IEPM

SLAC QBSS Testbed

SLAC Home Page
QBSS Measurements | QBSS Testbed | Evaluation of routes for QBSS tests
Differentiated Services (DiffServ) are cited in terms of increasing performance by labelling packets with "expedited forwarding (forward me first)" and "assured forwarding (drop me last)". However, IEPM is leading the way in making use of the QBone Scavenger Service (qbss), where DiffServ is used to provide less than best effort. Huge data transfer will be trickled across the network and will back down whenever other applications need the bandwidth.

Data trasfer tools

There are two tools we can now use to transport data over the Wide Area Network with packets marked for QBone Scavanger Services:

bbftp

bbftp is a software designed to quickly transfer files accross a wide area network. It has been written for the babar experiment in order to transfer big files (more than 2 GigaBytes). To perform our test we have introduced one extra-option, that allows for packets to be marked for QBSS.

bbcp

bbcp is a tool developed at SLAC by Andy Hanushevsky to "securily and quickly" copy data from source to destination. The application has been officially patched to allow packets' marking for QBSS; this means setting the TOS (type of service) of the packets to be 0x20.
The user can mark packets for QBSS using one of the application options:
bbcp -q tos-value .....

QBSS test bed

Cisco 7200

The following is the layout of the first QBSS test bed we created at SLAC.
We have two Cisco 7200 back to back, connected via a 10/full ethernet connection, this allows us to easily congest this link and to measure the perfomances of the variuos traffic types.

Router configuration

The configuration of the router consists of: This is accomplished in the router with the following commands:

class-map match-all qbss
match ip dscp 8
class-map match-all hi-priority
match ip dscp 10


policy-map qos-policy
class qbss
bandwidth percent 1
random-detect
class hi-priority
bandwidth percent 30
random-detect
class class-default
bandwidth percent 69
random-detect


interface ...
service-policy output qos-policy

QBSS test bed

Cisco 6509 SUP2/MSFC2

The following is the layout of the second QBSS test bed that we built using Cisco 6509 with SUP2/MSFC2 configuration:

The two switches are connected together using a 100 mbps line and they connect to the rest of SLAC and to the test network with a Gigabit line.

The MSFC2s have only virtual interfaces, the routing is performed by the PFC2 (Policy Feature Cards) on the switch, and traffic policing can be achieved configuring QoS (Quality of Service) on the switch ports.
Our bottleneck is a 100mbps line, of type 2q2t: this means we can define two queues, with two threshold each. The thresholds indicate the level of congestion that triggers drop of traffic. Traffic with different DSCP (Differentiated Services Code Point) can be assigned to the any of the two queues and to one of the two threshold.
Given there are only two queues available, we have assigned the QBSS to a queue by itself, and the Priority and the Best Effort traffic to the second queue. The QBSS queue gets assigned a 5% of the total bandwidth available and the rest is assigned to the second queue.
At this point we can use policing to limit the Priority traffic to get 30% of the bandwidth.

The configuration we have implemented is the following:

set qos enable(enable Qos)
set qos dcsp-cos-map 8:1(map QBSS to COS 1)
set qos dcsp-cos-map 10:7(map priority traffic to Cos 7)
set qos wrr 2q2t 255 5(define the bandwidth for each queue)
set qos map 2q2t tx 1 2 cos 0(map cos 0 [Best Effort] to queue 1 threshold 2)
set qos map 2q2t tx 2 2 cos 1(map cos 1 [QBSS] to queue 2 threshold 2)
set qos map 2q2t tx 1 2 cos 7(map cos 7 [Priority] to queue 2 threshold 2)
set qos policer aggregate policer1 rate 30000 policed-dscp erate 30000 drop burst 3200(define a policer for limiting priority traffic)
set qos acl ip acl1 trust-dscp aggregate policer1 ip any any dscp-field 10(define ACL to select priority traffic and police it with the previusly defined policer)

Revised September 21, 2001.
URL: http://www-iepm.slac.stanford.edu/monitoring/qbss/qbss.html
Comments to iepm-l@slac.stanford.edu