• Switched 10Mbps and 100Mbps Ethernet provide good migration paths for future (and current high performance) machines. Switched 10Mbps and certain 100Mbps Ethernet use the same cabling facilities as 10BASE-T shared Ethernet. Switched 10Mbps usually requires no change for client computers. However, 100Mbps usually requires a new Network Interface Card and a new set of drivers. 100Mbps is very similar to 10BASE-T from a software point of view, so software changes beyond the drivers are not usually needed.

• Switched 10Mbps and 100Mbps Ethernet are scaleable solutions that are supportable into the foreseeable future.
• Migrating to high performance equipment as needed is the most economical course of action.
• Performance gains will be realized with switched 10Mbps since UCDNet connections are currently constrained to shared 10Mbps.

ImplicationMitigation1Machines could overdrive Network 21 desktop facilities. Upgrade path for more bandwidth to the desktop has already been considered and planned. Departments wishing to adopt faster technology earlier than the planned upgrade cycle may be required to pay for the upgrade. 2Cutover priorities delay migration to high performance options. High performance upgrades may be addressed as soon as possible after cutover using post-cutover support (2.2). Post-cutover operational support will have processes and rates in place for departments to move, change, add, or upgrade their facilities.

Other OptionImplication1Install high performance equipment at start

• High start-up costs associated with this requirement would exceed the Network 21 project budget.
• The higher costs conflict with the goal to achieve the highest level of campus penetration.

2Install high performance upgrades as part of departmental cutover as time permits. Installing high performance options during cutover could delay cutover for other departments.

1.6: Connecting Custom High Performance Networks to Network 21 Infrastructure

• Abstract

  Some departments have already invested resources building custom high performance networks. While such custom networks may be incompatible with Network 21, a department may request a routed bulk connection allowing continued deployment of a custom network. Requests for routed bulk connections may be granted based upon an understanding of prerequisites, conditions of use, and performance and upgrade issues.

• Policy

  Departments with custom high performance networks may request and fund a routed bulk connection to Network 21 provided the following conditions and issues are met and understood:

  • The custom network must meet the following criteria:
    • The network may not be a 10Mbps shared or switched Ethernet since this is not considered to be a custom high performance network.
    • FDDI and 100Mbps Ethernet will be considered for a routed bulk connection.
    • In order to qualify for a routed bulk connection to Network 21, the departmental network under consideration must use a significant level of high speed or custom facilities. Having an installed base of 10/100 Ethernet NICs operating at 10Mbps or converting a 10Mbps network to 100Mbps at a very limited number of nodes will not qualify as a high performance network.
  • Department must understand that:
    • Addresses and network masks must be transitioned to Network 21 format. Even though existing networking infrastructure is used, Network 21 connections will not interoperate with IP addresses associated with classical UCDNet (including 128.120.0.0, 205.227.0-31.0, 208.1.128-175.0, and 169.228.128-175.0). Because Network 21 technology and protocols differ substantially from classical UCDNet technologies and routing protocols a dynamic, reliable coexistence is not possible. The old and new campus networks will communicate as independent networks but will not co-exist on the same infrastructure.
    • A routed bulk connection to Network 21 eliminates the possibility of using Network 21 features such as VLANs, real-time services, and switching.
    • A routed bulk connection will not perform as well as Network 21's switching architecture.
    • Any and all performance problems and/or other difficulties will be the sole responsibility of the department.
    • The department may be responsible for all or part of the upgrade costs should the department decide to convert to Network 21 infrastructure or one of its high performance options at a later date.
    • The connection to the router shall be made using 100BASE-FX or 100BASE-TX. 100BASE-FX is the preferred means of connection since fiber is typically more reliable than twisted pair. The department is responsible for converting its network architecture to 100 Mbps Ethernet for connection to Network 21.
    • The department is responsible for the additional cost of providing routed facilities, including but not limited to the router, interface cards for the router, cables, transceivers, and labor.

