Transcript The requirements of researchers using networks

SERENATE
work item i:
User Needs
Investigation into current and future requirements
of researchers using research & education
networks
Study carried out by the Academia Europaea and
the European Science Foundation
The report on User Needs will be
based on two inputs:
1. The replies to a widely distributed
questionnaire
2. The presentations and discussions in the
SERENATE Workshop for end-users of
research networks (Montpellier, 17-19
January 2003)
The Questionnaire
• About 4000 researchers were invited to answer the
questions in a Web-based questionnaire
• About 500 replies received:
–
–
–
–
–
–
–
Physics 95
Chemistry 40
Technology 45
Life and earth sciences 56
Medical 110
Social sciences 67
Humanities 37
The Questionnaire
some highlights (1)
• 43% say they are not (yet) using highbandwidth networks, 57% say they are using
high-bandwidth networks
– Of the latter group, 48% mention accessing distant
databases, 15% mention distant processing e.g. at
supercomputers
• 37% transfer only rather small files (< 1 MB),
14% regularly transfer rather large files (> 100
MB)
The Questionnaire
some highlights (2)
• What would be the most important network
development for your research in the next 3, 5 and 10
years?
– 41%: more bandwidth, removal of bandwidth bottlenecks
– 5%: improved access of resources (e.g. from countries with
less than average access
– 13%: Grid computing
– 11%: larger or more distributed databases
– 8%: videoconferencing, use of distributed video material
– 5%: virtual working / virtual laboratories
– 5%: remote modelling, real-time visualisation
The Questionnaire
some highlights (3)
• What would the impact be to your research if network
speeds were increased by one or two orders of
magnitude (say, international connections at 10 or
100 Gb/s?
–
–
–
–
1%: negative impact (less time to think)
15%: “I have no idea”
24%: no major impact
45%: positive effects, namely:
•
•
•
•
•
Remote working with collaborators: 8%
Remote or distributed computing: 7%
Videoconferencing: 4%
Accessing databases or moving data: 3%
Remote control of equipment: 2%
The Questionnaire
some highlights (4)
• Is your use of the research network currently limited
by the international connections, the national
network, the regional/metropolitan network or the
campus network?
–
–
–
–
–
–
43%: there are no serious bottlenecks
15%: there are bottlenecks but I don’t know where
23%: the bottlenecks are in the campus network
7%: the bottlenecks are in the regional/metro network
11%: the bottlenecks are at the national level
13%: the bottlenecks are at the international level
The Questionnaire
some highlights (5)
• Who should be paying for research and education
networks and how?
– 91%: there should be no charging at the point of use:
•
•
•
•
12%: the government or EU should pay
5%: the university/institution should pay
74%: no further comment
“networking should be free, just like water and electricity”
– 9%: there should be some form of charging:
• 4%: according to use
• 2%: as part of research grants
• 2%: only heavy users should be charged
End-users Workshop
Conclusions
WORKING
DRAFT!!
David Williams
CERN, also President TERENA
SERENATE End-users Workshop, Montpellier
19 January 2003
Introduction
• Despite the very high quality of the attendees
we were disappointed by the size of the
attendance at this workshop
• But the quality of the presentations and
discussions from those that did make the
journey made the effort worthwhile.
Overall
• The workshop wishes to acknowledge the strong
progress that has been made during the past five
years, leading to a rather reasonable environment for
R&E networking in many parts of Europe
• While there are many reasons to be grateful (to
governments and to the many different organisations
involved in funding and providing these services) we
absolutely cannot be complacent. The situation in
several areas of Europe is far from satisfactory, and
arguably getting worse, and overall developments in
the networking field are so rapid that very strong
efforts are required to maintain services in a healthy
situation
Three key components
• The workshop recognises that satisfactory network
performance for the end-user depends on a healthy
infrastructure in three areas:– On the campus
– Nationally
– Internationally
• Unless indicated otherwise, in this report the word
“infrastructure” should be understood in a broad
sense – not just cabling and transmission capacity
but also all software and services needed to provide
the services required by the end user
Campus-related issues
• There is serious evidence that for many European
researchers the major source of network
performance limitations is their local campus.
The NREN model
• The workshop recognises that the NREN
model - that one NREN is necessary and
sufficient for each (European) country - has
served well and remains appropriate
The European REN model
• In the final session of the workshop we were asked to address
the extent to which a model of multiple interconnected NRENs
would still be optimal in ~10 years time. Would it not be better
to envisage, at some point in time, a move to a single,
ubiquitous,European REN?
