Date: Tue, 28 Oct 1997 15:16:08 -0800 (PST) From: Elise Marie Meyer To: ken lindahl Cc: David.Wasley@ucop.edu, rgr@stanford.edu, cenic-tpg@ucdavis.edu Subject: Re: CalREN-2 design On Tue, 28 Oct 1997, ken lindahl wrote: > however, i'm concerned that there might be a non-engineering requirement > for ATM: when the NSF proposal was being developed, researchers from the > various campuses were asked to describe their applications that would > benefit from calren2 connectivity. if any of those researchers was > "promised" ATM as part of calren2, there may be a requirement for ATM at > that campus. i would be interested in hearing what campuses consider > themselves in that position. At UCSB we have at least one researcher that needs access to native ATM services. Will research projects get access to ATM services (if that is used for the inter-campus network) or just IP services? Will other campus involved in these projects be providing ATM access at their end? I am including a message drafted by one of our researchers with one of these projects. Elise Marie Meyer Department of Physics Santa Barbara, CA 93106 To: CalREN2 Technical Planning Group and CalREN2 Campus representatives, The CalREN2 project aims to upgrade the wide-area networking infrastructure available to California-based research institutions. Each of these institutions is also expected to improve their local network infrastructures to allow their users to benefit from the increased backbone network bandwidth. There are several competing technologies available for local- and wide-area networks, including Gigabit EtherNet and ATM. The CalREN2 description is rather vague about what networking technology is to be used, and what services and protocols are to be provided. It states that the IP protocol is to be the basis of the wide-area service, but also that the provision of ATM-like quality of service (QoS) is to be explored. I believe that many members of the UC R&D community have assumed that CalREN2 backbone would use ATM technology, and may be impacted by the choice of a non-ATM technology for the wide-area connections. Both Gigabit EtherNet and ATM technologies are "moving targets" with ever-changing features and price-performance figures. At present, it seems that Gigabit EtherNet is less expensive to implement, but that ATM technology provides better scalability and more advanced services. If the decision were to be made to use a non-ATM backbone for CalREN2(meaning that long-distance ATM "pure" connections were impossible), it would have negative impacts on all those projects that have assumed wide-area ATM network. These groups fall into three categories: (1) those needing end-to-end ATM QoS negotiation (and service guarantees), (2) those wishing to use non-IP protocols (such as GIOP-over-ATM), and (3) those desiring ATM-specific services (e.g., dedicated PVCs with user-controlled VP/VC managemen t). Future IP enhancements such as the RSVP protocol may address some of the requirements of users in category (1), but not the needs of groups (2) and (3). While it must be acknowledged that most users only (currently) require "fat pipes" for existing IP-based services such as FTP and telnet, there are nevertheless a good number of projects that now require (or will require soon) wide-area ATM QoS or ATM services. I include an informal and incomplete list of these below. It is my recommendation that the issue of the technology base of the backbone be raised at the highest level, and that the question be posed to all CalREN2 partners as to which sites have projects that require ATM QoS, ATM services, or non-IP protocols. Stephen Travis Pope CREATE/Music/UCSB stp@create.ucsb.edu -=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=- UC Projects/Partners requiring ATM QoS, ATM services, or non-IP protocols (This is only a sampling of the projects that I'm aware of or involved in.) UCSB NGNM: Next-Generation Networked Multimedia Project (non-IP + QoS) PMP: Parallel Media Processing Project (non-IP) Stone: Media Databases and Interaction (non-IP + QoS) IDIOT: Impact of Distributed Object Technology (services + non-IP) DRIVE: Distributed Real-time Interactive Virtual Environment (non-IP + QoS) HARP: Hypermedia Acoustical Reality Project (QoS) T&L: Thunder and Lightning (services + QoS) UCSD Transit: Sound Spatialization (non-IP + QoS) MusD: Distributed Sound/music databases (services + QoS) PD: Pure Data distributed signal processing language ( QoS) UCB CMT: Continuous Media Toolkit (QoS) CAST: CNMAT Audio Synthesis Tools (services + QoS) LLNL NTONC Consortium UCLA VR WorldServers (services + QoS) (There are certainly others.)