Connection Failure IET : Remote Access
Connection Failure

Remote Access

RAMP Report, 10/12/98


Table of Contents   |    Acknowledgements    |    Section:  1   |    2   |    3   |    4   |    5   |    Appendices




Appendices


Appendix A:
Remote Access Improvement

Appendix B:
Financial Analysis Detail

Appendix C:
Diagram of Expanded Modem Pool Services




Appendix A:
Remote Access Improvement

Current Campus Remote and Network Access Improvement Projects


ResNet

RAMP

Commercial Access Services (Contracts with Internet Service Providers)

Distributed Authentication Services

Magic WAND (Wide Area Network Davis)

Wireless

Dynamic Host Configuration Protocol

Public Network Ports

Approaches at Other Universities

Comparison Institutions to Be Used

Summary of Findings


return to top



Appendix B:
Financial Analysis Detail

One-Time Costs Already Funded

One-Time Costs for Expansion

Remote Access Servers (Modems)

Communication Line Installation Charges

Installation Costs

Step Function Switch Cost

Develop Billing System

Annual Operation and Maintenance Costs

Internet

Line Charges

Depreciation

Equipment Maintenance

Technical Support

Operations and Maintenance Support

User Support

Billing

Communications Resources Overhead





Appendix C:
Diagram of Expanded Modem Pool Services



return to top

Appendix A: Remote Access Improvement


Current Campus Remote and Network Access Improvement Projects


ResNet

This project installed network data ports in each residence hall unit, providing high-speed networking to the dorms. This project has reduced the load on the campus modem pool, as residence halls previously accounted for 25% of the dial-in usage.


RAMP

The Remote Access Management Program (RAMP) was a series of pilot projects IET conducted to evaluate the benefits, costs and feasibility of new remote access alternatives for the campus.

RAMP's Remote Access Pilot evaluated:
  • The cost and feasibility of installing, maintaining, and supporting new, software-controlled remote access technologies (56K modems with ISDN capabilities);
  • The benefits to campus in achieving its instructional and research goals, through providing this level of connectivity as a campus service;
  • The effect of segmenting the modem pools among user groups and types of services; and
  • Potential funding models for such services, including recharge.
IT obtained and installed equipment for testing which provided 96 33.6Kbps modems (with pre-paid upgrades to both 56K and ISDN) to three pilot groups, largely made up of academic and non-tenured faculty who connect to the campus network from off campus for the purposes of research and teaching.

Phase A of the pilot began in November 1997 with 12 faculty participants. Phase B began in February 1998 with an additional 100 faculty and 11 Technical Support Coordinators from 11 departments. The test service expanded in May 1998 to include members of the Academic Senate and Academic Federation, and the Technology Support Coordinators who provide them technical support.

The project also involved assessing the existing 14.4 modem pool.

Commercial Access Services (Contracts with Internet Service Providers)
  • Negotiated non-exclusive agreements with ISPs for campus users.
  • Provided Recommended Solutions documents to the campus community for use in selecting commercial remote access services.

Distributed Authentication Services

In the past, UC Davis affiliates using outside Internet Service Providers (ISPs) had problems accessing restricted on-campus servers and documents. That's because access to those resources was granted based on a list of authorized UC Davis IP addresses (128.120.*, 169.237.*, 152.79.*). Users connecting through an off-campus ISP don't have a UC Davis IP address; they have an IP address assigned by the ISP.

To solve this problem, IT's Distributed Computing Analysis and Support unit developed the Distributed Authentication solution. This allows webmasters and other information providers to authenticate incoming users with high-security Kerberos user names and passwords, instead of IP addresses. UC Davis affiliates can access restricted materials even if they are connecting through ISPs instead of the campus modem pool.

If a UC Davis affiliate uses an ISP to connect to campus, they will need a UC Davis Login ID and Kerberos password to access restricted resources through the Distributed Authentication solution. They will also need to use a Web browser that is capable of SSL (Secure Sockets Layer) connections, such as Netscape or Internet Explorer version 2.0 or above. Information on the UC Davis Login ID and Kerberos password can be found at http://computingaccounts.ucdavis.edu/. General information on the Distributed Authentication solution, a pointer to a page to perform a browser test and UC Davis Login ID and Kerberos passwords can be found at http://www.ucdavis.edu/authentication/.

