NEWS:  All Missouri State University Subnets to be Reconfigured.

Missouri State University is reorganizing its subnets into VLANs (Virtual LANs) with more address space for client stations per VLAN.  Changes for DHCP and BOOTP users will be transparent upon reboot.

All users on all three campuses will ultimately be affected.

June 4th, 1998

Missouri State University Networking Administration will be changing all network addresses in the process of reorganizing and upgrading the building network infrastructure over the next several months.   All client workstations will have to be changed, but fortunately, only a reboot (or possibly a DHCP RELEASE and RENEW) will be required for all standard DHCP and BOOTP users.

Users with static IP addresses and users with DHCP reservations must reapply for a new IP address (including all Computer Services users and all printers).  We will attempt to contact all users in these groups, but until recently, complete records of static and especially of reservation users were not kept.  All users who obtained these addresses from Networking were notified upon acquisition that these changes would occur and that they should contact Networking should their static or reserved address should stop working.  User Support workers and distributed Support Specialists should be on the lookout for this issue when dealing with a previously working machine.

Conversion time schedules, where available, and proposed VLAN numbers can be found on the VLAN Migration Planning page.  This document is a working document that will change as plans evolve, so be sure to check it often if you wish to follow the migration.

When you convert my building, what do I do?  How long will I be down?

If you do not plan on working the day of the conversion, please power your computer off prior to leaving.  If you are working on the network the day your conversion is to take place, be sure to save your work before the planned shutdown.  We will time the actual shutdown for 10:00 a.m. for the first two conversions (Ellis Hall and Craig Hall).   Check the time information on the VLAN Migration Planning page for the conversion time of your building.  About 10 minutes before the time listed, you should save all of your work and shut your computer down.

The network segment will then be shutdown and we will activate new DHCP and BOOTP address ranges, continue reprogramming all hubs, and will reconfigure the router interface supporting your building.  If all goes well, you will be able to restart your machine no later than 30 minutes after the shutdown with no further interruptions.  Actual times could be much shorter.

If you are using a DHCP reservation for your computer or printer, you must have reapplied for a reserved address on the new network.  Otherwise your device will not function on the network when restarted.

If you are using a hard-coded static address, you will have to reconfigure your device with the new address information received from Networking.   You can make these changes while we are reconfiguring the network itself.  If you do not reconfigure your machine properly with information obtained from Networking, your device will not function on the network.

Why Change?  Aren't things working just fine now?

The reorganization is being performed for several reasons:

1. The number of DHCP leases on 24 bit networks (i.e. a subnet mask of 255.255.255.0) is too limiting.
2. Current network overlaying techniques and superscopes introduce complexity and performance problems.
3. Combining multiple subnets into a fewer number of switched networks improves performance and flexibility.
4. Grouping network numbers by physical zone facilitates route advertisement reduction.
5. Increased DHCP headroom will facilitate the introduction of open ports for mobile computers.

Tell me more...

Only 250 or so stations can be placed on any given network segment.  This should not much of an issue for shared networks because they normally are not allowed to grow beyond about 200 stations.  This is because collisions climb greatly as the number of stations increase thus degrading overall performance.  Having 500 stations on a shared network, for example, would be crazy.  However, with a partially switched network, such as Carrington Hall, the 321 stations (as of June 4, 1998) are supported without problems.  About half of the hub stacks in each of the wiring closets are connected to a Cisco C1924-EN switch which breaks the collision domains up by hub stack but still allows for communications between physical segments on the same VLAN without going back to the main campus router.  The expansion allows for more stations per VLAN without overlaying the network.  Carrington Hall, for example, is scheduled to have all hub stacks connected to the Cisco switch by October 21st, 1998.

Overlaying a network segment involves telling the router that two or more subnets exist on the same physical network through the same interface.   This allows about 250 more stations per subnet overlay.  For example, Temple Hall is overlaid once and could have 500 stations.  This give us plenty of room to allocate old BOOTP and new DHCP address pools.  However, if a station on the primary subnet wants to communicate with a station on the secondary subnet, it causes traffic to travel to the main campus router in Cheek Hall, and then right back out the same interface to Temple again.  This is obviously very inefficient.

Overlaying also causes additional complexity in the DHCP server responsible for assigning workstations appropriate address information.  Addresses can be allocated from the primary or secondary scopes from an entity called a superscope.   Unfortunately, the Microsoft DHCP server has several unresolved problems with superscopes and patches have yet to be forthcoming.

By combining multiple subnets into a single VLAN, we can take more advantage of layer 2 (ISO network model) switching as opposed to layer 3 routing.   Switching occurs more or less at wire speed which means that the electronic delay incurred when transferring a packet from one wire to another is negligible.  The Cisco switches we use only check the first 64 bits of a packet before forwarding it to the appropriate network segment.  Routers buffer the entire packet in an input queue, examine the packet, decides where it should go, buffers it in an output queue, and then delivers the packet.  This is much slower.  Layer 3 switching has been introduced by several vendors, but Missouri State University currently does not own any equipment which supports it.

A much smaller benefit is that by grouping VLAN addresses (thus IP, IPX, and AppleTalk addresses by Missouri State University conventions), we have the ability to aggregate route announcements for IP.  This would eliminate a small percentage of network maintenance traffic.  Given current OSPF traffic statistics, this improvement would be very small, but an improvement none the less.

Switching will allow us to have separate VLANs for laptop or mobile computers.  In this way, the highly volatile address space supporting the laptops would not compete for addresses out of the normal building pool.  Most buildings, however, do not support VLANs to the port level at this time.  This means that any mobile computers in a building must compete for addresses with the normal building occupants which could exhaust the supply.  The solution, until complete switching is deployed, is to provide a large surplus of addresses so the supply is not consumed.  Doing this requires 23 or 22 bit subnet masks which allow for 512 or 1024 addresses per VLAN.

My brain's not full, tell me more!

For more information, please contact Mark Harsen