• Because of time pressure we agreed to return to this topic in
later workshops. Issues that might complicate such an approach
include the national political, funding and regulatory regimes,
and local languages. It was also pointed out that there may be
some concerns related to scaling, and that the structure of the
network tends to be inherently tiered.
• It was suggested that the correct approach is probably to
examine the functionality required at the various levels of the
network, and a study along that line will be prepared by
SERENATE.
Growth in requirements
• The workshop was very impressed by the
evidence for the growing “requirements” of
all areas of research
• It is clear that these “requirements” will
continue to grow dramatically, over the next
5-10 years, IN ALL DISCIPLINES and IN
ALL COUNTRIES
The move to services
• The workshop noted that user expectations have
evolved beyond the provision of pure bandwidth
towards the supply of more complex services, and that
it is important that NRENs take this evolution into
account
• Some examples of required services are given in the
following slides
Authentication services and good
remote access
• The workshop takes note of the move towards authentication (and related)
services, commonly referred to as AAA services
• Since end-users increasingly require good access wherever they may be
(especially from home and while on missions) the workshop looks forward to
the deployment of a set of AAA services, compatible across Europe (and,
more generally, on a global basis).
• It also encourages NRENs (and campuses) to provide good broadband remote
access
• One good example of the benefits of such an approach could come when
universities having paid (sometimes very significant prices) to provide closely
controlled licensed access to specific data resources for their staff, are able
make them available remotely to those staff without infringing the terms of
their license.
End-to-end service
• If an end-user experiences unexpectedly poor
performance while making international data transfers it
is normally hard for him/her to understand the reasons,
and to initiate corrective action. One major complication
for the user is the fact that five separate organisations are
typically responsible for different components of any
international data transfer.
• The workshop would like to see the NRENs (and
DANTE) cooperate to put in place a service which, when
triggered by the user, would determine whether the
performance obtained was indeed worse than normal,
and, if so, take responsibility for correcting the situation.
Flexibility
• The workshop noted with great interest the present experiments
being conducted by GÉANT, by several national networks, and
by the radio-astronomy community, to combine the signals from
several European radio telescopes.
• The workshop feels that this example, which fundamentally
incorporates the networks’ long distance optical fibres into a
new scientific instrument, is a very good and successful
demonstration of the need for NRENs to be attentive to user
requests and to adopt a flexible approach to service provision.
• The first ideas of the atmospheric physics community to moreor-less continually monitor lightning production on a global
basis form another example in the area of data acquisition.
• However, the request for flexibility is not related to just
communications hardware – it refers to all NREN services
Enhanced services and the relation between
NRENs and the R&E community
• The workshop notes that the trend towards the
provision of enhanced services (rather than simply
communication bandwidth) means that the NRENs will
need to engage in a deeper dialogue with campuses and
with their national body/ies responsible for IT
infrastructure
• Besides AAA services (mentioned above) we discussed
the extent to which NRENs should become more
involved in initiatives aimed at reducing spam and
protecting against viruses and hacking.
Education and Information Flow
• NRENs need to balance their approach to meeting two
potentially conflicting end-user requirements
– The wish to view “networking” as an always-available, highly reliable
and easily understood service, tending over time to simply increase in
performance and decrease in cost – “networking should be invisible”
– The need to be informed about changes in network technology and
services since they can have an important impact on the way in which
research is carried out
• The workshop felt that NRENs should increase the flow of
information, including road-maps of likely future service
developments, to their end-user communities, and make more
educational material available.
• The workshop also felt that in addition NRENs should devote
more resources to listening to their end-users, and actively
engage in dialogue with them.
Scientific communities and IT services
• People thinking about the broad introduction of grid services over the next 510 years are aware that each scientific discipline will need some internal
decision making process or structure if it is to obtain the expected benefits.
• Decisions will be needed, for example, in areas such as metadata definition;
ontologies describing common concepts in the discipline; the “default”
assumptions about the conditions under which computational or data
resources may be shared; resource discovery mechanisms that are compatible
across Europe.
• The workshop encourages the EU and others (ESF?, AE?) to think about
establishing small working groups to facilitate this process at the European
level. Such teams should be aware of any national work in their discipline,
and be able to speak with some authority for the discipline in the European
context.
• It such teams were formed, then their specialists on grids and networking
would, in future, form the obvious invitees for any future SERENATE enduser workshop.
User influence on NRENs
• The workshop discussed the ways in which end-users can and
should influence their NREN organisations
• As stated previously, there is a recognition that the NRENs are
generally operating efficiently and effectively, and end users do
not feel a strong need for more “control” over their NREN
organisations
• However, as also discussed above, there is a belief that the
NRENs would do well to better inform and consult their endusers. The general feeling of the meeting was that a better
“feed-back loop” should be constructed.
