From A Student Perspective May 6, 1997

How does CWRUnet live up to what should be provided?
      The purpose of this section is to assess the current status of CWRUnet – how we see that the Department’s goals are, or are not, being met.

      In the spring of 1996, the Administration announced what was hyped as the "Next Big Thing" in CWRUnet technology – a broad, aggressive push towards Asynchronous Transfer Mode (ATM) technology. One of the design goals of CWRUnet was to build a cutting-edge network, and keep it that way. Nine years ago, Ethernet technology represented the forefront of computer networking. Forecasting on the future, Information Services recently decided that ATM would be the new standard in networking technology, supplanting Ethernet. Consequently, aggressive plans were made to both convert the current CWRUnet infrastructure to ATM, and make sure that all future growth of CWRUnet was ATM-based.

      Before implementation, ATM was billed as a "faster, next-generation network," something that would return the University to the cutting-edge of technology and keep it there for awhile. As it stands now, 31 ATM switches and 640 ATM-attached hosts later, CWRUnet’s new topology has barely achieved one goal, and is currently missing the other. First off, performance gains have been severely limited. One reason for the under-whelming performance of ATM is due to the fact that the bulk of its traffic must pass through an emulation layer, LAN Emulation (LANE). ATM technology is natively connection-based, and under LANE, is forced to behave like a typical connectionless LAN. This is necessary in order to keep current network applications functioning, and to facilitate communication with Ethernet-connected hosts. This gives rise to another contributor to poor performance – the fact that most ATM cells need to undergo a conversion to Ethernet frames, thus imposing severe latency penalties. Finally, CWRUnet’s overall performance is impacted by stability and congestion issues inherent to ATM technology.

      In terms of "next-generation" networking, it can be argued that ATM will not become the next big transport mechanism for end users. Corporations, Universities, etc. are currently sticking with Ethernet and its next-generation derivatives. This quote, from the recent "PC of the Future" issue of PC Magazine, emphasizes this point:

Fast Ethernet rolled over competitors and emerged as the favorite way to link file servers, databases, and Web servers into local networks. …Fast Ethernet is now a core technology that will power networks far a decade to come." (Frank J. Derfler, Jr., March 25 1997, PC Magazine)

      It seems as if the rest of the world is sticking with enhanced Ethernet technology. However, CWRU’s investment in ATM can be validated if ATM-specific applications are developed. Currently, there aren’t any ATM-specific applications running on CWRUnet, and it doesn’t appear as if there are any planned for the future.

      On the whole, CWRUnet’s Ethernet segment has remained fairly stable. Due to its large size and bridged nature, however, it is still hampered by some performance problems. Some Ethernet switches have been installed recently, improving the situation somewhat. However, broadcast storms are still problematic. Also, with the advent of ATM, CWRUnet’s backbone has been changed from very reliable, proven Fiber Distributed Data Interface (FDDI) technology, to Switched Ethernet, and now to ATM. Consequently, large-scale ATM crashes can wreak havoc with all of CWRUnet, as opposed to just squashing the aforementioned 640 ATM-attached nodes.

      Alongside the migration to ATM, the CWRUnet Access Servers also experienced a migration from the Serial Line Internet Protocol (SLIP) to the Point-to-Point Protocol (PPP). First and foremost, it should be noted that this was definitely a smart decision. PPP is currently the industry standard for providing remote access to networks. Thus, considering how advanced CWRUnet’s LAN topology is, this upgrade was necessary for its dial-up links to keep pace.

      The new dial-in method, PPP, has several advantages over the older SLIP-based system. Firstly, PPP is a far more robust protocol, able to support a number of protocol layers (IPCP, IPXCP, NetBIOS, etc.) through one common network layer. Also, PPP has many strong security and end-user authentication schemes. Most importantly, PPP is currently the industry standard, and looks to be the basis for the remote access platform of the future. Consequently, it is likely that future improvements in remote network access technology will occur on the PPP platform. Therefore, by moving to the PPP remote access platform now, Information Services has built a solid foundation for the future. Unfortunately, this move has been met with some resistance among the CWRU community.

