) by the
Nanometrics' program "Archiver". It runs under OS/2 warp 4 in a PI 133MHz
machine. The "Archiver" has been modified so it now generates full SEED-volumes:
they include the response information. After its creation each volume is
sent via FTP to the linux SFTP server; this FTP transmission is done in a
private network. Two copies of each are made, one into the backup directory
and the other into the temporary directory. The linux server runs on a PI
120MHz computer.
The DATA CENTER requests the data every 30 min. The incoming data
is stored by the client machine itself (a PIII) in the datafarm (80Gb HD).
At the same time the plots are generated. We use SAC (Seismic Analysis Code)
for Linux, that generates postscript pictures. Once all this process is finished
the real-time analysis is done in the same machine. It is based on spectral
analysis of the incoming data (next section for details). The figures with
the results are generated, in postscript, using Gnuplot. The Image Magick
package is used for the conversion of postscript figures to web compatible
formats.
Figure 4. ISEV website snapshots.
a)Daily 24h waveform plot. Corresponding to Fabra Station (FBR) on the current day (15 October 2002). b)Zoomed view of the previous 24h waveform plot. c)24h Waveform plot of a selected day from the datafarm. In this case the day when the Sicila earthquake occurred (6 August 2002). d)Temporal segment from the datafarm corresponding to a local event occurred the 5th October 2002. It shows the SHZ channel from FBR and FONT stations. Click figures to see larger images.
Both the web server and the AutoDRM run on a Sun Server
under SunOS 5.6. This machine only administers the requests for the ISEV
and runs the cgi-Perl scripts. All the pages, plots and data shown in the
webpage or accessible via AutoDRM are stored in the Linux machine. This data
is mounted in the Sun Server using NFS.
The ISEV website provides the following information: zoomable
24h waveform plots for the three last days, 24h non-zoomable waveform plots
for all the days in the datafarm, waveforms for selected temporal segments
in SHZ channel (Figure 4) and the results of the RT analysis (see next section).
Real Time Analysis
The computer in the DATA CENTER that manages the reception and storage of the data stays most of its time idle. All this free time could be used for other tasks as preliminary analysis of the incoming data. This analysis can carry out many different tasks as:- Quality control of data
- SOH of the station operation
- General characteristics of the seismic signal
- Any other specific need
The implementation of new tasks to be applied to incoming data after each request is only limited by the CPU capabilities of the receiving system. These tasks are executed sequentially before the incoming data are removed from the temporary incoming directories at the end of the process.
Figure 5. Snapshots of the RT analysis results.
a) Minimum spectra for FONT station. b) Time series for the frequency of the maximum, FBR station. Sudden changes of this value not related with earthquakes can show changes in the station behavior. Click figures to see larger images.
In the ISEV website the minimum spectra, and the time series corresponding to the amplitude of the maximum and its frequency are shown (Figure 5). These plots, mainly the one showing the evolution of the frequency of the maximum, allow the distant monitoring of the behavior of the sensors and recording system.
Further developments. The SISLook system
Currently the datafarm at the Department of Astronomy contains also the data from stations in Tunel del Cadí (CADI), Monestir de Poblet (POBL) and Power Plant Vandellòs 2 (VAN2). This data is also made visible on the SISLook website. At this moment this data is not transfered in real-time to the Data Center. However, a system like the one presented here would be easily implemented when these stations get access to the Internet and this is our next main goal. The SISLook provides the users the possibility to do a preliminary visual inspection of the data before doing any request. This service has been proved to be very useful when waveform data with certain characteristics is required. In particular, for those analysis that are very sensitive to S/N ratio. This preliminary work may reduce waste of time in useless requests and analysis.The SISLook web page also links to corresponding stationbooks where full information of the operation of each station can be found. Included history, response files, noise studies, log books and any other incidences that may have affected the normal operation of each station.