The following table and footnotes will help webmasters and service providers understand which features of distributed authentication can be used with a particular combination of hardware and software.

Features Supported
Server Basic Pass Through Middle-ware High Volume High Security Logging
UNIX            
Apache ¤ ¤ ¤ ¤ ¤ ¤
NCSA ¤ ¤ ¤ ¤ ¤ ¤
Netscape ¤ ¤ ¤ ¤ ¤ ¤
             
Windows NT            
Apache ¤ ¤ ¤ 2 ¤ ¤
Netscape ¤ ¤ ¤ ¤ ¤ ¤
IIS - ISAPI ¤ 1 6 ¤ ¤ 5
IIS - Perl ¤ ¤ ¤ 2 ¤ ¤
             
Macintosh ¤ ¤ ¤ 3 4 ¤
             


Notes:
  1. The current version of ISAPI does not allow the authentication process to set the REMOTE_USER variable. This means that subsequent authorization processes cannot distinguish securely between a successful and unsuccessful authentication. By default, IIS is installed with the ISAPI. The Perl tools must be installed explicitly.
  2. Calling Perl as a CGI module from an NT server imposes a performance penalty.
  3. Macintosh systems do not currently offer true multitasking, and are generally not stable as high volume servers.
  4. High security requires that the authentication module be able to read specific information from an AFS directory. Current Macintosh systems cannot do this.
  5. Logging of authentication attempts requires either an API that can open and write files in a specified directory, or the use of the Perl version of distributed authentication. The current version of ISAPI does not allow the authentication software to specify a location for the log files.
  6. The ISAPI solution only works with middleware if no further authorization is required. See note 1 for more detail on this issue.
Explanation of Features
  • Basic: Features supported by distributed authentication on every platform, including access control lists, automatic authentication by IP address, and Kerberos authentication. Authenticated users can access protected directories, and unauthenticated users are barred.
  • Pass Through: Both authenticated and unauthenticated users are granted access, and a local authorization process decides which specific resources each user can access. The authorization process must be able to determine the result of the authentication process in a reasonably secure way. This is accomplished by having the server set an environment variable called REMOTE_USER. A user that bypasses the normal authentication process and attempts to fool the authorization process into granting access to restricted resources will not be able to set the environment variable. This allows the authorization process to distinguish between successful and unsuccessful authentication attempts.
  • Middleware: Software that mediates between the Web server and other software, such as a database. Cold Fusion is one example of middleware. While all versions and configurations of the distributed authentication will work with middleware, not all middleware is available for all platforms. Cold Fusion is only available for Windows NT and Sun Solaris UNIX.
  • High Volume: The server handles high levels of traffic, with significant concurrent access.
  • High Security: The distributed authentication software can provide several levels of security. The highest level of security requires that the user have a valid "cookie" in their browser, and then checks the contents of that cookie against a file in AFS space that can only be written by the campus Kerberos server. To take advantage of this high security option, the Web server API or Perl must be able to read files in AFS directories.
  • Logging: A log of all authentication attempts, both successful and unsuccessful can be maintained.

Magic WAND (Wide Area Network Davis)

The project involves working with local property owners at off-campus apartment complexes in Davis to install high-speed connections (ISDN or better), on a pilot basis. The project team also provides technical and other assistance to apartment complex property owners to help encourage the installation of high-speed connections to the Internet from small computer labs at apartment complexes and from students' apartment residences.

Students at such apartment complexes would be able to use the lab computers or plug their own computers into network connections in their apartments to check electronic mail, obtain on-line course materials from campus computers and conduct on-line research via the high-speed network connection.

A team of IT staff, apartment complex property owners, and community organizations developed four pilot sites, where small computer labs, located in apartment complexes, had ISDN connections installed. An additional pilot will extend network connectivity to adjacent study halls and apartments, through a wireless access point. Students with notebook computers and PC/PCMCIA card slots would be able to check out wireless cards from the complex office or purchase wireless cards, and connect their machines to the Internet through the wireless access point.

Based on the pilot's results, team members will develop a Guide to Networking Apartment Complexes in order to educate local apartment complex property owners and managers about the benefits of Internet connectivity for their business and their residents.


