Summary: Library for reading Mark6 scatter-gather recordings
Name: mark6sg
Version: @version@
Release: 1
License: GPL3
Group: VLBI
BuildRoot: /var/tmp/mark6sg-@version@-buildroot
Requires: gcc >= 4.1.0
Prefix: /usr
Source: mark6sg-@version@.tar.gz

%description
A library intended to help reconstruct VLBI scans that were written in
a special scatter-gather format onto a JBOD array of disks.

The library provides a VFS-like read-only file interface (standard 
open(), read(), and so on calls) that allow programs to read these 
scattered files as if they were a single file.

The library assumes that files were written on a 10GbE network data capture
computer with the Mark6 software (dplane and cplane; using versions at least 
until 11/2014). A single 10GbE capture session in Mark6 software produces a
series of scattered files, for example:

/mnt/disks/1/0/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/1/1/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/1/2/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/1/3/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/1/4/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/1/5/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/1/6/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/1/7/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/0/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/1/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/2/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/3/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/4/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/5/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/6/wrtest_kjcc_test_2014y314d10h59m09s.vdif
/mnt/disks/2/7/wrtest_kjcc_test_2014y314d10h59m09s.vdif

The Mark6 recordings are identified by their base filename. In the above 
example the recording is "wrtest_kjcc_test_2014y314d10h59m09s.vdif".

The several files associated with a recording each contain hundreds
of typically ~10MB-sized data blocks. During recording, the incoming 10GbE 
UDP/IP data are combined into blocks by Mark6 software. To prevent 
slower-throughput disks from forming a performance bottleneck, Mark6
software scatters blocks to files in the order as earlier writes
are completing (i.e. not Round-robin scheduling). Hence the time order 
of data across files is somewhat random.

Upon open()'ing a recording the library thus needs to check the metadata
of all files associated with the recording. This can take a few seconds.

After this read() and lseek() can be used to read data from a recording.
There are no special restrictions.

Interpreting the data is up to the user application.
Typically it is VDIF format VLBI data.

%prep
%setup -q -n mark6sg-@version@

%build
%configure
make

%install
%makeinstall

%files
%defattr(-,root,root)
%{_libdir}/libmark6sg.la
%{_libdir}/libmark6sg.so.0.0.0
%{_libdir}/libmark6sg.so.0
%{_libdir}/libmark6sg.so
%{_libdir}/libmark6sg.a
%{_includedir}/mark6sg.h
%{_includedir}/mark6sg/mark6_sg_utils.h
%{_includedir}/mark6sg/mark6_sg_vfs.h

%changelog
20 Nov 2014 Jan Wagner <jwagner@kasi.re.kr>
28 Jul 2014 Jan Wagner <jwagner@kasi.re.kr> added new function to API
09 May 2015 Jan Wagner <jwagner@kasi.re.kr> simpler prefetch code, multiple files openable simultaneously