Minutes from the 1^st NWG
12 May 2000
Mayantigo Building, Santa Cruz de La Palma

Introduction
Current Network Topology
Current Network
Future Network Requirements
Network Design and Alternative Providers
Security
AOB
Date for Next Meeting

Introduction

Meeting opened by Rene Rutten director of ING, describing reason the panel was created and, why it was felt that we should meet at all. Namely Operations Sub-Committee (OSC) have had various discussions in the past, on issues that ere felt to lie outside the technical expertise of members of the committee. After the last OSC meeting it was decided to set up an advisory group to inform the OSC, of networking difficulties and solutions. At this point DC was appointed chair of the meeting and CM secretary.

During this introduction a series of points were passed to the meeting from the OSC for discussion namely;

1. The speed of the existing link. Concerns were expressed about the reliability and speed of the existing link
2. Current topology of the network and the technology involved.
3. Development plans, it is clear that the future networking needs of User Institutions (UIs) need to be considered and that the IAC, as the connectivity provider must be involved in these discussions.
4. The existence of bottle necks on the network connection, and what can be done to ease the load on them.
5. Use of alternative providers, for Internet access and the link between the SLO and ORM.

CM at this point expressed concerns about the nature of the problems on the network. The exact nature of the perceived problems and the existence of hard figures was queried. This statement was supported by other members of the NWG. Dr. Rutten at this point admitted to a lack of hard figures for network connectivity, and network throughput, making reference to a graph produced at the OSC meeting by the HEGRA representative. JC proceeded to show the graph in question, and it was agreed to table a discussion on connectivity issues later in the meeting.

Dr Rutten left the meeting at this point. DC formally took over the meeting at this point and raised the issue of responsibility for the network and at which point each UI, is responsible for the network. Concern was expressed about the lack of a more formal agreement defining areas of responsibility and largely agreed upon by all UIs. Discussion on this issue, was again tabled for later discussion.

DC also mentioned his belief that the NWG were an advisory panel reporting to the OSC, thus all we could do was recommend a series of actions to the OSC and let them decide on the ultimate solution, as they had the ability to actually spend money once advised by the NWG.

Diego Sierra proceeded to inform the members of the meeting on the current topology of the network both at the ORM, and to the rest of the internet. DS also detailed proposed changes to the network both at the IAC, affecting the ORM directly and some of the changes outside the IAC.

DS informed the meeting that the IAC have currently undertaken to document the existing network layout, and the future plans for the same. It was agreed by DS to make this information available to all the members present at the meeting once it was finished, and some form of access was discussed and agreed.  Plans for an improvement of the link between the ORM and La Laguna, via some provider other than Telefonica were mentioned. Although no clear plan was mentioned through a lack of information, the process is thought to be still in the development phase, and will more than likely turnout to be some sort of public bid. Therefore it is important for UIs to express the future needs to make

this process effective.

It was also mentioned that the connection plans for the GTC were unclear, i.e. whether or not they planned to use the existing link to the Internet or had other plans was unclear to all members present at the meeting. Clearly their plans affect the future network topology and bandwidth availability.

Discussion of the network topology and changes then led to a resumption in the discussion about the network responsibility. AZ and DC both expressed concerns about the lack of clear boundaries of responsibility, and also the lack of notification when work was to be carried out on the network. DS agreed at this point to notify all responsible persons (i.e. those present at the meeting), of planned work on the network, before actually carrying out the work.

The meeting then returned to the issue of network connectivity, and whether or not there really was a problem. CM stated that in his opinion a large amount of the perceived problem did not really exist as no one could really draw up hard figures for connectivity time.

In addition the issue of network throughput was one that is often pointed to as being a problem area, but in reality no one has hard evidence to support this claim.

Some members agreed and added that without hard figures it was impossible to discuss this issue further. IS stated that the Liverpool John Moores University team had been trying out network tests for the last two years and found that the network performance was more than adequate for their needs, he then showed a paper to the meeting summarising LJMs results.

CM pointed out that he had in place a very simple script to transfer a 1.0MB and 0.3MB file, from ORM to Stockholm, the 1.0MB test file being chosen as a more representative file size for transfers, and the 0.3MB used when the network load was high. It was proposed that a similar test be carried out collaboratively by the UIs, with a view to getting a clear idea of the network throughput to home institutes. All members agreed to standardise the tests, and use the same script and file.

