NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
Tonight, at around 6 PM, we should be back to "normal", whatever that is, as our web caches refill up with banner ads. :-) Fail back to me then. Thanks, and sorry. Things are rapidly improving even as I type this. ******************************************************************************* Gilbert Sebenste ******** (My opinions only!) ****** Staff Meteorologist, Northern Illinois University **** E-mail: gilbert@xxxxxxx *** web: http://weather.admin.niu.edu ** Work phone: 815-753-5492 * ******************************************************************************* ---------- Forwarded message ---------- > From: NIU Network Engineering ** High Priority ** The primary degradation in performance that we experienced Tuesday has been mitigated to some degree. ITS modified some parameters on the packet shaper and this modification will take an extended period of time to have a full affect. The shaper re-learns packet handling based on the new parameters and the dynamic traffic. We also "flushed" the local cache engine to ensure that locally cached content would not have error prone content due to the traffic congestion problem. This resulted in a self-imposed performance hit as the cache engine re-populates content. Checks of bandwidth utilization as relates to certain classes of data indicates an overall trend of higher traffic which is pushing the limit of our current bandwidth. High error rates are not present. There is no indication of any equipment failure. There are minor denial of service events occurring - but these occur almost all the time and we deal with them if we can get a handle on a consistent source. We do have fairly large access control lists in place on the edge routers to block some of these sources. Unfortunately, there is a balancing point where the complexity of the access control lists and power required to process packet info against them - outweighs the addition of more control information. Also, by the time the culprit traffic gets to NIU's control list screening, the bandwidth has already been used to get to us---it competes for incoming traffic bandwidth. So, in summary, we have done all we can at this point to address the most recent performance issue. Performance appears to be improved today. We continue to monitor and tune packet shaping efforts and watch for security related issues. J R Fatz Director, Enterprise Systems Support
ldm-users
archives: