P. Vicat-Blanc Primet

P. Vicat-Blanc Primet ₁

Tools & software for flow

processing in Grid

networks

[email protected]

NEGST project

P. Vicat-Blanc Primet ₂

Summary :

our NEGST

Y2007 activity

Context

 GRIDNET-FJ : INRIA associated team  NEGST : JST- CNRS (Naregi-Grid5000)

Current activities

 TCP variants evaluation with GNET10 & Grid5000  L2 Switches behavior investigation

 Packet capture (GNET1) for traffic analysis

 Flow Scheduling & Network resource reservation  MPI evaluations in Grid5000

 Grid5000-Naregi link establishment

Some achievements

 2 Pfldnet2007 papers (AIST & RESO), HSN07, GridNets07, Cluster07 sub.  Collaborative experiments /design RESO & AIST with GNET1 & GNET10  Grid5000-Naregi link is quite ready to use

Tools and software for flow processing in Grid

networks

P. Vicat-Blanc Primet ₄

Introduction

Grid traffic is a mix of highly heterogeneous flows with specific

characteristics and constrains

We propose to combine tools and software for processing flows at

different levels /places of the grid :



GNET1 & GNET10 (AIST): for efficient flow/packet processing

PSPacer (AIST): for simple packet pacing



GridMPI (AIST): for efficient MPI WAN communications



BDTS (RESO): a Grid Service for controlled bulk data transfers



FlowCTL(RESO): simple API & software for fine grain flow time/rate control

MetroGRID (RESO): A Grid Service to measure and model grid traffic

P. Vicat-Blanc Primet ₅

Packet processing: GNET1 & GNET10

Latency emulation: explore wide area behavior

Precise measurement (delays, aggregate throughput)

High speed packet generation



TCP variants evaluation (10G - 1ms to 200ms)



L2 limits in congestion situation (starvation phenomenon)

Evaluation of rate limitation mechanisms (Pspacer)



Compare hardware and software emulation solution (collab with

MESCAL- Grenoble)

High Speed Packet capture



Header extraction



Header concatenation and sending



Precise and fine grain traffic & flow analysis

P. Vicat-Blanc Primet ₆

TCP variants evaluation

...

...

10 GbE

iperf_iperf

iperf iperfiperf_iperfd

Grid5000 backbone

Or 10Gb/s WAN emulator (GtrcNET1 or GtrcNET10)

1 to N (=40) nodes with 1 or 10GbE interfaces

1 to N ( =40) nodes with 1 or 10GbE interfaces

Romaric Guillier, Ludovic Hablot, Yuetsu Kodama, Tomohiro Kudoh, Fumihiro Okazaki, Pascale Vicat-Blanc

Primet, Sébastien Soudan and Ryousei Takano

Next step: Naregi-Grid5000 link

GNET

_G5K

A study of large flow interactions in high-speed

shared networks with Grid5000 and GtrcNET-10

P. Vicat-Blanc Primet ₇

Reverse traffic impact on efficiency and fairness

[Guillier - Soudan - Primet - High Speed Networks workshop of Infocom2007]

19 flows forward /19 flows reverse Affects flow performance by 20% & global throughput, unstability

2 flows forward / 2 flows: reverse : Affects flow performance by 50% & global throughput, unstability Without reverse traffic:

Efficiency,

Fairness & equilibrium Around 470Mb/s per flow 940Mb/s global

Congesting reverse traffic: No efficiency,

Fairness & unstability Around 160Mb/s per flow

Without reverse traffic: Efficiency,

Fairness & equilibrium Around 470Mb/s per flow 9200 Mb/s global

With reverse traffic: No fficiency,

Fairness & unstability Around 400Mb/s per flow 8200 Mb/s global

Packet capture with GNET1

plan

Yichun Li, Yuetsu Kodama,

P. Vicat-Blanc Primet ₁₀

Activities around Grid MPI eval in G5K

Ludovic Hablot, Olivier Gluck, Pascale Primet

with the help of GridMPI team at AIST

P. Vicat-Blanc Primet ₁₁

Bulk Data Transfer Scheduling (BDTS)

Goal of the service :

Improve Bulk Data Transfers:

-delay determinism & reliability by using

=> admission control & scheduling at flow level

=> dynamic provisioning at physical (optical) level

-Performance :

ie : Resource utilization & transfer duration (utility functions)

⇒

optimize resource allocation

⇒

handle congestion control (TCP performs better)

⇒

optimize on demand resource provisionning

Sebastien Soudan, Dinil M. Divakaran, Chen Cheng, Pascale Primet

P. Vicat-Blanc Primet ₁₂

P. Vicat-Blanc Primet ₁₃

BDTS: Reservation flexibility

A bulk data transfer can start from any time after its arrival, at

any and even time variant bandwidth value, as long as it is

completed before its deadline

P. Vicat-Blanc Primet ₁₄

BDTS problem : objective function

Previous works minimized makespan given network

capacity constraint, our approach is to

minimize

congestion given time constraints:

1. Can deal with the case when different tasks has

different (hard) time constraint

2. Suitable for networks where capacity is provisioned

based on the scheduling result (e.g. optical or

overlay network)

3. Minimize congestion improves performance of

coexisting best-effort interactive traffic

P. Vicat-Blanc Primet ₁₅

Bulk Data Transfer Scheduling problem

P. Vicat-Blanc Primet ₁₆

P. Vicat-Blanc Primet ₁₇

P. Vicat-Blanc Primet ₁₈

P. Vicat-Blanc Primet ₁₉

BDTS current work & perspectives

BDTS service specification and API definition

BDTS scheduler implementation & deployment in Grid5000

End-host based time/rate control mechanisms (FlowCTL) &

interaction with High Speed TCP variants

Interaction with candidate users (upper layers : workflow

engines, SAGA, Distributed File System, RPC…)

Interaction with Bandwidth/light path reservation services

(lower layers)

P. Vicat-Blanc Primet ₂₀

Conclusion and perspectives

Grid traffic is a mix of highly heterogeneous flows with specific

characteristics and constrains



Large flows with QoS requirements (transfer delay & reliability)

Small messages sensitive to slow start effect



Grid management & control traffic

High performance encrypted channels

We propose to combine tools and software for processing flows

at different level /place of the grid to



Measure & monitor the traffic (awareness, admission control)

Differentiate the packets (IP QoS, traffic & flow management)

Schedule the flows (congestion control)

P. Vicat-Blanc Primet ₂₁

Future plan for our collaboration

Context

 GRIDNET-FJ : INRIA associated team  NEGST : JST- CNRS (Naregi-Grid5000)

 PhD student, Sebastien Soudan at AIST ( 6 months -

bourse Lavoisier

Topics

 TCP variant benchmkg (eWAN) on G5K-Naregi tb (Romaric Guillier - PhD)  Packet Header capture on Grid5000 - Naregi link (Yichun Li - IE)

 Header capture & sampling(GNET10) for traffic analysis (Patrick Loiseau

-PhD )

 FlowCTL, BDTS & G-Lambda GNS integration (Sebastien Soudan - PhD)  Virtual clusters interconnection optimization ( Dinil Mon Divakaran - PhD)  MPI evaluations in Grid5000-Naregi (Ludovic Hablot - PhD)

Other possible perspectives

 Collaborative experiments G5K & Naregi on

 Grid Traffic measurement (Naregi), GridFTP (Osaka)

 Collaborative Demo at SC07 (with AIST: GNET, eWAN, G5K, Naregi…)  Standardisation activities (GHPN, NML)

RESO is hosting GridNets2007 (210-07) & CCGrid2008 & probably

OGF23 in 05-2008 à Lyon