From A Student Perspective | May 6, 1997 |
How can CNS meet these expectations?
The CWRUnet access servers are swamped. Fortunately, the administration has a
new plan in order to fix things: the recently publicized multi-tiered service
scheme. The issue of remote students, faculty, and staff paying for a normally
free service is very complex, and must be handled carefully. It cannot be
stressed enough that the administration needs to establish a better rapport
with the students. Through this link, they would be able to more fully explain
their position, and be able to better accept feedback and constructive criticism.
This must be imperative so that any plan enacted can achieve a broad base of
support from the community.
First off, many of the issues surrounding dial-in access to CWRUnet haven’t been made clear by the Administration. For example, no hard data has been released concerning the utilization of the current dial-in pool, projected budget for remote access equipment, etc. Also, nothing has been published regarding what type of research Information Services did regarding alternatives to the current remote access system, and their decision process was also never revealed. There are many ways to charge users for dial-up access, yet it seems like this current plan was chosen out of convenience. Let’s take a moment to summarize the plan in question:
(note: time limits will be enforced)
Pool 1 - the current 92 modems
Pool 2 - minimum 1 modem per 10 subscribers
Pool 3 - minimum 1 modem per 8 subscribers
Pool 4 - minimum 1 modem per 4 subscribers
[Jeff Gumpf, cwru.net.general, 3 Mar 1997 19:35:15 GMT]
The current plan, as we have been given to understand it, seems to be born of convenience, and doesn’t seem very well thought out. Any plan for revamping CWRUnet’s remote access service must strive to be easy to access, reliable, fast, and easy to use. Ease of implementation shouldn’t be a key factor. This plan is convenient because it is fairly simple to implement, requiring no major changes to the TACACS authorization servers. Billing is simply assessed according to what service the subscriber signs up for, and the University already has such a billing infrastructure in place. This plan doesn’t seem well thought out, because it appears as if its implications haven’t been carefully considered. By separating out the heavy users from the less demanding users, the normal bell curve distribution that defines port utilization will be disrupted. It may be, that all of the heavy users will pay the monthly fee for a low user to port ratio, and then spend long hours connected – locking out all of their similarly paying brethren. Furthermore, this plan only addresses the problem of inadequate user to port ratios, neglecting all of the other problems with CWRUnet’s access servers.
There are many other ways to implement a fee-based dial-in service. One such example is currently being implemented at the University of Minnesota (http://www.nts.umn.edu/services/modembilling.html). Basically, Minnesota is in the process of implementing a pay-as-you-go scheme. In essence, the first 30 hours of connect time a month are free. From then on, users must pay an hourly rate in order to connect to the network. Online usage and billing is freely available to users. The system also gives warnings, and if so configured, blocks access when the 30-hour limit is reached (so that users don’t "accidentally" incur charges).
To summarize, the ability to remotely access CWRUnet is one of the most vital services that the University provides, considering that the majority of the student population lives off campus. Furthermore, recent changes to the remote access functionality of CWRUnet have been both good and bad. It was a good move for Information Services to provide PPP rather than SLIP-based access. However, it was a bad move to shrink the number of ports available to the users. Consequently, Information Services needs to placate their user base, and implementing a multi-tiered payment scheme in order to lower the user to port ratio isn’t going to do it. Information Services needs to re-examine the current situation, and try and accomplish several things. Firstly, the current PPP dial-in pool should be expanded, at no cost to the user. If the necessary expansion cannot be accomplished within the given budget constraints, then an alternative method must be fully researched. However, as we have seen, before going ahead with any plan it is imperative that Information Services seek as much input from the CWRU community as possible – in order to help the plan become a success.
We have already evaluated the current status of the ATM-based portion of CWRUnet, it has been received in a less than stellar light. The current ATM infrastructure hasn’t been met with wild admiration from the CWRU community. However, with so much money already invested in ATM technology, it is impossible to turn back. Therefore, this section deals with the current plans for the expansion of ATM on CWRUnet.
Careful consideration must be taken regarding the expansion of the ATM network. When ATM was first implemented, it was so raft with problems that it was more or less unusable for an entire semester. This lack of usability could have been avoided if some better planning had been done. For example, one of the first problems encountered were the horrifically overcrowded Emulated LANs (ELANs). If an ATM load generator had been utilized, enough traffic could have been generated on the ATM network to adequately simulate expected usage, thus predicting the overcrowding problem. Thus, from the example of first semester, it is apparent that a great amount of foresight and planning needs to be given to expansion of the network. Hopefully, some performance testing can be done, to determine in advance how an additional 800 new ATM-attached hosts will impact the network. In this scenario, it should then be possible to eliminate points of failure in advance. Consequently, an exemplary level of service could be provided to incoming freshman right from the start. However, it must be stressed that simple foresight cannot eliminate all future problems. For example, another fault of the ATM network first semester was an insidious bug in the Edge Switching code between the campus ATM switches and Ethernet hubs. Therefore, if at all possible, the expansion of the ATM network should be retarded, until the ramifications of edge switching and general traffic management are more understood. This would be a great cost saver. Currently, ATM technology is rather expensive, but the price is coming down. Waiting even another 6 months could save the university hundreds of thousands of dollars. Furthermore, slowing the growth of the experimental arm of CWRUnet would give the engineers much more flexibility in dealing with unforeseen problems as they crop up. Also, since fewer students would be on the experimental network, fewer people would be adversely affected by such problems.
