The aim of this forum is 1) (short-term) to discuss on the possible solutions for the improvement of connectivity among Asian Tier sites and their status of domestic network environment and 2) (long-term) to monitor periodically the state-of-art of the established network environment through this forum and to organise a body with a broader agenda embracing not only the network but also common issues that could be arisen among Asian Tier sites.
The target of this forum is mostly Asian Tier sites, however, the forum is open to every interested parties, in particular, distributed computing coordinations of LHC experiments and network experts on OPN/ONE.
Registration is closed. Please contact the forum organizer via email if you would like to participate to the forum.
Use the following link to join Vidyo Conference for this forum: vidyo link
Korea Institute of Science and Technology Information, Daejeon Headquarters. Please refer to Travel Information below to know how to reach KISTI.
Banquet will start 19:00 on Wednesday (2nd day of the forum, 23rd September) at The Kingdom Buffet & Wedding (Picaso hall on 3rd floor). It takes 15m by walk from the Dunsan Graytone Hotel and 4m by metro. The location of the banquet place with directions from the hotel can be found here (Warning!! Korean). The walk path is shown here (Warning!! Korean).
How to reach Daejeon from Incheon Airport:
To check bus schedule, go to schedule and click “Bus Routes” then do bus search with the term “Daejeon” and refer to “Daejeon(Deluxe)” column and your final destination is “Daejeon Government Complex”.
The bus stop heading for “Daejeon Government Complex” is located at the outside of the exit number 9D on 1st floor (note that 1st floor means ground floor in Korea, so the same floor with Arrival).
The bus destined for "Daejeon(Sejong City)” will bring you to “Daejeon Government Complex” as well however, please note that it will pass through “Sejong City” so that it will take 20 minutes more than “Deajeon(Deluxe)”.
When you arrive Daejeon, you are recommended to take a taxi and to follow the instruction in this or you can refer to this to get to KISTI, Daejeon Headquarters. Please note that the forum will take place at KISTI, Daejeon Headquarters since there are several branches of KISTI in Korea .
The name of hotel is Dunsan Graytone Hotel and its location is here (situated at the near of City Hall and metro station (City Hall station) allowing easy access to shopping and sightseeing)
Shuttle service (KISTI <-> Hotel) will be provided by the hotel for the participants to this forum. Detailed schedule will be provided later on.
How to reach the hotel:
After Tatsuya’s presentation on the new T2 site in Tsukuba, Japan, William asked about what LHCONE connection from Tsukuba means and Tatsuya said it is not connected yet.
Gungwon asked about who the user of Tsukuba site is and Latchezar(?) said user for Tsukuba T2 is worldwide users like other T1 or T2, it is the resource contribution to the community (e.g. ALICE).
Vikas added India (kolkata) also got the same questions many times and it is convinced that we can get access to the whole experiment resources (such as detectors, whole computing resources) with just few contribution.
After Jianlin’s presentation on Wuhan site report, she added Wuhan is trying to connect to IHEP in Beijing and Latchezar said internal connectivity seems OK in China however going to outside is somehow difficult because of some monitoring policy and at some unknown reason some are blocked, as a result Wuhan is in “semi-“connected state to ALICE community.
(Also, Latchezar pointed at some point later on, due to asymmetric routing accessing to Wuhan site is not possible, that is why we see asymmetric bandwidth test results at alimonitor.cern.ch)
After Taufiq’s presentation on Bandung site report, Latchezar pointed out the site situation is not the problem but the discontinuity on human resources is (new system admin, Aditya).
Edoardo asked about what the two clouds means in slide number 6 and Taufiq answered both are internet connection while the cloud connected to 330Mbps through 45Mbps link from LIPI is the commercial one.
Taufiq confirmed from Asif’s question that 100Mbps is the main link while 45Mbps line is for backup and connected to commercial link.
Sang-Un asked other NREN in Indonesia and Taufiq said there are but those are for other institutions, it seems that NRENs are not shared among institutions rather each institutions has their own backbones.
After Brij’s presentation on Mumbai site report, Chinorat added TEIN goes to HongKong via Singapore.
Brij confirmed Sang-Un’s questions that TIFR connected to the router of SINET(JP) and VECC uses the same route to SINET; Mumbai directly peered with CERN in 10Gbps and has a backup to 2.5Gbps to EU.
After Vikas’s presentation on Kolkata site report, Kolkata is connected to POP in Mumbai with 1Gbps while TIFR is connected with 2Gbps.
Latchezar added that involving with EU (well established connection to EU) does not help because the sites in Asia has to compete with EU sites already having good connectivity among them and he pointed out that the strong point of T2 is to store reconstructed data which is easily forgettable. While CERN has whole raw data and T1 shares a portion of raw data (no single site has whole data especially for reconstructed data), the large portion (about 60-70%) of grid activity is simulation and 20% is user analysis which are mostly relied on T2.