      Aside from the usual problems accompanied by any change, there were several "outstanding" difficulties that have troubled the bulk of dial-in community. The use of remote access is on the rise, as more and more students decide to both live off campus, and try to do their work from home. It should be obvious, then, that utilization of the remote access servers has been increasing, a trend that is likely to continue. Unfortunately, the number of available lines, or ports decreased in the transition from SLIP to PPP. Consequently, the new PPP pool wasn’t able to support as many simultaneous users as the old pool, causing one common complaint: horrific busy signals that deny access to the network.

      Alongside with the port contention issues, the dial-in portion of CWRUnet has also faced some other reliability challenges. All remote access users must be authenticated in order to login. Currently, CNS is using TACACS-style authorization, which basically means that there is one machine that must be asked whether or not somebody can login, for every user on the system. Whenever this machine goes down, all remote access users are effectively locked out of the system. Along with this point of failure, the current access server platform chosen for PPP seems to have some disagreeable modems. Remote access users frequently report modems that randomly drop their connection, or connect at low speeds (<19200 baud).

      Concurrently, CWRUnet dial-in seems to be somewhat inflexible, in terms of services offered. For example, CNS/LIT doesn’t seem to officially support the use of network software (software library) via a PPP connection. This isn’t to say that the PPP link can’t actually perform this task, rather it hasn’t either been tested or it has just been ignored. Admittedly, running IPX applications over a modem link is rather arduous (read: slow), but it can be done, given enough patience on the part of the user. And adventurous users have been able to get these and other protocols to work over the university-supplied PPP link. It would be nice if these would officially supported, and more applications were provided via the dial-in link. In doing so, the large disparity between on-campus and off-campus service could be significantly narrowed. Consequently, there is plenty of room for CWRU’s remote network access service to improve, and given how useful and viable it is, improve it must.

      On the plus side of things, earlier problems with off campus connectivity (extremely high load on single outbound T1) have recently been remedied by the addition of 4 other T1’s, bringing the total up to 5. Additionally, a 155Mbit, OC-3c link into the Internet has been promised, but not delivered at the time of this writing. Fortunately, this isn’t a very large issue - 155Mbits of bandwidth is far greater than most WAN links that make up the internet, and thus this new link would probably tend to be underutilized.

      Now we move onto the historically thorny issue of Service and Support for network users. There are currently two methods of getting Service/Support: the Service Desk, and the newsgroup, cwru.net.general. The newsgroup is useful in that it appeals to a small niche of students who are knowledgeable enough to both get on the network in the first place, and to ask appropriate questions. As it stands currently, this form of communication between staff and students works pretty well. However, since the newsgroup caters to such a small portion the community, the other option, the Help Desk, must pick up the slack.

      In general, the CNS Help Desk has been oriented to the end-all of Service/Support at CWRU. General hardware problems along with network software problems are directed to this service. As a result, the Help Desk seems to be overwhelmed. Typical experiences with the staff behind the Help Desk are usually less than pleasing. More often than not, the staff is rude, condescending, or unresponsive. Phone calls frequently go answered, and if a message is left, the call is rarely returned. A number of users are frustrated and angered by this level of service.

      Fortunately, there are other ways in which students can get help with CWRUnet issues. For example, a rudimentary form of Computer Based Training exists on the university web page. There are sample tutorials in place dealing with hooking up the fiber optic cables, connecting to CWRUnet, obtaining a personal CWRUnet ID, etc. This content may or may not be effective because besides being hard to find and little known, one needs a working computer in order to see it!

      Other support options provided that seem to work well, include the Kelvin Smith Library Help Sessions, and the CWRUnet in a Nutshell tutorial. When used these are very nice additions to the Service and Support portfolio. Unfortunately, not very many people fully utilize these services. Also, some of the material (especially in the Nutshell manual) may be outdated.

      Finally, there seem to be some disparities between the way that the administration treats access to internal information. The CWRU community seems to be left in the dark regarding network issues, whenever possible. There are no diagrams of network topology, no documentation on the composition of CWRUnet, etc. In fact, there is also very little interaction between students and CNS staff in general. The newsgroup, cwru.net.general, is still the only source of network information/interaction. Aside from Mr. Gumpf’s monthly ATM status updates, there isn’t even a regular source of network information on the newsgroup! It is pretty obvious that strong ties between CNS and the community need to be fostered.


Prev   Next


Table of Contents:
[ Abstract | Why CWRUnet? | Services | Status | Solutions | Conclusion ]


Comments?