Closing the gap (1/3)
• There is very clear evidence that we are very far away from
“equality of networking opportunities” for researchers from all
countries in the European Research Area.
• There is a widely held view that the relative situation (for
example, the bandwidth available to researchers in the
economically weakest countries compared to that available to
researchers in the economically strongest countries) is still
degrading
• There is good evidence from our survey that researchers all over
the ERA wish to do similar work, and that the resources of
researchers in the economically weaker countries of Europe
represent an intellectual force which is being seriously underutilised because of poor computer networking infrastructure
Closing the gap (2/3)
• Computer networking and grids are all about improving human
communication and resource sharing
• The workshop feels strongly that the provision of good IT infrastructures –
including networking and grids – is one important factor among several
which can influence the migration of high quality researchers.
• One important outcome of the SERENATE project must be to draw the
dramatic nature of this gap – Europe’s internal “Digital Divide” – to the
attention of politicians
• The workshop notes that regulatory liberalisation in the EU-15 played a
major role in reducing prices and improving services, and we have to insist
that the elimination of telecoms monopolies, and the very rapid introduction
of effective competition among several operators, will be crucial factors if
this gap is to be eliminated in the next say 5 years.
• However, any attempt to implement “equal opportunities” across the ERA in
the next ~5 years will depend on strong political commitment, and spending
significant sums of money. Can we estimate how much money would be
needed to have a significant impact?.
Closing the gap (3/3)
• The workshop suggest that the annual reports being produced for the
accession countries should be required to specifically address the extent to
which their R&E network has been brought up to EU levels.
• In order to demonstrate the size of the gap, and to monitor its evolution and,
hopefully, its eventual disappearance, we recommend that the EU should
commission a project to provide monitoring data for end-user performance to
and from a selection of universities and research centres in the countries
involved and in the EU-15 for comparison.
• Examples of the sort of data that we have in mind can be found via
http://www-iepm.slac.stanford.edu/
Unexpected gaps
• While we think of most of the EU-15 countries as well-connected,
workshop participants mentioned that there can still be some expected
problems with certain R&E sites.
• Highly specialised instruments and/or institutes which are located offcampus are often poorly connected. While we heard of some dramatic
examples of this effect close to major cities, the situation tends to be
worst in very rural or mountainous locations, or on islands.
• The situation is highlighted when wanting to connect to such
specialised sites with high-performance high-reliability links – for
instance as part of a grid infrastructure.
Communities beyond research and tertiary
education? (1/5)
• SERENATE has been asked to look at how communities related to those of the
traditional NREN communities (which we can define as publicly-funded
research and tertiary education) could be served.
• This is a complex issue, and it is hard to imagine a uniform pan-European
approach. The most that we can do is probably to indicate some general
principles which may (or may not) be applicable in a given country.
• We start by reiterating the reasons to have an NREN in the first place
– R&E computer networking requires a specialised infrastructure (not just hardware)
for collaboration among researchers, and this infrastructure is not available as a
“standard service” from commercial providers
– The community is tolerant of experimental services at the leading edge
– The last ~20 years have shown strong economic benefits resulting from using the
community as a motor for change
Communities beyond … (2/5)
Different ways to connect
• Three possible approaches for handling the NREN
connectivity requirements of new communities
(shorthand for groups of users who do not obviously
fall into the classic NREN communities of publiclyfunded research institutes and post-18 education) are:– Direct connection to the NREN
– Connection to a dedicated network which has wellengineered connectivity to the NREN
– Connection to the commodity Internet, with ist connectivity
to the NREN
Communities beyond … (3/5)
Possible methodology
• A possible methodology would be
– to look at each possible new community and decide to what extent it
meets any of the requirements bulleted above (1/5; why we have an
NREN)
– to determine the extent to which strong interaction between the R&E
community and the new community is desirable
– to consider issues of of physical size and scale – in countries like
Luxembourg and Slovenia there are few universities and quite strong
economies of scale in handling at least secondary education via direct
connection to the NREN
– to consider issues of required services and expertise – all communities
need more than simple hardware connectivity
– to think about applicable financial models
– and come to a conclusion
Communities beyond … (4/5)
• Whatever approach is adopted in a given country it
seems clear that each NREN will have to pay more
attention to its interconnection technology, services and
infrastructure for supporting these “related
communities”
• We have to recognise that the choice (does community
X have direct connection to the NREN –yes or no) may
be hard to make. There will always be grey areas, and
to some extent this is an issue for national taste
• Basically we need to ensure that all communities
wishing for good connectivity to the core NREN
community are able to interface effectively to it.