• Assumptions
  • Departments with custom high performance networks have the resources to maintain such networks.
  • Departments with high performance networks must participate in Network 21 for campus and Internet connectivity.
  • These departments may wish to implement Network 21 high performance options when it is more economical to do so on a large scale.
• Benefits
  • Departments may continue to use their custom networks while using Network 21 to meet their internetworking needs.
• Implications / Mitigations

  ImplicationMitigation1Custom high performance networks may exceed the performance levels of standard Network 21 connections.

  • The bulk, routed connection is provided at the highest level of Network 21 capability.
  • Under normal circumstances, about 80% of the traffic on a LAN does not leave the LAN. The traffic is generated among the resources on the LAN and does not need to leave the LAN. Given this traffic distribution, the LAN's bandwidth levels may exceed those of the campus network without significant loss of performance.

• Other Options Considered / Implications

  Other OptionImplication1Connect departments with custom high performance networks to Network 21 with high performance options.

  • Forcing a department to convert to 100Mbps Ethernet would prove too costly on a large scale.
  • 100Mbps Ethernet could be lower performance than the original LAN.
  • Using high performance options may delay the cutover process.

1.7: Converting to Network 21 IP Address Space

• Abstract

  The allocation of fixed 250 host subnets on UCDNet has exhausted the 128.120.0.0 address space. With Network 21, variable length subnet masks will be implemented in order to assign properly sized subnets to a department, providing better utilization of network layer address space. Since the average UCDNet subnet consists of about 45 hosts, Network 21 subnets will tend to be smaller than current UCDNet standard.

• Policy

  In order to participate in the Network 21 project and cutover, departments shall convert their current IP addresses to their newly assigned address space. Upon completion of cutover to the Network 21 infrastructure, departments shall discontinue use of their previous addresses and return the subnet to IT.

• Assumptions
  • In general, subnets will be much smaller than the current UCDNet standard of 250 addresses.
  • VLANs will serve to consolidate remote sites into a single subnet; therefore, departments will require fewer subnets.

  Benefits

  • Proper sizing of subnets and more efficient utilization of address space will enable each department to have its own address space. The elimination of shared address space among departments will result in increased security.
• Implications / Mitigations

  ImplicationMitigation1Departments will need to exercise tighter control over their address space to stay within the size of their subnet. Departments will need to keep complete and accurate records of address assignments. IT will provide templates for tracking relevant information. IT will also provide known information about the configuration of the department's current subnet. 2Variable length subnet masks make host configuration more difficult and not uniform across campus.IT will work closely with the departmental Technology Support Coordinators (TSCs) and network administrators to reduce the difficulties often associated with variable length subnet masks. For more information about the Technology Support Program, please see the IT web page at http://www.ucdavis.edu/IT.

• Other Options Considered / Implications

  Other OptionImplication1Continue to use 250 hosts per subnet.Campus networking address space would be quickly exhausted.2Use a small subnet (16 hosts) and assign multiple subnets per department.Although this would appear to make efficient use of address space and provide a common subnet mask across the entire address space, the default router address would be different for each subnet. This presents greater configuration difficulties and uses a larger percentage of addresses for routing.

1.8: Converting to Network 21 Variable Length Subnet Masks and Routing Protocols

• Abstract

  Variable length subnet masks will be implemented with Network 21, meaning that the subnet mask will vary from site to site depending on the size of the department. Variable length subnet masks require the use of a new backbone routing protocol; however, departments will continue to receive routing updates similar to those currently provided by UCDNet.

• Policy

  In order to participate in Network 21, departments shall reconfigure the subnet mask on hosts as necessary.

  IT shall continue to provide RIP routing updates to departmental hosts. RIP routing updates will contain default route information only.

• Assumptions
  • Variable length subnet masks are necessary for a network of the size and complexity of Network 21.
  • OSPF will be the routing protocol used on the backbone to implement variable length subnet masks.
  • RIP will be the routing protocol used to present routing information to departmental hosts.
• Benefits
  • Variable length subnet masks allow better utilization of IP address space.
  • Variable length subnet masks allow for better scaling for small departments as well as large ones.
  • RIP is widely implemented on many platforms and is currently in use with UCDNet.
  • Hosts using only Proxy ARP for connections will work with Network 21 without changes.
• Implications / Mitigations

  ImplicationMitigation1Departments will need to exercise tighter control on their address space to stay within the size of their subnet. Departments will need to keep complete and accurate records of address assignments. IT will provide templates for tracking relevant information. IT will also provide known information about the configuration of the department's current subnet. 2Variable length subnet masks make host configuration more difficult and not uniform across campus.IT will work closely with TSCs and network administrators to reduce the difficulties often associated with variable length subnet masks so this is not a widespread issue.3RIP will not pass variable length subnet mask information to departmental hosts. Most departments do not need complete routing information for the entire campus.

