Slide 1: preliminary

preliminary 
measurement specification 
for Internet routers

15 january 1999

ISMA panel

            Daniel McRobb (CAIDA)
            Greg Miller (MCI)
            Stan Hanks (Enron)
            Randy Bush (Verio)   [remote!]

moderator: kc (caida)

Slide 2: outline

outline


purpose of specification
outline of initial strawman document 
open issues
discuss discuss
conclusions, next steps

Slide 3: purpose: tell vendors what to measure and how

purpose: tell vendors what to measure and how


needs

capacity planning
peering engineering
SLA verification
tracking topology and routing changes
ATM/cell level errors

Slide 4: initial strawman document

initial strawman document 


collecting application-specific data
      e.g. reconstruct full HTTP headers 
      from packet level data

flow aggregation parameters
      granularity
      timeout
      how much configurability

queue lengths

focus on IP level

Slide 5

what really needed? 


operationally
basic traffic characterization (how much)
AS matrices
traffic import/export
routing/address space coverage
security/vulnerability protection 	
(DOS attack traces, filtering)

more researchy
interarrival time behavior
protocol-relevant (dups, packet sizes)

Slide 6: open issues

open issues


how important is ATM/cell level stuff?
how to report missing flows/stats?
impact on router performance (any?)
sampling
SLAs - passive indicators? delay measurements?
timestamp deltas in sniffed packets
correlate w data from active tools, snmp
metrics of jitter
how will MPLS change all this?
how will IPSEC change all this?

Slide 7: microscopic ("tcpdump i/f card") sniffing would allow

microscopic ("tcpdump i/f card") sniffing would allow


interarrival times 
packet run lengths
interarrival time distributions

protocol-relevant 
TCP: retransmissions/dup acks
packet size distributions

security related 
DOS attack traces
on-card kernel packet filtering, a la bpf

Slide 8

discuss discuss
 
 

             (k/elves take notes)

Slide 9: conclusions

conclusions


     will write into revised spec
     and part of ISMA final report 



comments to 
kc@caida.org