Communities beyond … (5/5)
Examples of communities that are “interested” (in many countries)
and for which decisions need to be made include:• 24*7 access for researchers and post-18 students
• Schools (at all levels)
• Musea, Archives, Memory-preservation
• Hospitals/Healthcare
• Ministries
• Libraries
• SMEs in science parks
• SMEs in arts and creative industries (Berlin cf London)
Charging
• SERENATE recognises that there is no free lunch. Networks do cost money
• There is a strong reluctance to charge for “standard” use at the point of
delivery. Basic research networking is a common good.
• Some people feel that it might be acceptable (maybe even desirable?) for
specialised heavy (de Laat Class C) users or any others with special network
needs) to be able to contribute to funding for those special needs.
• No clear consensus emerged at the workshop on this last idea. As a minimum
such “advanced use extra contribution” should only be introduced when the
situation is well understood and stable– both the financial issues and the fact
that the technology works. Up-front charging introduced too soon is a good
way to stifle progress and advanced applications.
• For many disciplines the network will become part of the “computer” in the
future
Things the NRENs should also do (1/2)
• Develop (in conjunction with academic community)
split of tasks between NRENs, any “national grid
organisations” and campuses.
– Authorisation (AAA) - compatible across Europe
– Anti-hacking, Anti-spam
– Better understanding of standard copyright and related legal
issues, including DB access rules. Make policies more
obvious to end-users. Update “user charters”
– Possibly lobby (politically) against NRENs being held
responsible for the content that they transport?
– Develop and reinforce best-practice use (what is this
supposed to cover?)
Things the NRENs should also do (2/2)
• Encourage a slightly more formal organisation of
research disciplines in their approach to IT
infrastructure. Would be useful both nationally and on
a pan-European basis
• Topics to be considered could include:–
–
–
–
development of metadata standards
and discipline ontologies
evolution of networking requirements
definition of default resource (data, CPU, etc) sharing
policies for grids
SLIDES FOR SERENATE
IN GENERAL
(NOT SO MUCH FOR END-USERS)
Technology
• The main point is that greater deployment of optical
networking seems inevitable.
• This will provide several opportunities
– Support for mixed functionality (all-to-all “conventional
access” for Class A and B users) and more dedicated links for
Grid-like and VPN-like (Class C) users
• Present tests to optimise single TCP/IP flows towards 1
Gbps end-to-end looks as though they will lead to
production deployment over the next ~12 months
• Need to pay a lot of attention to issues such as mobility
(not so much access while in a car/train, more
accessing all net facilities from any location – inside
country and out) and security
Regulatory Issues
• The 1998 EU regulatory package (“Bangemann
deregulation”) is in the process of being significantly
modified
• The new regime was adopted as a set of EU Directives
in 2002, and is required to be implemented nationally
by July 2003 at the latest
• For NRENs one very important issue is whether they
wish (or will be) considered as a “public” or as a
“private” network. The position/choice may vary in
different countries.
Regulatory – “Own Fibre” (1/2)
• Directives are – in principle – supposed to be
technology neutral
• One detail which is not tech-neutral is that in some
countries the last-mile access regulations - requiring
operators with SMP (significant market power) to offer
reasonable access to either public or private networks –
only apply to copper infrastructure and not to fibre
• The Directives make it clear that NRENs can – should
they so wish – install their own infrastructure, and
specifically fibre
• So there are no telecoms regulatory barriers to NRENs
physically installing their own trenches and fibre
Regulatory – “Own Fibre” (2/2)
• However, any NREN planning to do that needs to carefully
investigate the local rules for RoW and construction permits. It
may take some time – say 6-18 months - to organise physical
installing. Longer delays should not be possible (in EU and
countries aligned on EU rules).
• We cannot exclude that in some countries old NREN statutes
would have been framed to exclude the possibility of the NREN
owning infrastructure. NRENs should probably check for this.
• While true “own installation” can be a useful bargaining
position, we imagine that most NRENs would much prefer to
acquire any such fibres from (conventional or alternative)
operators on a medium to long-term leasing basis.
• Or as a wavelength service
Security Issues
• We are increasingly expecting that Internet services
needed by the R&E community are provided from
remote locations, and accessible from “anywhere”.
• Campuses are investing significant resources to repel
human and algorithmic break-in attempts
• Authentication and authorisation technology is starting
to be deployed which is a fundamental improvement on
widespread and incoherent
The end
End-users Workshop