Speaker
Prof.
Soh Suzuki
(KEK)
Description
Formerly most of HEP experiments in japan used
the centralized computing model.
Originally HEPnet-J had only one instance which is
connected to Internet, and recently it has many closed
network which connects domestic sites.
At that time, the network connectivity in Japan
was very poor and the main purpose of HEPnet-J was
providing enough connectivity for interactive use
over domestic and international links funded by KEK.
During last 10 years,
the domestic and international connectivity
provided by each university and NREN has been
dramatically improved and it is enough for
manual transfer of typical mDST files.
The rapid growth of data volume
makes it unable to apply same model to new generation experiments.
As the LCG tier structure for LHC has proved that
the distributed computing model over collaboration sites
is really applicable to the huge scale experiment,
the external connectivity for international collaboration sites
should be more faster, and more secure.
For example,
the Belle II experiment in KEK
will have huge data repositories in US and EU.
The expected throughput from KEK to US is 20Gbps,
thus we have to bypass the security device like a firewall
for this data path.
The computing facility in KEK (KEKCC) is not enclosed
in the firewall of KEK, but it has a dedicated firewall that
is managed by the independent policy.
Its throughput is sufficient to handle an ordinary activity
over Grid VO.
But it is not enough for the mass-data transmission for Belle II,
we have to bypass it in the same way of LHCONE.
Now bypass lines for Belle II are prepared and under tasting.
We will report how we renovate HEPnet-J for experiments
driven by international collaboration.
Author
Prof.
Soh Suzuki
(KEK)
Co-authors
Fukuko YUASA
(KEK)
Malachi Schram
Takanori HARA
(KEK)
Tomoaki Nakamura
(High Energy Accelerator Research Organization (JP))
Yukio KARITA
(KEK)