• Other Options Considered / Implications

  Other OptionImplication1Continue to use 250 hosts per subnet.Campus networking address space would be exhausted quickly.2Use a small subnet (16 hosts) and assign multiple subnets per department.Although this would make efficient use of address space and provide a common subnet mask across the entire address space, the default router address would be different for each subnet and would present greater configuration difficulties.

1.9: Implementing Virtual LANs (VLANs)

• Abstract

  Network 21 technology will make it possible to join servers on remote networks virtually as though they were on the same subnet. Virtual LANs (VLANs) provide system and network administration conveniences but may be subject to certain conditions before implementation. Please refer to VLAN Information for a general discussion of VLANs.

• Policy

  If a department is spread among several sites, the department shall gather information about clients and servers at all remote sites to be included within a single VLAN. Since this information is vital to the planning process, this information must be gathered before the department proceeds with cutover in any of its sites.

• Assumptions
  • VLANs will be used to consolidate remote departmental networks making shared resources easier to provide.
  • Use of VLANs is subject to technical limitations, such as the number of VLANs a machine may belong to (one).
• Benefits
  • Network 21 VLANs will simplify IP configuration among the remote sites participating in the VLAN.
  • Hosts may move to any of the VLAN sites without configuration changes, making employee relocations easier for the department.
• Implications / Mitigations

  ImplicationMitigation1Participation in multiple VLANs is not possible.

  • Machines may still use IP for resources not included in the departmental VLAN.
  • Future technology may relieve the single VLAN restriction.

  2Departments will not be able to fully utilize benefits gained with VLANs until all the department's remotes sites have been cutover to the Network 21 infrastructure.Departments may start implementing limited VLANs before all remote sites have been transitioned. Phasing into VLANs provides a smoother transition and prevents a large amount of work to implement VLANs after the last departmental site is cut over.

1.10: Using Network Layer Protocols with Network 21

• Abstract

  In UCDNet the standard network layer protocol to use when communicating with machines in other buildings is the Internet Protocol (IP). With the use of VLANs, departments may use non-IP protocols to communicate with machines that are in different buildings and on the same VLAN.

• Policy

  Departments will be able to use Network Layer Protocols that were formerly unroutable through UCDNet under the following conditions:

  • Use of non-IP protocols shall be limited to the departmental VLAN.
  • The Internet Protocol will be required to communicate with machines outside the VLAN.

• Assumptions
  • Local departmental computer resources will tend to use proprietary protocols by default.
  • Configuring proprietary systems to use IP may result in additional system overhead that will reduce communications efficiency.
• Benefits
  • Protocols that are not easily routed through an IP internetwork may now be utilized within the VLAN environment without modification.
  • Using proprietary protocols in the VLAN may provide excellent security by guaranteeing access denial to IP users outside the VLAN.
• Implications / Mitigations

  ImplicationMitigation1If IP is not used as the Network Layer Protocol, users outside the VLAN will not have access to resources.

  • Most modern software is able to do IP encapsulation so access to resources on the VLAN is not normally an issue.
  • In some cases, gateways may be used to translate between IP and the proprietary protocol. An example of an IP gateway is the Shiva FastPath and Cayman GatorBox, which tunnel AppleTalk through an IP network. Gateway devices most likely will not need to be purchased because they are widely in use on UCDNet.

  2Network administration decisions become more complicated with the choice of non-IP Network Layer Protocols. IT Network Specialists may provide assistance when determining which Network Layer Protocol to use on the departmental VLAN.

• Other Options Considered / Implications

  Other OptionImplication1Require IP for all interbuilding communications. This would impose unreasonable constraint and financial burden on departments converting their LANs to Network 21.

