General Background 3

Technical State of the Art

In the previous paragraphs, the technical state of the art of this sector has been explained. As regards the HPCN technology more details are reported in the next section.

Please note that, in the field of music, there exists several European projects mainly belonging to Library Area:

MOODS project will take into account the results produced by these projects and in particular by projects CANTATE, E-MOLL and HARMONICA. This is guaranteed by the presence of SHYLOCK among the partners of this project. SHYLOCK has been a partner of CANTATE, E-MOLL projects, and it is presently a partner of HARMONICA project.

HPCN Technology Related Issues

According to the preparatory study for this project {LIOOEVOLUTION96}, it should be noted that a huge quantity of information has to pass on the network connecting lecterns in conditions of score transferring due to loading/reloading of refrains. By considering 200 lecterns and the dimension of a score page in electronic format (5000 bytes), an acceptable transfer time (1 second) is obtained with at least 10 Mbps (bit per second) of throughput. Given the high cost that a fault can produce during the execution of public performance, the communication must be strongly reliable. To this end, real-time constraints will be imposed on each transmission; this leads to select for the low-level support a solution consisting of a network having a limited time of wait, such as FDDI (ISO 9314), Token Ring (802.5), 100 VG-AnyLAN (802.12), ATM, {100VG96}, {ATM95}, {Nassar96}, {Bird94}.

On the basis of the early analysis, the network support must provide: (i) a throughput bigger than 30 Mbps; (ii) a very fair access and bounded latency, (iii) a predictable behavior (jitter free or limited), (iv) a support for monitoring network behavior and reliability in real-time, (v) the possibility of connecting at least 200 nodes without decreasing point-to-point performance (as it happens in token-based networks, FDDI and Token Ring), (vi) a low-cost solution (differently from FDDI and ATM solutions), (vii) the possibility of using both twisted pairs (low cost) and optic (noise immunity) wiring, (viii) a possible evolution towards faster versions in the future, (ix) the possibility of implementing fault tolerance networks, (x) a mature technology, (xi) the possibility of using broadcast and multicast addressing.

For these reasons, the adoption of HPCN technology for providing a safe and fast communications among lecterns/editors, is mandatory.

Moreover, with the preliminary study the 100 VG-AnyLAN was selected for its high predictability, throughput (100 Mbps nominal with 95 % of actual utilization) and reliability. In fact, by using this network support, fault tolerant networks can be built. Moreover, high reliability and predictability are obtained, since this allows the use of two levels of priority: normal and high (it is based on Demand Priority protocol instead of collision detection such as Ethernet and Fast Ethernet, CSMA/CD protocol), and has bounded latency and intrinsically supports the monitoring of the network by means of specific nodes. The topology is a HUB-based star with a maximum of 5 levels.

Unfortunately HP have announced the abandon of 100 VG AnyLAN technology from the market. With this notice, and with the aim of maintaining low the cost of the lectern the only possible solution is the fast Ethernet, 100 BT. Moreover, since all work for the MOODS prototype and project will be performed by using TCP/IP and PVM, the substitution of the 100 BT with a faster network will be possible with low expenses.

At higher level, please note that, several supports for distributed systems have been proposed in the market -- e.g., CORBA, OLE Microsoft, etc., but they are not suitable for guaranteeing the fulfillment of real-time constraints. In addition, these supports are quite heavy to be adopted since they present

several internal levels of software which in practice impede the exploitation of capabilities of 100 VG-AnyLAN. This is mainly due to the fact that such supports for distributing objects are mainly focussed to implement non-safety critical applications.

For these reasons the DSI has implemented a low-level object-oriented support based on PVM {TheNET96}, {PVM394}, {Chang96}. This support will be customized and optimized for constituting the low-level network support of MOODS.