However, in the event that expansion of the ATM network cannot be limited, steps must be taken to ensure reliability. One way to increase the robustness of the ATM network involves implementing the LANE 2.0 standard as soon as it becomes available. This upgrade to the current LANE standard adds many additional stability and reliability features. For example, under LANE 2.0 it is possible to maintain backup LAN Emulation Servers (LES) on the network, eliminating a possible point of failure. Providing ATM-attached servers would also enhance robustness. The benefits would be twofold. First, failure of the Edge Switch wouldn’t necessarily cause ATM-attached hosts to be separated from their data. Secondly, ATM-attached hosts would be able to reap some of the benefits of ATM technology, by transferring data at increased speeds.
Alongside the reliability issues, CWRUnet’s ATM branch is also having some problems reaching its "next generation" network goal. In order for ATM to be considered a truly ground-breaking, next generation network, important new ways of using the network, that rely on key ATM technologies, must be implemented. As currently implemented, an ATM connection into CWRUnet isn’t much different from an Ethernet connection. Most people do the same things with their ATM connection that they did with their Ethernet one (TCP/IP, IPX, etc.), only in some cases they’re doing it a little bit faster. And as we’ve previously mentioned: most people judge networks, and computers in general, not by how they’re made, but by what they do. Thus, Information Services needs to explore new and groundbreaking applications, and their potential effects on education.
ATM technology holds the promise to enable several new forms of communication, thus creating many different educational opportunities. This is great, considering the fact that CWRU is a university, and one of the university’s primary goals is the education of its students. The fact that ATM can guarantee both bandwidth and data delivery, along with its low latency, enable many applications in the area of real-time voice/video delivery. Complete lectures could be given over CWRUnet, broadcast to the computer of every student in the class. And with real-time MPEG2 compression and decompression, these lectures can even be interactive. Other areas to explore include, but are not limited to: Voice/Video over IP, Multicast, "Server Push" technology, and Videoconferencing. In more detail, Voice/Video over IP consists of using standard TCP and UDP packets to transfer real-time audio and or video over the network. This has advantages over previous technologies in that it can work over any transport mechanism that supports TCP/IP. Multicast technology is simply a mechanism that allows one host to transmit the same data to multiple hosts. In essence, this allows group applications, in which many people participate in a common environment. Server push technology represents the latest trend in the evolution of the World Wide Web. Basically, instead of clients actively seeking data on the network, they simply subscribe to "channels". These channels then dump information directly to the end user, much like a television. Any one of these technologies might become the "next big thing" in computing.
With the search for new applications, many great educational and research opportunities would be created. There are many among the general student community who aren’t knowledgeable about computer networking, and ATM specifically. Yet these technologies are intertwined with the future of computing. Therefore, any experience that students can gain from CWRUnet should be considered beneficial. Moreover, many of the aforementioned technologies are still in their infancy, and much research has yet to be done. This is another great opportunity for the CWRU community. In fact, for a university that thrives on leading research, it shouldn’t be too hard to truly enter the forefront in another fundamental technology of the future: computer networking.
Other areas of research exist with in standards groups, such as the ATM Forum, and within advanced networking projects, such as the Internet 2 consortium. Although it should be noted that CWRU is on the list of universities lending support to this project, as of yet, it doesn’t appear as if any real work has been done on the part of Information Services. Judging by the early documentation provided at the Internet 2 website, it appears as if the latest incarnation of CWRUnet puts the University is in a unique position. All of the ATM technology that has been installed should allow the University to leap to the forefront of this project, if Information Services were so inclined. There are many other areas of research that Information Services should consider participating in, such as the vBNS, and research projects with the National Laboratory for Applied Network Research (NLANR).
The methods with which Information Services attempts to both provide service and support for CWRUnet seem to be fundamentally sound. Unfortunately, one in particular gets bogged down in implementation: the CWRUnet Help Desk. The short falls of this service have already been noted, but unfortunately, the possible solutions seem murky. Firstly, the problems with the Help Desk could be boiled down to a simple staffing issue. It could be entirely possible that the current staff cannot deal with the volume of calls that they receive on a daily basis. If this is the case, then it should simply be a matter of re-apportioning funds in order to obtain proper staffing. However, these problems might be cased by a simple lack of focus – the mission goals for the Help Desk may not emphasize courtesy and respect of customers.
Information Services seems to be addressing this apparent shortfall in management style, with their plans to "revamp" the Help Desk for the 1997-1998 school year. Current plans call for Digital Media Services (DMS, formerly Library Information Technologies) to take control of the service desk. It will be moved to the new Kelvin Smith Library, and they are actively seeking a full-time support engineer. If properly carried out, these changes could fix the perceived shortcomings of the CWRUnet Help Desk.
The issues surrounding network reliability have been an ongoing theme throughout this document. It should be obvious that mission-critical servers, and more importantly, the services they provide, need to be available when needed. Therefore, it seems obvious that these vital servers should be run on stable, robust hardware platforms, and should have viable backups whenever possible. Many of the changes discussed can be implemented simply by changing management strategy. Transport, or the physical media on which data is moved, should be de-emphasized. Content services should be the main focus of CWRUnet. Quite simply, this means that the administration needs to recognize the fact that the CWRU community simply wants to get their work done, regardless of the "leading edge" manner in which their bits get there.
Table of Contents:
[ Abstract |
Why CWRUnet? |
Services |
Status |
Solutions |
Conclusion
]