| |
|
At this point the project goals are defined
in a general way, i.e. “demonstrate end-to-end QoS”,
“demonstrate end-to-end scavenger service”. and
“demonstrate the use of e2e Qos by live Grid applications”
There need to be translated into a more detailed set of
technical goals which are
- very specific
- technically achievable
- well enough defined that the success metrics can be
defined as well as the likely content of the final project
report.
- well enough defined such that the subsequent tasks
can be carried out
|
| |
|
This task encompasses all of the work needed
to define the types of traffic we need to generate and use
during the project, as well as the measurement systems required
to characterise the network, and measure and demonstrate
expected traffic behaviours.
A very important part of this task is to define the measurements
and presentations (e.g. plots) which will be used to demonstrate
that the project technical goals defined in task 1 have
been achieved. I.e. the success metrics.
This task is divided into three parts:
- definition of the load traffic and the software and
hardware is required to produce it;
- definition of the measurement systems required and
specification of the software and hardware is required;
- production of necessary software and scripts for both
of the above.
|
| |
|
This task is (perhaps artificially) separated
from the previous task as it may involve benchmarking, hardware
selection, negotiation of collaboration agreements and purchasing
procedures. |
| |
|
All security implications are considered within
this task. |
| |
|
Define aggregate traffic classes and policies
to be defined. It is envisaged that a set of policies of increasing
complexity are defined to administer the behaviour of the
network and the experimental tests will provide a measure
of their effectiveness, i.e. the defined policies can be measured
against reality. |
| |
|
Much useful work can be done with the routing
equipment in a limited laboratory environment, before connecting
everything to the WAN. This makes sense anyway as the roll-out
of some of the WAN connections appears to be likely to slip.
The work will include:
- Learning how to configure the routers for packet classification
- Configuration of policing
- Learning how to configure chosen PHBs
- Production of IOS templates for wider deployment.
- verification of measurement techniques
- creating and observing QoS behaviour,
- in particular “proof of principle” of project
technical goal
- establising baseline throughput performance.
Tests devised for this purpose could form a useful basis
for the more-complex WAN networks tests whilst also assisting
in establishing a ‘learning curve’ for devising
the router configurations. |
| |
|
The network configuration is defined here.
It is shown as a set of increasingly complex configurations
starting from basic IP connectivity to include elements
of diffserv, MPLS and combinations thereof. In addition,
the number of administrative domains increases from a simple
“core” model to the “man - core -man”
model.
|
| |
|
This task is intended to undertake a systematic
and complete set of demonstrations which allow us to achieve
the first goals of the project.
In fact much of the work will have been done piecemeal
during many of the previous tasks, and it may be in the
end that there is little more to do.
The key point is that the focus of this task is different.
Whereas previously the focus was equipment configuration,
the focus of this task is a final output document.
Experience shows that it is only when writing such a document
that missing measurements become apparent.
The work includes:
- Revision of success metrics. The success metrics have
been defined in task1. These are however likely to need
revising in the light of experience to this point
- Planning of a systematic set of e2e measurements corresponding
to the above
- Carrying out systematic set of measurements.
- Scheduling regular technical presentation sessions (2
weekly ?) to present ongoing results to technical group
for feedback and direction.
- Evolving descriptive document and presentation.
- Preparation of some form of publication.
|
| |
| |
The preceding work (up to task 8) will have
led us to successful demonstration of e2e QoS across multiple
UK domains.
Further work will be required to go from this point to
the point of UKERNA having defined a “Scalable Managed
Bandwidth Service” (i.e. replacement for old ATM pilot)
as required of them by their SLA.
Work is required here to define what this means precisely
over and above e2e Diffserv capability (for example it may
only require providing and supporting nationally the associated
set of traffic classifications, static (or possibly dynamic)
SLAs , and policing mechanisms nationally).
|
| |
| |
This and the following task appear at the
end as they can only rationally follow much of the work
of the previous tasks. However this should not be misunderstood
as diminishing their importance. Demonstration of “QoS”
across international boundaries was a stated objective of
the original proposal.
The generic task is
- Identify end points in EU and US, in particular identify
named people at the far ends who agree to be contacts
for this project.
- Identify which services can be extended, and identify
all relevant connectivity issues.
- Negotiate and lobby as appropriate to obtain in principle
agreement for e2e connectivity where feasible.
- Convert 3 into concrete series of steps to arrive at
point where a set of tests (which should follow task 8)
can be undertaken
- Make measurement. It is suggested to programme
Of course the above does not contain much specific detail
by the very nature of the task. In all probability most
of the work is political. The details can only be apparent
after following various lines of opportunity.
For example (but nothing below represents any firm agreement
as yet)
- Negotiations are already well underway with DANTE as
part of the DataGRID/DataTAG projects for piloting of
IP premium. On the face of it this matches well with the
UKERNA programme of IP premium deployment. These lines
must be followed to establish a coherent plan for connectivity
to CERN and other European sites. Very likely INFN (IT)
will be a key end point also.
- We have already spoken with Les Cottrell at SLAC. Les
has already performed QBSS tests which were presented
at the GGF in Toronto. Les is very interested to extend
these QBSS tests to MB-NG.
- We have contacted Ian Foster, and have good contact
with STARLIGHT and IWIRE in Chicago. Discussion of some
scheme to extend QoS through to sites on IWIRE may be
the next step.
|
| |
| |
The project proposal includes an objective
to take network services up to the applications layer in
some form. This task is to investigate and pilot suitable
mechanisms for doing this.
The task is open ended, and may extend to any level of
complexity (including dynamic allocation) as time and resources
permit. It is not envisaged that applications would be able
to set code points directly, rather that some indirection
is provided where an application requests a service from
some broker which itself has the right to negotiate with
the netowrk.
It could also start from and directions to be proposed
by collaboration members, and J.Crowcroft has already made
suggestions in this direction.
It could use the GARA work which has been done in the Globus
context (Volker Sandler – well known to CISCO). As
understood so far this provides access via APIs to resource
reservation, and was based around CISCO routers. |
| |
| |
This task will define the high performance
programme for MB-NG. It will specify the tests and experiments
required to achieve high performance high throughput networking.
It will also define suitable demonstrations of sustained
data transfers over the network using various transport
and application protocols.
This task will use the deliverables from Task 2 that define
the test and background traffic patterns and generation
methods, the corresponding specification work in this task
is only to select those tests/measurements that are applicable.
This task will also be driven by Task 7, in that the network
configurations and operational policies for the high performance
tests and demonstrations will be selected from those specified
by Task 7. It is expected at all tests will be operated
on a set of increasingly complex configurations and QoS
policies starting from basic IP connectivity to include
elements of diffserv, MPLS and combinations thereof.
This task is divided into several parts:
- Definition of the work required to investigate TCP/IP
operation on high bandwidth links.
- Definition of the work required to investigate non-TCP/IP
protocols.
- Specification of the data transfer tools to be tested.
- Organisation of the high performance demonstrations.
|
