[Fwd: [e2e] Network optimization for bandwidth provisioning]

[Sireen Habib Malik]

Hi Jing,

Ref. your first question.....

>> Could we accept that assumption? To my understanding , if link capacity is overprovisioned by 
>> 15%, routing paths>> computed by SPF may deviate from that of original network.


If SPF is taking "number of hops" as the shortest path then there will not be any route changes as long as the topology is intact. This maybe realized by setting the same OSPF weight on all interfaces. However, if the network is using dynamic OSPF weight setting which happens to be some function of capacity then flows may re-route.

I have not read the paper but the mentioned assumption appears to me as a pre-requisitve for such a study. I mean,  for such a performance comparison, all the variables except capacity need to be kept the same. If for one reason or another flows re-route because of the change in capacity then there would be no ground for such a performance comparison.  

Example : one flow is going from source node A to destination node D. It follows a path A-B-C-X-D (B, C and X are the intermediate nodes). Say this flow faces a 10msec e2e delay. Now increase the link capacities. If the network has dynamic OSPF weight setting, it will react and re-route the flows. Say, our example flow now follows the path A-X-D. Assume, resultant dealy is 5msec - an improvement. Question is, what brought about this improvement? Increased capacity or the online-optimized route?  So,  to measure the impact of over-provisioned capacity  the flow should keep A-B-C-X-D path i.e. routing should not change.

Sireen Malik,

Technical University of Hamburg-Harburg
Digital Communication Networks
FSP 4-06
Denickestr. 17  
21073 Hamburg, Deutschland

Tel: (040) 42-878-3351 Fax: (040) 42-878-2941 E-Mail: s.malik at tu-harburg.de


The information contained in this e-mail is intended for the recipient
or entity to whom it is addressed. It may contain confidential
information if you are not the intended recipient, you must not copy,
distribute or take any act in reliance on it. If you have received this
e-mail in error, please notify the sender immediately and delete from
your system.





Jing Shen wrote:


>> Thank you very much , and sorry for my poor question.
>>
>> Fraleigh's paper , " Provisioning IP Backbone Networks to Support
>> Latency Sensitive Traffic", stated that its analyis is based on
>> assumption that network topology and routing path remains in its
>> original shape after link bandwidth is overprovisioned.
>>
>> My first question is: Could we accept that assumption? To my
>> understanding , if link capacity is overprovisioned by 15%, routing
>> paths computed by SPF may deviate from that of original network.
>>
>> In order to verify its conclusion, my second question focuses on the
>> bandwidth provisioning state of current backbone network and its
>> edge-to-edge delay statics.
>>
>> Thanks again.
>>
>> Best regards
>>
>> */Tao Ye <tye at sprintlabs.com>/* wrote:
>>
>>     I don't understand the first question, but for the second
>>     question, you
>>     can always look at Sprint's ipmon page for Sprint's backbone link
>>     utilization.
>>
>>     For example, BW, utilization and other statistics of a number of
>>     backbone
>>     links are available:
>>
>>     http://ipmon.sprintlabs.com/packstat
>>
>>     Delay statistics are available for several dates as well, such as
>>     2002/11/21.
>>
>>     http://ipmon.sprintlabs.com/delaystat
>>
>>     On Sun, 16 Mar 2003, Amgad Zeitoun wrote:
>>
>>     >
>>     >
>>     >
>>     > --__--__--
>>     >
>>     > Message: 1
>>     > Date: Sun, 16 Mar 2003 16:39:12 +0800 (CST)
>>     > From: =?gb2312?q?Jing=20Shen?=
>>     > To: end2end-interest at postel.org
>>     > Subject: [e2e] Network optimization for bandwidth provisioning
>>     >
>>     > Hi,
>>     >
>>     > I just read a paper published on InfoComm'03 named "Achieving
>>     ! > Near-Optimal Traffic Engineering Solutions for Current OSPF/IS-IS
>>     > Networks", it is stated that
>>     >
>>     > only 15% BW overprovisioning is needed to guarantee as low as
>>     3ms e2e
>>     > delay requirement.
>>     >
>>     > My question is:
>>     >
>>     > 1) a network optimization result with both network topology,
>>     routing
>>     > path preassumed reliable? How could we consider routing path
>>     changes
>>     > when link BW is overprovisioned?
>>     >
>>     > 2) what's the state of link BW in current backbone network ? is
>>     there
>>     > more than 15% BW idled?
>>     >
>>     > Thanks in advance.
>>     >
>>     >
>>     >
>>     >
>>     >
>>     >
>>     > Jing Shen
>>     >
>>     > State Key Lab of CAD&CG
>>     > ZheJiang University(YuQuan)
>>     > HangZhou, ZheJiang Province 310027
>>     > P.R.China
>>     >
>>     >
>>     >
>>     >
>>     > --__--__--
>>     >
>>     > _
>>     >
>>
>>     -- 
>>     ciao,
>>
>>     Tao
>>
>>
>>
>> Jing Shen
>>
>> S! tate Key Lab of CAD&CG
>> ZheJiang University(YuQuan)
>> HangZhou, ZheJiang Province 310027
>> P.R.China
>>
>>
>> ------------------------------------------------------------------------
>> *Do You Yahoo!?*
>> "¸ü¶à¾ªÏ²£¬Í¬Ñù¾«²Ê£¬NetVista A30 ÈÈÂô"
>> <http://ad.cn.doubleclick.net/clk;5313999;7930402;p?http://www.ibm.com/cn/promotion/pc/netvista_a30/index.shtml>
>  
>