Wireless

This project will expand the range of communications service options that are available to the campus involving wireless technologies. Wireless communication gives its users greater flexibility and mobility and provides an alternative to the traditional constraint of being bound to a specific physical location. The voice service users would be able to program their desktop telephones so that all calls to their campus number are forwarded to a cellular-like device capable of receiving calls at any location, whether the user is on-campus or on the other side of the country. Wireless data users would have network access from almost any location on- or off-campus within range of a signal transmitter. Wireless services, however are limited by their speed and cost. For example, wireless data communications tend to provide speeds similar to modem dial-up access but can be among the most expensive of the connectivity options.

While numerous campus wireless pilot projects have been carried out over the past four years, wireless voice and data network access services are finally feasible as a regular type of service, because a sufficient variety of shapes, sizes and capabilities are appearing on the local market. A campus wireless services project team was appointed for the purpose of surveying user interest and vendor capabilities, to conduct pilot testing services and to perform a cost/benefit analysis. On the basis of these criteria, the team made recommendations about new wireless services for the campus. Other tasks included the identification of the Voice/Data functional characteristics that are considered desirable as well as the pricing and cost options that must be considered prior to a service rollout.

A team of staff from IT and other university departments, was assembled to formally address this subject in detail.

Wireless technologies were evaluated in three pilot projects including spread-spectrum/narrow-band microwave wireless, cellular wireless, packet data wireless and specialized integrated voice/data/pager wireless devices and services (e.g., PocketNet). For each, a pilot test plan was developed. In addition, team members conducted a current market survey (inventory of technology and services), and a survey of wireless projects and services at other institutions. Based on the research and the pilots, the team completed cost-benefit analyses and user cost profiles, and made a final report with recommendations to the Director of Communications Resources (CR).

The team is evaluating the introduction of AT&T's Wireless Service in the immediate future, with additional installation of microcells to improve campus coverage. One reason for an early rollout is to provide the benefit of government pricing to potential users of the service. It would also be a way of gauging the potential demands for wireless service.

Test plans were developed for each of the three pilots, pilot participants have been recruited, equipment has been identified and loan/purchases are underway.

A survey was distributed to wireless campus customers at two phone number swap exchange sites at the beginning of March 1998. The survey is being analyzed for usage information and for interest in different types of wireless service technology.


Dynamic Host Configuration Protocol (DHCP)

Dynamic Host Configuration Protocol is a protocol that, through a server, automatically assigns IP addresses to computers as they connect to a network. If this service were provided at UC Davis, campus notebook computer users whose equipment's Ethernet card was registered with the campus, could more easily access the network as they move around between office, classroom, library and home.

The major deliverable for the project team is a report which evaluates the costs, benefits, and recommended architecture of any DHCP services potentially provided by IT, in the Network 21 post-cutover environment. The report would include recommendations as to whether IT should deliver such a service. A preliminary report that was completed during Spring of 1998 covers three activities:
  • Results of evaluating and testing a service that allows units such as student housing to administer a large pool of semi-mobile computers, and implementation of the service for the residence halls as part of ResNet.
  • Results of evaluating a support service for the mobile computing end-user, to handle IP address assignment for mobile devices (laptops, network labs, faculty machines moved between offices and classrooms).
  • Issues associated with a centralized DHCP support service for departmental network administrators and lab administrators.
While DHCP service has already been implemented for the residence halls, any campus-wide implementation of a centralized, IT-maintained DHCP service as well as a distributed support infrastructure would not occur before potential costs and support issues are thoroughly evaluated.


Public Network Ports

This project:
  • Provides an alternative to computer labs for students who own a computer notebook, by providing network access ports in convenient locations (libraries, lounges, and other on-campus public spaces).
  • Helps campus departments develop this service for their areas by providing information and support on cost, maintenance, and monitoring of site usage.
  • Results in a well-maintained campus-wide list of notebook access ports available to the campus community.

Approaches at Other Universities

Information gathering in this area is still underway.


Comparison Institutions to Be Used
  • Blue Cross/Blue Shield
  • Duke University
  • Harvard University
  • Massachusetts Institute of Technology
  • Michigan State University
  • Stanford University
  • State University of New York, Buffalo
  • Texas A&M
  • University of Illinois at Urbana - Champagne
  • UC Berkeley
  • UC Irvine
  • UCLA
  • UC San Diego
  • UC San Francisco
  • University of Michigan, Ann Arbor
  • University of Virginia
  • Yale University
  • University of Pennsylvania (See http://www.upenn.edu/computing/remote/)
  • University of Maryland
  • University of Wisconsin
  • University of Texas, Dallas (UTD) (See http://www.utdallas.edu/ir/rna/)
  • Cornell
  • University of Washington
  • North Carolina State
  • Florida Atlantic University
  • University of Indiana
  • University of Minnesota
  • University of Ohio
  • University of Nebraska
  • Penn State
  • Purdue
  • University of Iowa
  • Northwestern
  • Arizona State
  • University of Phoenix
  • University of Oregon
  • Memorial University Newfoundland (See http://www.mun.ca/cc/dcg/modem.cfm and http://www.munet.mun.ca/)

The following universities participated in CAUSE remote access survey in 1996 (see http://www.georgetown.edu/acs/people/hassler/)
  • Florida State University*
  • University of Alabama/Birmingham*
  • University System of Georgia*
  • Brigham Young University*
  • George Mason University*
  • Georgia State University*
  • Hamilton College*
  • Lewis & Clark College*
  • Pepperdine University*
  • Rice University*
  • University of Montana*
  • University of Tennessee/Chattanooga*
  • University of Tennessee/Martin*
Back to top   |   back to Table of Contents.


Summary of Findings

Of 34 institutions planned for the survey, two are presently discounted for lack of data (Blue Cross and University of Phoenix). We have at least partial data from Web sites and e-mailed correspondence on the remaining 32 sites.

Subsidized Access
  • 17 institutions offer fully subsidized remote access on some level
  • 8 institutions offer partially subsidized access, either in the form of UNIX-only access or by offering a limited amount of time, with charges for additional time
Charged Access

Only one institution, Penn State, offers access on a paid basis at a rate comparable to ISPs, and it's not apparent from the Web page whether that rate applies to students or only to faculty

ISP Dependent Access

Two institutions are entirely ISP dependent. University of Michigan offers "free" access that is paid for through a student technology fee. Michigan State offers limited free access, with a charge for additional hours.

ISP Partnerships

Eight additional institutions list an arrangement with an ISP for remote access. A number of institutions seem to be passively encouraging ISP use by offering only low end services (both Purdue and Arizona State offer only UNIX access) or impossible contention ratios (Duke has 128 modems to service 32,000 accounts).

From the University of Texas, Dallas, webmaster (UT Dallas offers 72 free modem lines and a good ISP partnership): "

We're trying to get out of the business of maintaining modem pools.

1. They break/die too frequently.
2. They need constant firmware upgrades.
3. Faster speeds require newer hardware investments annually.
4. Too many man-hours required for good support."

From Arizona State University, which offers only UNIX access to students, the statement:

"ASU encourages individuals to use a commercial ISP."

ISDN Services

Both of the following alternatives are currently being used:
  • Institutions offer ISDN only through an ISP.
  • Institutions offer their own ISDN.
Resource Allocation

We didn't specifically look at resource allocation, and there is no clear pattern on handling it. Many institutions limit access time by session length, by monthly usage or both.

Memorial University of Newfoundland, in response to a 1997 study, has devised a "Fair Allocation" program (15 hour time allocation to start, plus 35 minutes per day; users can bank up to 15 hours in their accounts. Additional hours can be purchased at 10 Hours for C$7.50, plus $.50 for additional hours. Usage time is charged against the account at full use during prime time, 50% from noon to 4 p. m. and 25% from 1 a.m. to noon.)

Many institutions, however, don't address the issue at all, as University of Virginia says: "

Faculty, staff, graduate, and undergraduate students all compete for access to our free modems using the same rules and probability for access."

Back to top   |   back to Table of Contents.




Appendix B: Financial Analysis Detail


Cost Descriptor Variable Costs Fixed Costs
Per Port Per User Step Unit
One Time Implementation Costs Cost Cost Cost
Remote Access Servers

374.77

N/A

N/A

N/A

Line Installation

99.04

N/A

N/A

N/A

Installation Costs (PR Tech Midrange - $58,838/yr)

7.66

N/A

N/A

N/A

Step Function Switch Cost (per 24 Access Servers)

N/A

N/A

6,564

N/A

Develop Billing System (PAIII Midrange - $68,928/yr - 4 months)

