Case for e-Science Studentships at UCL
(P.Clarke, S.Bhatti, J.Crowcroft)

We request the award of one or more e-science studentships to work in the area of high rate data replication and the network services which are essential to the operation of most, if not all, Grid applications. This studentship bid is made jointly between the Physics and Computer Science departments at UCL. The two groups have matched interests, which provide an excellent base to host studentships to work in an area which is core to e-science operations and manifestly essential to the delivery of PPARC objectives. The HEP Physics group brings a very focused application (the Particle Physics and EU DataGrids) and a demonstrated interest in the networking issues associated with Grid applications. The Computer Science department brings a wealth of expertise in the entire area of data transport protocols, QoS and traffic management techniques, which will underpin all Grid operations The research programme we are following encompasses many several key areas, each of which will offer a research student(s) will gain valuable and practical experience:

High rate/High volume/Reliable Data transport applications. Grid data rates will exceed 100s of Mbit/s over long latency routes. Such transfers will rely upon the availability of high rate/high volume/reliable data transport applications. All three elements are key: the combination of rate and volume are clear, but it is also essential that such applications can recover from faults and complete the transport. The transport protocols will also have to deal with efficient replication and update to multiple sites. It is already apparent that protocols such as “standard” TCP based FTP are unlikely to be adequate and therefore today we do not know how to satisfy the DataGrid demands. This is an area where significant work has been done in the CS community based upon modeling and controlled measurements, however it is vital to understand these effects in the context of real Grid traffic patterns running on the wide area network (WAN). The work includes measurements and comparisons to theoretical models; investigation of the suitability of enhanced TCP based applications (e.g. taking account of explicit knowledge of resource availability and congestion notification) and addition of reliability strategies to the use of FTP. In the longer term there are proposals to use non-TCP based transport over the long haul parts of the route and for multi-site replication and update. The research will focus upon DataGrid testbed requirements between UK sites and from UK to CERN and the USA.

QoS and traffic management: Grid operation will require new services of the network fabric. In the immediate future  (DataGrid testbeds) there will be need for traffic management mechanisms to provide managed bandwidth over IP. There are various techniques including IP packet classification, DIFFSERV and MPLS, but none have yet been demonstrated in the Grid context and in particular across multiple administrative domains. In this context both groups are involved in a collaborative project involving UCL, Manchester, CLRC, UKERNA and which is being submitted to a leading supplier to the networking industry for support. The project aims to demonstrate traffic management services between main Grid Tier-N sites in the UK using the SuperJANET4 development network, and if possible extend this onward to CERN and the USA. This project will address multiple administrative domain issues and stress such end-to-end services using both pseudo and live Grid traffic. [Note: this work is core to all Grid activities. In this respect we expect to approach both the OST and DTI].

In the longer term Quality of Service (QoS) provision is probably crucial to HEP applications. QoS depends upon traffic classification at ingress to the network, and suitable treatment of aggregated classes at all points within the network. Eventually applications, which can respond to explicit congestion notification, must be developed. This Networked systems using IP is an area where the UCL CS group has wide and internationally recognised expertise (see http://www-mice.ucl.ac.uk/), including active participation within the IETF.

Network information services: Grid matchmaking services rely upon the availability of resource information. Such resource information includes knowledge of the static (i.e. latency, packet loss) and dynamic network conditions (i.e. on demand measurement of throughput between specific nodes).  Both groups both have prior experience, which can be used to develop a suitable information service for the DataGrid. A pilot project is underway to develop an LDAP based solution.

Data management and replication middleware: We are starting work based upon real testbed requirements in 2001, to demonstrate the use of LDAP based meta-data catalogues access to remote data from a local application using Globus services (GASS and GRAM) which link to http, ftp and other transport applications. This work will initially focus on the CDF testbed.

We have a wide programme of work directed at general e-science requirements and is by construction working toward PPARC e-science goals. The PhD studentship programme we propose would give experience in all of these areas through immersion in the combined programme. We envisage that the specific research topic to concentrate upon data transport middleware and protocols, and the provision of network information to the middleware layer. We have also a potential PhD CASE studentship programme concentrating upon the traffic engineering aspects of our programme. This will be the subject of a direct CASE application later but we would readily accommodate this now if more than one studentship is awarded (converting to a CASE later).

The studentships will be jointly supervised by Physics and CS.  The academic staff involved are:

Peter Clarke: Reader in Physics, HEP LHC group leader, active in HEP Object Oriented applications and networking for 4 years, EU-DataGrid PTB member and WP7 representative, coordinating UK network fabric activity and working on the PPARC DataGrid proposal.

S.Bhatti: Lecturer in CS, researches into QoS (applications and networks), network management, network security and mobile systems.

J. CrowCroft: Professor in CS. Research in multi-media Communications, traffic engineering and QoS. Member of the ACM and the British Computer Society, a fellow of the IEE and the royal academy of engineering and a Senior Member the IEEE.

Thu, 9 August, 2001 2:11