Title: Meeting on ATLAS Remote Farms. Copenhagen 11 May 2004
1Networking for ATLAS Remote Farms
Richard Hughes-Jones The University of Manchester
DataGrid WP7 Dante Tests on the GÉANT
Core End-2-End Measurements from the 4th Year
VLBI Project at Manchester New TCP stacks the
effect on throughput Some Simple Network Tests
CERN-Manchester
2DataGrid WP7 Dante Tests on the GÉANT Core
- Set-up
- Supermicro PC in
- London GEANT PoP
- Amsterdam GEANT PoP
- Smartbits in
- London GEANT PoP
- Frankfurt GEANT PoP
- Long link UK-SE-DE2-IT-CH-FR-BE-NL
- Short Link UK-FR-BE-NL
3Tests GÉANT Core UDP throughput
- UDP Throughput
- London-Amsterdam
- Available BW to packet on wire
- Then 1/t
- Wire rate 998 Mbit/s for packets gt 1400 bytes
- Packet Loss
- None for large packets
- Dips in BW lined to packet loss
- SysKonnect NIC int. per packet
- CPU load important
4Tests GÉANT Core Packet re-ordering
- Effect of Packet size
- London-Amsterdam
- Packets at 10 µs line speed
- 10,000 sent
- Packet Loss 0.1
- Re-order Distribution
5Tests GÉANT Core Packet re-ordering
- Effect of LBE background
- Amsterdam-London
- BE Test flow
- Packets at 10 µs line speed
- 10,000 sent
- Packet Loss 0.1
- Re-order Distributions
6Tests GÉANT Core Packet Jitter
- Amsterdam-London
- BE Test flow
- Packet spacing 80 µs
- BE Test flow Background 60 BE 1.4Gbit 40
LBE 780Mbit
- IPPremium Test flow Background
7Tests GÉANT Core 1-way Delay
- Amsterdam-London
- IPPremium Test flow
- Packet spacing 80 µs
- BE Test flow Background 60 BE 1.4Gbit 40
LBE 780Mbit
- BE Test flow Background 60 BE 1.4Gbit 40
LBE 780Mbit
8VLBI Project Test Topology
SURFnet
Manchester
JIVE Dwingaloo
Jodrell
SuperJANET4
Adam Mathews Steve OToole Univ of Manchester
9VLBI Project Throughput
- Manchester to Dwingeloo
- 2.0G Hz Xeon 1.2 GHz PIII
- Re-ordering vs Offered Load
10VLBI Project Jitter 1-way Delay
- 1472 byte Packets man -gt JIVE
- FWHM 22 µs (B2B 3 µs )
11VLBI Project Packet Loss Long Range Effects?
- Aggregated Variance Method
- Divide time series length N into blocks of size m
- Calc mean of each section Xm(k) k 1 N/m
- Calc variance VXm of these Xm(k)
- Vary m size of the blocks
- Plot on log-log fit slope ß
- Hurst parameter H ß 2H -2
- Measureß -0.355 which gives H 0.822
- H 1 no long range dependence
12Traffic Flows
- Manchester NetNorthWest - SuperJANET Access
links - Two 1 Gbit/s
- Access linksSJ4 to GÉANT GÉANT to
SurfNet
13High Performance TCP DataTAG
- Different TCP stacks tested on the DataTAG
Network - 128 ms round trip time
- Drop 1 in 106
- High-Speed
- Rapid recovery
- Scalable
- Very fast recovery
- Standard
- Recovery would take 10 mins
14High Performance TCP MB-NG
- Drop 1 in 25,000
- Rtt 6.2 ms
- Recover in 1.6 s
- Standard HighSpeed Scalable
15Some Network Tests
16Lab Test TCP Request-Response Histograms
- PC router PC
- BE Test flow
- Request spacing 0 µs
17Man-CERN TCP Request-Response Latency
- DataTAG PC backup link
- BE Tests
- Request spacing 0 µs
- Win size 2.5Mbytes
- Compare with UDP latency
- Large differences
- Rtt of 20 msdelaybw 2.5 Mbytes
- 1Mbyte data 690 pkts ? interesting bursts !
18Man-CERN UDP Throughput Packet Loss
- DataTAG PC backup link
- BE Tests
- Throughput
- Packet loss
19Traffic Flows
- Manchester NetNorthWest - SuperJANET Access
links - Link to PC in M/c
- Access links
- 1 GE Man to NNW
- Total Man to NNW
- NNW to SuperJANET4
20(No Transcript)
21VLBI Project Packet Loss Is it Poisson?
- Divide time series of packets into 1000 slices of
50 packets - Total lost packets 1410
- Average number / slice 1.4
- Calc Poisson Probability P(n, µ) µ n e -µ
n! -
- Curves close but not exact
- Could be more than 1 process
22Traffic QoS Classes on GÉANT Backbone
Max Throughput on 2.5 G PoS
- Normal Traffic
- Radio Astronomy Data
- Less Than Best Effort
- 2.0 Gbit/s
- Normal Traffic
- Less Than Best Effort
- 2.0 Gbit/s
- Normal Traffic
- Radio Astronomy Data
- 500 Mbit/s
23Some Measurements made during ER2002