Contents:

1) IGMP Snooping by HP and Foundry Switches
2) Get NDO Approval Before Equipment Purchase
3) Change In Traceroute Outputs Due To Change In UIUCnet Backbone
4) Networking Basics Class on 5/15 in L532 DCL
5) NAS Workshop on Drive Imaging
6) Staff Changes: Alicia Cross is the new Host Manager

1. IGMP Snooping by HP and Foundry Switches

Cisco 4000's, and 5000's w/o a Supervisor III module and a NetFlow Feature Card, as well as most other Cisco switches, don't do IGMP snooping. The HP and Foundry switches do. The implication of this for users is that they will experience a severe slowdown of traffic on their entire network if they use multicast protocols to broadcast data. Applications that use multicast data include Ghost, Imagecast, and multimedia players such as the Real Media Player. The reason why the network slows down is that switches that don't do IGMP snooping will transmit the multicast data across all network segments, even to nodes that have no interest in receiving such data. The nodes are then unable to transmit new packets onto the shared media for the period of time that the multicast packet is flooded. To learn more visit http://www.globecom.net/ietf/draft/draft-ietf-idmr-snoop-00.html.

2. Get NDO Approval Before Equipment Purchase

Please remember that network equipment MUST be approved by the NDO before beginning a purchase. The proper procedure is for the respective network's administrator to contact their designer. While P.O.'s do get caught, this is not the approval procedure. Failure to comply with the official approval procedure can result in loss of free NAS and Lan maintenance services.

3. Change In Traceroute Outputs Due To Change In UIUCnet Backbone

A fairly major configuration change was performed on the UIUCnet backbone recently, to address some performance issues with the core layer 3 switches. Instead of one sweeping backbone network, there are now several of them in a point-to-point mesh configuration. This allows us much greater control at the IP level, rather than the Ethernet level.

Since the core is now comprised of several distinct IP subnets instead of one, a side effect of this change is that each core device will show up in a 'traceroute' from your network to other locations on or off campus.

For instance, where you may have seen the following prior to the change:
Tracing the route to IP node www.uic.edu from 1 to 30 hops

1    <1 ms   <1 ms   <1 ms uiuc-example-net.gw.uiuc.edu [130.126.2.1]
2    <1 ms   <1 ms   <1 ms exit1.gw.uiuc.edu [128.174.1.137]
3     4 ms    5 ms    8 ms t-dmz.gw.uiuc.edu [128.174.0.209]
4     5 ms    5 ms    5 ms uic-mren.gw.uic.edu [192.17.8.39]
5     7 ms    7 ms    6 ms bgrc-cat5500-2-atm.gw.uic.edu [128.248.120.232]
6    11 ms    10 ms    8 ms cedar.cc.uic.edu [128.248.155.115]

You might see this now instead:
Tracing the route to IP node www.uic.edu from 1 to 30 hops

1    <1 ms   <1 ms   <1 ms uiuc-example-net.gw.uiuc.edu [130.126.2.1]
2    <1 ms   <1 ms   <1 ms core2-1.gw.uiuc.edu [128.174.1.154]
3    <1 ms   <1 ms   <1 ms core4-1.gw.uiuc.edu [128.174.1.142]
4    <1 ms   <1 ms   <1 ms exit1.gw.uiuc.edu [128.174.1.137]
5     7 ms    7 ms    5 ms t-dmz.gw.uiuc.edu [128.174.0.209]
6     8 ms    7 ms   10 ms uic-mren.gw.uic.edu [192.17.8.39]
7    11 ms   10 ms   13 ms bgrc-cat5500-2-atm.gw.uic.edu [128.248.120.232]
8    11 ms   11 ms    9 ms cedar.cc.uic.edu [128.248.155.115]

While at first glance this appears to be two more "hops" and thus a longer path, it's important to understand that the topology of the core has not changed, and just as many devices are being traversed now as prior to the change. The difference is that these are now IP hops and thus visible to traceroute, whereas before they were layer 2 Ethernet hops and thus not visible. Since the gigabit Ethernet core switches perform both layer 2 and layer 3 switching in hardware for most functions, no performance has been lost, and no latency or delay introduced because of this change.

Note: Due to the rapid growth of the Internet, and that fact that many other networks are making similar changes, the total number of hops required to reach many destinations is increasing. Some older TCP/IP implementations have a low default maximum number of hops which can make some sites unreachable. For example, Windows 95 sets the default time to live (TTL) setting to 32 hops. Microsoft documents this in "Windows TCP/IP Registry Entries" at: http://support.microsoft.com/support/kb/articles/Q158/4/74.ASP; however, many utilities and registry files have been posted to make it easier to increase the TTL setting in Windows 95 than can be done with their documentation. See: http://www.winguides.com/registry/display.php/31/
for an example.

4. Networking Basics Class on 5/15 in L532 DCL.

The third NAS training session for Spring 2001 will once again cover Networking Basics. The class will be held in room L532 DCL, on Tuesday, May 15th, from 2:00pm to 5:00pm.

5. NAS Workshop on Multicast and Drive Imaging on 5/25 in 2240 DCL

NAS will host a workshop on Drive Imaging. The workshop will include a short talk by Charley Kline on multicast switch issues, and discussions on Ghost, Imagecast, and other related issues. The workshop will be held on Friday, May 25th, from 3:00pm to 5:00pm in 2240 DCL. Note: Imagecast version 4.5 was recently posted to the CCSO Site Licensed Software Distribution page at: https://www-s.uiuc.edu/sitelic/dist.

6. Staff Changes: Alicia Cross is the new Host Manager

Alicia Cross is our new DNS manager. Alicia (a-lee'sa) is new to the University but not new to the community. Her background is in accounting and computer science, a great combination for a DNS manager. She is now answering hostmgr mail; please bear with her as she learns the ropes. As far as possible, make DNS changes via the URL http://www.uiuc.edu/ph/www/hostmgr. DNS Documentation is also available, for further reference, at http://www-commeng.cso.uiuc.edu/dns/docs/. If the Host Registration website fails to make the DNS changes, contact Alicia at hostmgr@uiuc.edu.