1.11: Using the Internet Protocol after the Network 21 Project

• Abstract

  The Internet Protocol remains the most widely-used means for computer internetworking internationally. Even with new advances in technology, Network 21 will still depend upon the Internet Protocol for communications to the Internet and between VLANs.

• Policy

  For inter-VLAN and Internet communications, departments shall use the Internet Protocol (IP).

• Assumptions
  • Departments already connected to UCDNet are familiar with IP support.
• Benefits
  • IP is the most widely-used internetworking protocol in the world.
  • Support for IP already exists on campus.
  • Most communications software is able to use IP.
• Implications / Mitigations

  ImplicationMitigation1IP support may require a gateway for protocol conversion when systems are not able to encapsulate other protocols within IP.

  • Shiva FastPath and Cayman GatorBox are IP gateways in use on campus. These devices encapsulate (or tunnel) AppleTalk across an IP network to another AppleTalk network.
  • All software that is UCDNet compatible (with the exception of DECNet which may be addressed with a VLAN) is already IP capable in one form or another.

1.12: Security Implementation with Network 21

• Abstract

  Network 21 has the potential to provide some limited forms of security. Although these security measures can improve information integrity, they should not be relied upon in cases where data is especially sensitive.

• Policy

  Network 21 may provide the following security measures. These measures should be viewed as added security benefits and should not be solely relied upon where absolute security is a priority.

  • Static Ethernet hardware address per port allows only a particular Ethernet card to use the Network 21 port. This option will not function if a local hub or repeater has been used to expand the number of network devices connected to a single NAM.
  • Ethernet switching provides privacy as unicast frames will not be broadcast to all ports of the local network (assuming departments have not installed hubs or repeaters).
  • Protocol obscurity hides proprietary protocols from the outside world using VLANs within the department. This option is only valid for non-IP protocols used within the department's VLAN.

  If problems arise with a secure port, IT staff will assist troubleshooting the port only with the registered TSC or network administrator for the given network. This will eliminate unauthorized changes in network security.

• Assumptions
  • Departments desiring additional security measures are willing to make the trade-off between security and convenience.
  • Departments requiring enhanced security will investigate other security measures such as end-to-end data encryption.
• Benefits
  • These security options are fairly easy to implement and maintain and should be discussed with IT personnel during the cutover planning process.
• Implications / Mitigations

  ImplicationMitigation1

  • Security hinders the convenience and flexibility of the network.
  • A single Ethernet address restricted to a single port eliminates the possibility of computer mobility within the department.

  • IT will assist the departmental TSC or network administrator in troubleshooting a secure port.
  • IT turnaround time for reconfiguring a secure port shall be minimized to reduce the inconvenience associated with reconfigurations on secure networks.

  2Since IT will troubleshoot secure ports only with the assistance of the TSC, network administrator, or other designated department staff, problems may remain unresolved in their absence.IT will encourage departments to designate alternate contacts to reduce the chances of sustained downtime caused by lack of designated troubleshooting personnel.

Post-Cutover Support Policy (Network 21 Support)

2.1: Network Operations Center

• Abstract

  As a part of the enhanced UCDNet, IT will maintain a Network Operations Center to be the focal point for trouble calls, information, and network monitoring and analysis.

• Policy

  The Network Operations Center (NOC) shall be a 24x7 (24 hours a day, seven days a week) operation to maintain the highest, around-the-clock reliability of Network 21 on a large scale.

  During business hours, the NOC shall provide all services (trouble calls, information, network monitoring and analysis).

  After business hours, the NOC shall provide reduced services such as network trouble calls, network monitoring, non-critical maintenance, and statistics consolidation.

• Assumptions
  • The NOC may be reached by e-mail, phone, world-wide-web, radio, or other commonly used forms of communication.
  • The Network 21 Trouble desk phone number is (530)754-NW21.
• Other Options Considered / Implications

  Other OptionImplication18x5 (eight hours a day, five days a week) NOC supportSupport during business hours will not accommodate usage by network clients.216x5 (sixteen hours a day, five days a week) NOC supportExtending support to midnight will meet the demands of most staff; however, many important automated jobs are run between midnight and 6:00AM. 16x5 does not address weekend network use.

2.2: Service Orders for Network Services

• Abstract

  Service orders (moves, changes, additions) should be submitted to the Network Service Center. Service Order processes are being designed for faster service.

• Policy

  IT shall receive post Network 21 service orders through on-line forms, mail, or in-person delivery.

  Typical service orders shall take no more than two working days to complete, providing infrastructure is available. In cases where infrastructure must be installed and does not require special treatment (boring holes through concrete barriers, etc.), the turnaround time shall not exceed 10 working days.

  Service orders for new installations in areas not covered by Network 21 shall be turned over to a network designer/estimator for review and action. An estimate for installation costs shall be done within 3 working days. After departmental approval, installation shall proceed as quickly as possible, pending receipt of equipment needed for the job (which may involve custom parts).

• Assumptions
  • Much of the service order process may be automated and optimized.
  • The campus community will be willing to submit requests electronically in order to expedite work.
  • Faster turnaround time is contingent upon installation and readiness of Network 21 infrastructure.
• Benefits

  A decrease in time between service order and installation.

• Other Options Considered / Implications

  Other OptionImplication1Leave the current network order process in place.The current turnaround time for network orders is viewed as unacceptable in such a large scale network environment where moves, adds, and changes will be centralized.

2.3: Hours for Post-Network 21 Trouble Response

• Abstract

  IT shall provide rapid technical response to trouble reports during regular business hours.

• Policy

  During regular business hours, and under normal conditions, IT shall provide 2-hour technician response to single trouble reports.

  After-hours response shall normally entail a process of remote triage and, if necessary, a remote port reset by Network Operations Center staff.

  At all times, IT shall respond to large-scale outages (BDF or larger) as soon as possible. After-hours, large-scale support may include technician callback if the problems are not solved with remote diagnostics.

  After-hours departmental level (IDF and NAM) support may be requested by an authorized signature on a time-and-materials basis pending availability of appropriate technical staff.

• Assumptions
  • Single port failures within a department are not usually a critical problem after hours.
  • Large-scale outages have the potential to affect large number of clients and should be attended to as soon as possible.
• Benefits
  • Flexibility and economy of technical staff by guaranteeing after-hours service on a broad level and allowing departments the option of paying a fee for accelerated repair of smaller scale problems.
• Implications / Mitigations

  ImplicationMitigation1Connectivity to the desktop is not guaranteed 24 hours a day.

  • The cost of keeping technicians on site 24 hours a day would be excessive, considering the small number of critical single NAM failures.
  • In the case of an after-hours failure, departmental personnel can often use another workstation.

  2Technical response is not guaranteed if appropriate personnel are not available.Departmental needs span many areas of expertise (wiring, network analysis, network operating systems, computer hardware, etc.) that cannot possibly covered by the single on-call technician. Technicians working outside their specialties may or may not resolve the problem.

Policies Regarding Older UCDNet Backbone Infrastructure

3.1: Older UCDNet Backbone Infrastructure Equipment

• Abstract

  Older UCDNet backbone equipment, including some routers and broadband network equipment, will be redeployed to provide network services to campus sites not currently connected. NOTE: This policy does not apply to department owned equipment such as hubs and repeaters that have typically resided within the IDFs. Departments may choose to redeploy this equipment within their own networks as part of the cutover to Network 21 infrastructure.

• Policy

  In areas covered by Network 21, the backbone infrastructure equipment used to build the original UCDNet will be replaced with Network 21 equipment.

• Assumptions
  • Older UCDNet backbone infrastructure hardware will be available for redeployment shortly after area cutovers and after legacy desktop equipment is removed from the network.
  • Departments may choose to redeploy department-owned equipment within their own networks as part of the cutover to Network 21 infrastructure.
• Benefits

  Recycling older backbone hardware will help provide more economical networking services to portions of the campus that are not currently funded by the Network 21 project.

• Implications / Mitigations

  ImplicationMitigation1 Much of the the investment in the existing infrastructure could be lost. Departments have the option to leverage their investment in equipment used with classic UCDNet infrastructure by redeploying this equipment as part of the cutover to Network 21 infrastructure.