N/A

N/A

N/A

22,976

Total Costs

481.47

N/A

6,564.00

22,976

Annual Costs
Internet

N/A

2.75

N/A

N/A

Line Charges

207.18

N/A

N/A

N/A

Communication Server Depreciation

124.92

N/A

N/A

N/A

Existing Router & Switch Depreciation

N/A

N/A

N/A

32,231

Step Function Switch Increase Depreciation

N/A

N/A

2,188

N/A

Equipment Maintenance

31.25

N/A

191

4,575

Technical Support (2 PAIII Midrange - 10%)

N/A

N/A

N/A

13,786

O&M Support (PAIII Midrange - 25%)

N/A

N/A

N/A

17,232

User Support (Step Increase after 20,000 users fr 3.5 to 4.5 FTE)

N/A

N/A

60,000

210,000

Billing System

N/A

2.00

N/A

N/A

Communication Resources Overhead

N/A

N/A

N/A

41,604

Total Costs

$363.35

$4.75

$62,379

$319,428

  1. Variable costs consist of those costs that are sensitive to the quantity of ports or quantity of users. These costs may scale linearly or as a step function.
  2. Fixed costs consist of those costs that are not sensitive to quantity of ports or users.
One-time Costs Already Funded

The campus has already made some investments in remote access equipment that is highly scalable and that requires only minimum augmentation at high use thresholds. This equipment includes:
  • Authentication Servers
Two SUN workstations were purchased to run the authentication programs (RADIUS/Kerberos) for an estimated maximum user population of 25,000.
  • Router
One Bay BLN-2 router was purchased for $67,221.28. This router services an infrastructure that allows for communication server expansion. A BLN-2 router could support up to four 3Com Corebuilder's. Each Core Builder can support up to 31 3Com Link Switches. Each Link Switch provides either 12 or 24 connections and each CISCO AS5300 support 96 modem ports but require two connections (one 100Mbps link and a secondary backup 10Mbps link. Therefore,a potentially, fully populated and dedicated router, CoreBuilder, Link Switch configuration can support a maximum of 142,848 ports.
  • ATM Switches
An investment of $12,569 was made for two 3Com link switches. Additionally, system design, testing, and documentation costs were funded and completed through a $40,000 contract with INS.

User support conversion costs were funded, and the amount of time necessary to assist various types of users (faculty, staff, and students) through the conversion process. Often, this involved indirect needs, such as the need to upgrade the mail package, or in worst cases, the operating system, in order to use the higher speed remote access services. The following table provides the details of the conversion support model.

TOTAL Students Faculty Staff
Number of users

18,910

16,059

1,037

1,814

Avg hr assistance/user during conversion based on pilot experience

-

0.2

0.3

0.25

Total

3,977

3,212

311

454

Phase-in period (mo)

6

6

6

6

Hrs/mo

663

535

52

76

FTE

4

3

0.3

0.5

Hourly wage (SAIV for students, PAIII for faculty/staff)

-

$7.50

$25

$25

Total Support Cost

$43,206

$24,089

$7,779

$11,338



Back to top   |   back to Table of Contents.



  One-Time Costs for Expansion

Remote Access Servers (Modems)

Due in part to the demands of the dynamic ISP market, and in part to the consolidation of higher numbers of ports/server, remote access hardware and software costs are dropping rather dramatically. For example, in the year between acquiring new modem equipment for the faculty pilot (which started in Fall of 1997), and the new modem equipment which will provide interim remote access services this Fall, per port costs of the Cisco equipment almost halved. In the summer of 1997, two modem communication servers (Cisco AS5200) providing 96 33.6/56Kbps ports (48 ports per server) were purchased, and then this summer (1998) Cisco AS5300 equipment (four servers providing 96 ports each for an additional 384 ports) was purchased at nearly half the cost per port.

An even greater concentration (432 ports per remote access server) is also available, for example in the AS5800 which could further reduce per port costs. The downside of highly concentrated port/server ratios is the single point of failure problem: the greater the concentration, the higher the number of ports that would become unavailable at one time. The optimum configuration is one that balances per port cost against preventing single point of failure.

Communication Line Installation Charges

Line installation charges include PRI interface, T1 trunking, charges for dial plan, DID, PBX trunking, alternate routing, and PBX base additional trunking. For the recent installation of 384 ports, the total communication installation charges were $38,031.50, or $99 per port. This rate is expected to remain consistent during the short-term planning horizon.