Asif asked Vikas about statistics on data transfer between EU and Asia and Vikas answered there is no statistics because the data transfer is managed centrally by ALICE policy and Latchezar added that there is no priority on jobs.
Brij added that CMS does have because PhEdEx or FTS for CMS have data transfer statistics.
(Someone) pointed out that inefficient connectivity cause slow data transfer and transit limits a lot of connectivity.
William added that Internet2 in US frequently does transit while ESnet in US does as well but restricted only to scientific community.
Edoardo added that the problem of transit will be discussed at LHCOPN/LHCONE workshop in a month in Amsterdam.
After Chinorat’s presentation on Thailand site report, he stressed that hardware is not a problem but network is problem, in fact people working for network is problem and promoting physics community in Asia is the most important thing in a sense that Asia people do collaborate with EU or US people but not with Asian people this is the one of reasons why we still do not have good connectivity among Asia sites.
Chinorat also confirmed Brij’s question that perfsonar instance is installed and registered.
After Hsin-Yen’s presentation on ASGC site report, he added that ASGC has connection to TEIN through Hong Kong with 2.5G and it will be upgraded to 10G.
Amol asked about the architecture concerning the number of nodes of OpenStack implementation at AGSC and Hsin-Yen commented that the detailed on the architecture of OpenStack implementation has to be asked to ASGC expert while he added that the cloud system implemented at ASGC is now in production.
After Toru’s presentation on Hiroshima site report and domestic network environment of Japan, he added that SINET will keep the 10G line, for the moment, to Singapore to link with the TEIN community.
After Saif’s presentation on COMSATS site report, Sang-Un asked about the difference of PAKGRID and COMSATS and Muhammad answered that PAKGRID serves CMS as an associate member while it also provides services to ALICE.
After Edoardo’s presentation on LHC services and status, William added that the technology is well understood currently while the challenges are P2P and Edoardo added that the problem of P2P is because of multi domains.
Concerning on security issue, e.g. authorization, and the importance of the link with very well defined bandwidth allocation and protection for P2P in case of VPN(?) extension were mentioned.
In the presentation of Michael’s concerning LHCONE site connections, the thick lines of ESnet LHCONE Reachability represents the number of prefixes advertised by peers.
In the presentation of Edoardo’s concerning LHCONE in Asia, he pointed out that traffics in Asia should stay in Asia to keep path short and he proposed two solutions: use all paths currently available in Asia and each network providers allow transits traffics concerning LHCONE.
Latchezar added that some of Asian countries, e.g. India, Taiwan and Korea, already have well established networks to EU and US while going to EU or US directly means competition with well established sites in EU and US so staying in Asia would be more potentially beneficial to Asia sites.
According to Patch Lee’s presentation of TEIN status for LHCONE, TEIN has 4 POPs in Mumbai(IN), Singapore, Hong Kong and Beijing(CN) and its NOC is located at Beijing (Tsinghua Univ.) and it reaches GEANT in Madrid from Mumbai with 2.5G.
TEIN will upgrade all paths up to 10G soon and already in procurement phase: SG-EU (procured in October) to be resilient by SG-IN-EU (10G), SG-TH will be 1G while current 600M link remained, HK-AGSC (currently 2.5G) will not be upgraded for the moment due to limitation of resources (budget?)
Patch Lee added that currently there is no direct link to Kreonet but 1G connection will be established by the end of this year.
TEIN has a plan to upgrade SG-US and SG-EU links up to 100G and HK-US connection in discussion.
TEIN configured 3 links for VRF to India, ASGC and Tokyo while to GEANT not yet operational (to be done as soon as SG-TH link upgraded)
* TEIN has Juinper router
** Marine cable through Indian ocean is the most expensive one
Edoardo started the discussion to review the current internet connections among sites.
In general, path selections considered based on LOCAL_PREF, AS_PATH and MED (Multi-Exit Discriminator), mutually exclusive(?)
1. local preference, greater better
2. AS-PATH shorter better
3. MED lower better
Latchezar raised an issue on unused bandwidth of OPN (technically disallow other traffics other than T0-T1 or T1-T1) and concerning the possibility to use them, US has one link to EU for T1 with logical connection one for OPN the other one for ONE in the same link.
It was confirmed that all asian sites are connected to TEIN, except China (ORIENT, its backbone for HEP) and some traffic from Indonesia goes to Kreonet via Hong Kong through commercial connection.
Upon the proximity, some sites connected to Hong Kong while some others to Singapore.
Pilippines connected to Hong Kong with 1G.
Some added that Singapore, Malaysia (CMS), Hong Kong (ATLAS) and Pilippines are interested in WLCG.
Edoardo pointed out that we are trying to solve at least the problem concerning the sites presented here and if it works, the solution can be extended.
It was confirmed that TEIN, Kreonet and ASGC have their own route to US and EU and the three parties were agreed to transit one another once they establish their VRF for LHCONE.