The ALV’s Abstraction layer is equivalent to the MVC’s Model layer, while the ALV’s View component layer merges both View and Controller layers of MVC. The ALV’s links connect abstractions and their views, using references to remote objects. Consistency in ALV is kept Figure 4. Model-View-Controller-Session (MVCS) ob- jects showing an external broker maintaining session states. We have tested ALICE’s remote collaboration and nav- igation services in the several VR systems in our lab, and also in sessions with the Girona University (located 100 Km. from the Barcelona campus) through a 1OMb wide area network connection. In our lab we have available HMDs, a stereoscopic table, a CAVE, a MiniVR system and flat displays; and a similar setup at the Girona campus. The results obtained from our tests can be seen in the following table. The scene used on these tests (the interior of a ship) contains 50.000 polygons, but on purpose does not have complex textures that could skew graphics performance. The table shows the results obtained in the communication of 2, 4 and 8 workstations using unicast addresses from both sites. In the table we observe the average total number of mes- sages sent through the network in a series of repeated navi- gation trials, each test lasting 4 minutes. The total network time (in milliseconds) gives information about how much time ALICE spent in the transmission of messages during these 4 minutes tests (this means that only around 0.1-0.2% of total time was spent in network communications). The roundtrip time is also indicated in milliseconds. Since for this test we use unicast addresses, roundtrip time increases as more peers participate in the session. Finally, the table shows the average rendering framerate achieved for each Figure 5. Peers collaborate on an inspection (the “camera” and “upecito” avatars), as seen from an incognito client’s viewscene. Videos of peer interaction may be seen at hitp://crv1.Isi.upc.es/ vtheok/siacg04/.