Installation Costs

A small per port cost, representing a fully-burdened principal technician is included for additional remote access server installations.

Step Function Switch Cost

Given the existing configuration, additional network switches would be required for every purchase of 24 96-port remote access servers.

Develop Billing System

If the financial model decided upon requires recharge, additional costs associated with software development or acquisition of a billing/tracking system will be required.

Annual Operation and Maintenance Costs

The model assumes a three-year depreciation cycle; that is, every three years, all of the equipment charges described would need to be replaced.

Internet

An analysis of traffic across the border router indicates that the existing population of users account for 16% of the campus Internet costs and this should scale proportionately by the quantity of users.

Line Charges

Line charges are the monthly Pacific Bell costs associated with connecting the modem pool to the public-switched telephone network.

Depreciation

The remote access server depreciation is based upon three years and is scalable based upon 96 port increments. The main infrastructure depreciation supporting the modems (Router and ATM switches) is basically a fixed cost depreciation because of its large scalability. A step function increase in switching depreciation is required with the addition of 13 remote access servers beyond the current configuration, and for every 24 96-port servers after that.

Equipment Maintenance

There has been a minimal requirement for maintenance on the Cisco equipment or the T-1 lines. All have operated throughout the entire pilot without failure. A scheduled code upgrade was performed on the communications server without impact to the users. Additionally, while troubleshooting a reported slow response, interface cards were swapped to isolate the system, again with no impact to the users. Vendor maintenance agreements with the router and switch vendors are in place for high value equipment, and the cost of a 2% sparing level is assumed for the lower value equipment.

Technical Support

During the period of pilot testing, two part time network analysts, spent approximately 10% of their time resolving authentication issues. Procurement of two new SUN Ultra30 stations provided the required computing and capacity to maintain authentication with Kerberos and TACACS. Midway through the pilot, Radius was chosen to provide permit authentication due to its improved speed in processing the Kerberos permit. It is assumed that some level of technical support will be required on a continuing bases to deal with similar technology changes in the future.

Operations and Maintenance Support

Computer Resource Specialists and student analysts were used to troubleshoot configuration and password issues with new clients. 58 of the 457 trouble tickets opened between November 19, 1997 and April 19, 1998 dealt specifically with the Remote Access Pilot (13 %). Based on this experience, the Network Operating Center (NOC) estimates that technical support costs will be .25 FTE. Included in these support costs are the following items:
  • Communications server
  • Modem interface
  • Switch, router and circuit/wiring infrastructure
  • Maintenance and triage/repair of the Remote Access System
  • Monitoring of system performance
  • Recommendations for enhancements and system tuning that may affect service level.
  • Systems support maintenance of the TACACS/RADIUS servers and assist in authentication and authorization procedures and administration.
  • Recommendations for system upgrades, scheduling and coordinating system downtime.
  • Ownership of vendor supplied services as the single point of contact for vendor management from a systems perspective.
User Support

The IT Express currently provides support to the campus and its affiliates for dynamic IP dial-up network connections (PPP/SLIP). The IT Express also supports TCP/IP to the extent that it is required to set up PPP on both the Macintosh and Windows operating systems. IT Express also pre-qualifies (gateway ping) network trouble calls before escalation to the NOC for troubleshooting at their help desk. Details of the supported software packages are available for the IT Express Web site.

IT Express support cost estimates for two user population levels are provided for
  • 18,910 users, approximately $210,000 (for labor, benefits and equipment for 3.5 FTE)
  • 28,000 users, approximately $270,000 (for labor, benefits and equipment for 4.5 FTE)
Billing

If the financial model decided upon requires recharges, additional costs associated with software licenses and modifications, administrative processing, paper, mail, etc., will be required. These costs were estimated to be approximately $2.00 per user per year although there is not sufficient data to accurately determine this figure.

Communication Resources Overhead

A portion of the administrative costs associated with the operation of Communication Resources is distributed to each service that it provides. An equitable share of these costs based upon the size of the service and the rate structure approved by the rate committee is approximately $42,000 per year.

Back to top   |   back to Table of Contents.




Appendix C: Diagram of Expanded Modem Pool Services




This diagram indicates the architecture adopted for the deployment of expanded modem services through the Provost's Office allocations.



Back to top




Table of Contents   |    Acknowledgements    |    Section:  1   |    2   |    3   |    4   |    5   |    Appendices


Connection Failure