The issue of connectivity was then discussed, JC stating that on occasions he had lost the link during the day. DC stated that he had figures that showed the connectivity was better than 95%, and offered to make the figures available to all, however the exact down time figure was hard to determine. CM asked DS if it was possible to supply members with figures for connectivity and where certain faults lay, i.e. at the IAC or Madrid (RedIRIS), he said that past figures were not possible but he could certainly make future down time statistics available.

DC asked for SNMP access to the IAC router, so that when/if the router goes down he has the necessary notification. DC also agreed to make this information available to all members, after several requested access.

LJMT will send approx 50MB of data after every night. The data will be compressed and reduced as it impossible for them to send raw data. A high level of connectivity is very important to the telescope, for this reason they are interested in alternative link to the mountain i.e. 64kbps modem, ISDN etc etc. However IS was unable to quantify an acceptable level of connectivity that would be required. He also accepted that 100% connectivity was an impossible target.

Discussion then turned to the availability of ISDN on the mountain, and DS sated that there were some ISDN sockets on the IAC router.

Future requirements unclear, although the possibility of a new form of observing mode (service) was mentioned. The only data transfer mechanism envisioned at present is the ftp transfer of raw data, although this is by no means clear. The lack of definite future plans, makes it difficult to be exact about future network requirements.

TNG have no plans to transfer raw data anywhere, although their future requirements are unclear as well, as they have just changed director. AZ expressed concern about the lack of information from GTC,both in the initial and final operational phase of the telescope. This concern was mirrored by all present.

MAGIC (17 m telescope) will be ready summer 2001, it will produce 20MB/s and there are no plans to transfer this data via the network.

At present HEGRA transfer 100 MB of data a night and have no problems with this. The advent of MAGIC, will not increase this in any way as the data rate produced by the telescope is not conducive to ftp.

The only future requirement is for a high level of connectivity although he was unable to provide an exact figure for acceptable levels of connectivity.

No future requirements over and above those already in place. Diego Sierra then informed the meeting of the organisational changes at the IAC, the old Centro de Calculo, has been divided into sections providing separate services. DS is now in charge of networking issues, Carlos Martin security and Pablo Lopez Unix system administration. The networking group is now composed of four people and the cover should be better although there will be no cover at weekends and outside IAC working hours as there is no provision at the IAC for this.

The issue of security was briefly brought up, and discussed. General consensus was that it was better dealt with at another meeting as the topic was too big to discuss at the current meeting. Salient points discussed and to be forwarded to the security meeting were;

1. Some of the networks on the mountain are too open, and this is unacceptable. Some form of protection should be in place.

2. Some form of minimal security policy (ORM wide), implemented by all UIs. Measures over and above this to be the responsibility of each UI.

3. Common firewalls (yes or no?)

4. SSH etc

IS explained that no matter what the other UIs settle on they will implement some form of individual firewall solution. The nature of their operational model (remote observing) makes this essential, and their requirements are easily achievable.

Currently Telefonica provide the link from La Laguna to ORM. There was little alternative to this until recently. IAC have plans to put in some form of new link, the new carriers being sought by tender action. Therefore it is important for all UIs to provide IAC with requirements as they are drawing up the document on the tender action at present. DC asked for clarification on this process i,e, whether it definitely means another carrier or another provider.

Meeting closed.

Actions

20000512.1 DS : Mail all present at meeting every time he is going to work on the network. Make statistics of downtime of link between La Laguna and Madrid available in the future. Network topology documentation (present and future) to be made available to all UIs once finished.

Make SNMP connection available to DC.

20000512.2 DC : Make last years network down time figures available to all.

20000512.3 IS : Make LJM network reliability/throughput paper available to all in postscript format and mail all UIs when ready. Make 1.0MB uncompressible file for CM to include in his script for UIs network throughput tests. Supply NWG with exact figure for acceptable level of connectivity. Send security requirements to NWG for security meeting as he will not be attending.

20000512.4 JC : Supply NWG with exact figure for acceptable level of connectivity. Send security requirements to all UIs as he will not be present at the security meeting (neither will the other HEGRA representative Toni Coarsa).

20000512.5 CM : Send network throughput test script to all

20000512.6 ALL : Think about requirements for network usage both current and future, ad the level of security required so that definite proposals can be sent to OSC.

7. Date for Next  Meeting

Security Meeting : 11 o'clock 27/06/00 Mayantigo.

NWG : Friday 10/11/00 La Laguna Starting time to be agreed.

File location : http://www.ing.iac.es/~cfg/ormnwg/20000512.html

Minutes written by : Claudio Moreno
Revised and converted to HTML by : Don Carlos Abrams