Transcript NLANR Passive Measurement and Analysis

Passive Infrastructure:
Opportunities and Challenges
Jörg Micheel
NLANR/MNA
San Diego Supercomputer Center / UCSD
<[email protected]>
March 30th, 2005
PAM2005, Boston, MA
1
define: infrastructure
• “The term refers to the often unseen network of services provided
for the community. Infrastructure that territorial councils provide
includes roads, water, wastewater and stormwater pipes, treatment
facilities and parks.”
• “The framework of interdependent networks and systems comprising
identifiable industries, institutions (including people and procedures),
and distribution capabilities that provide a reliable flow of products
and services essential to the defense and economic security of the
United States, the smooth functioning of governments at all levels,
and society as a whole.”
• “Goods and services which, while in themselves not normally
directly productive, are regarded as essential to the functioning of a
developing economy. Infrastructure includes such things as power,
transport, roads, housing, education, health and other social
services.”
March 30th, 2005
PAM2005, Boston, MA
2
So infrastructure is …
•
•
•
•
•
Network of systems, policies, structures
In itself non-productive
Serving the community
Providing a reliable flow of products
Essential to a developing economy and
the society as a whole
• A cheap way of getting a lot done!
March 30th, 2005
PAM2005, Boston, MA
3
But infrastructure is also …
• A large scale investment, often under budget pressure
• Which requires an organization and specialized,
experienced staff for development and maintenance
• Inertia due to size: dynamics often in conflict with a
rapidly changing environment
• Common denominator – serve the nine tenth of the
iceberg that is under water
• Won’t necessarily be an exact fit to your particular needs
- inconvenience factor for the individual is unavoidable
• Corollary: don’t judge the impact of an infrastructure by
the tip of the iceberg that is sticking out
March 30th, 2005
PAM2005, Boston, MA
4
This workshop is an opportunity
• To define the infrastructures that you think
will best address the needs of the network
and research communities
• To search for common denominators:
types, scale, support, interaction, goals,
targets, timelines
• To ensure the resources will be there if
and when you need them for your future
research
March 30th, 2005
PAM2005, Boston, MA
5
PMA infrastructure
• Serve the network research community
with readily available data for analysis of
the evolving nature of the Internet
• Support research networks with needed
performance data on HPC network links
• Provide for a historical track record that
can be studied retrospectively
March 30th, 2005
PAM2005, Boston, MA
6
PMA strategic planning (2001)
• View IP header trace files as “movie clips” or snapshots
from real world of packet level communications
• Preserve base line 8x90 seconds historic sampling
• Diversify locations: access, aggregate, backbone,
wireless, research vs. commodity Internet
• Diversify visible window: hours, days, weeks, 24x7
• Provide timing details at single point (router
instrumentations) and multiple points (GPS/CDMA)
• Push real time analysis to provide instantaneous richer
views of packet network behaviour
March 30th, 2005
PAM2005, Boston, MA
7
PMA strategic planning (2005)
• This page intentionally left blank …
March 30th, 2005
PAM2005, Boston, MA
8
PMA strategic planning (2005)
• Aligned with the research agenda of the
community
• Coarse grained, hitting major targets on
the way
• Things that cannot be achieved by
individual research groups
• Things that take time to accomplish, things
that have lasting time span
March 30th, 2005
PAM2005, Boston, MA
9
Opportunities in conducting
Passive Measurement & Analysis
• Most precise way of reflecting the inner
workings of packet level communications
• Non-interference or second-guessing
• Direct attachment of CPU to network data
link allows for new models of network
comm’s to be computed in real time
• Potential to launch new kind of network
management framework and services
March 30th, 2005
PAM2005, Boston, MA
10
Challenges in conducting
Passive Measurement & Analysis
•
•
•
•
•
•
•
Costly to begin with (at present)
Technologically demanding (OCxMON)
Networks and links under constant change
Difficult to provide dense coverage
Research questions are moving target
Security of infrastructure is ongoing concern
Data sensitivity gets in the way of certain
analysis works
March 30th, 2005
PAM2005, Boston, MA
11
Public versus Private
General purpose infrastructure
Multiple smaller facilities
Serves the need of the community
at large (all that’s under water)
Serves a specific purpose only (tip
of the iceberg)
Sacrifices the needs of individuals
and small groups
Addresses the needs of small
groups
Cost efficient for the bulk of all
problems
Expense targeted to selected
problems only
Very little flexibility, difficult to
adapt apart from strategic goals
Very flexible for particular needs,
useless for others
Data easy to find for others
No data to correlate
Your favourite soap box …
… here.
March 30th, 2005
PAM2005, Boston, MA
12
Infrastructure drives Technology
Technology
Infrastructure
OC192MON
Router
instrumentation,
backbone
instrumentation
March 30th, 2005
PAM2005, Boston, MA
Research and
Operations
End-to-end
facilities
available to all
users
13
Questions to be answered
• Purpose of measurements and
infrastructure – who is the audience and
what are the goals
• Public and/or (multiple) private
infrastructure(s)
• Level of engagement (size)
• Data collection and analysis schedule
• Feedback and interaction between
infrastructure and community
March 30th, 2005
PAM2005, Boston, MA
14