Dynamic Network Resources Allocation in Grids through a Grid ...

INGRID 2007 – Instrumenting the GRID Second International Workshop on Distributed Cooperative Laboratories

Session 2: Networking for the GRID

Dynamic Network Resources Allocation in Grids through a Grid Network Resource Broker

Davide Adami, Stefano Giordano, Michele Pagano CNIT Research Unit Dept. of Information Engineering - University of Pisa 1

Davide Adami – S.Margherita Ligure, Italy, April, 16-18, 2007

Outline • Introduction • Target and motivations of the research activity • Grid Network Resource Broker Architecture • WCBDS (Wang-Crowcroft with Bandwidth and Delay Sorting) Path Computation Algorithm • Simulations results • Conclusions

2


A parallel renderer/encoder

Frame Sequencer

GoP Assembler

Output Store DivX Encoding

Parallel Rendering

3


Instrumentation/Computing Grid Environment High Speed Optical Network with G-MPLS Control Plane

Grid Concept “A Grid is a collection of distributed computing resources over network that appear to an user or an application as one large virtual computing system” 4


Grid Networking Issues 1. A network infrastructure which prevents degrading the throughput of grid applications due to network delay and network fault is required 2. It is necessary to carry out network resource scheduling as well as computing resource scheduling. 3. A network design and deployment methodology for complicated grid networking is necessary. Grid Application Grid Application

Grid Application

Diffserv-aware MPLS TE network

5


Scheduling Process: Enhanced Deployment Cycle •Application nodes are mapped into computing resources •Cumulative bandwidth requirements are given

Weighted Task Interaction Graph of the application •Vertex: Computational Cost •Edge: Communication Cost

•Network Query (list of candidate solutions) •Reservation

6


Grid Network-Aware Environment GNRB Monitoring and Management Area Grid Network Resource Broker

Grid Application Manager

Diffserv-aware MPLS TE network

PC Cluster

7


GNRB Functional Blocks Grid Application Grid Application Manager

Visualization GUI Admission Control Module Network Information Database

Path Computation Element Network Resource Scheduler

Measurement Sampling Modules

Measurement Database

Network Resources Manager

Network Monitoring System

Network Element Configuration Manager

8

Topology Discovery Service

Sampling

Sampling

Capturing device

Capturing device


GNRB Architecture • Network Resources Manager • Policy-based provisioning • Path computation • Network Resources Scheduling

• Topology Discovery Service • Network Element Configuration Manager • Service provisioning

• GNRB and Network Monitoring System • Link utilization • QoS measurements (packet loss, delay, jitter) • Implemented: SNMP agent for bandwidth utilization • Implemented: UNIPI MPLS/DiffServ network traffic monitoring tool 9


Network Monitoring System •

DiffServ/MPLS network traffic monitoring •

10

Graphs for each LSP/PHB pair are available (measured and predicted values)


GNRB Network Services • Network Topology Discovery • Provides information about the topology of the network and QoS metrics associated to the links • Best-effort connections

• Weighted Topology Discovery • Best paths, according to a metric specified by the GAM are computed by the NRM • Network resources may be allocated

• QoS provisioning • Premium service (Peak Rate, Burst Size, Latency) • Better than BE service (Mean Rate, Burst Size, Mean Latency) • End-to-end connections with QoS constraints are established 11


Path Computation Algorithm Goal Given a set of N LSP set-up requests, the basic function of the PCE is to find N network paths that satisfy their QoS constraints (Bandwidth Bmin, Delay Dmax) QoS Metrics •Bandwidth: Concave metric

l

k

m

j

B(p)=min[B(i,j); B(j,k); .. B(l,m)] i •Delay: Additive metric D(p)=D(i,j)+D(j,k)+…D(l,m)

Path p = i,j,k,l,m

12


End-to-end Delay Propagation delay End-to-end Delay

Transmission delay Queueing delay

Queueing delay: Deterministic Upper Bound Delay for LSP i

k M j Di = + ∑ Si j =1 ri

M = max burst size r = guaranteed rate LSP i i

Node j Delay in case of WFQ scheduling discipline

Si j =

Lmax Li + Rj ri

Lmax = max packet size Li = max packet size LSP i R j = output link bandwidth

13


The WCBDS Algorithm N Requests

WC Algorithm Z Requests accepted Z=N?

N Requests

Yes

EXIT

No Bandwidth Based Re-ordering

Wang-Crowcroft Algorithm WC Algorithm Z Requests accepted Z=N?

Yes

EXIT

No Delay Based Re-ordering

N Requests

1. Set dij= ∞ if Bij < Bmin 2. Compute the path P with the minimum delay 3. Calculate the delay D* of P 4. If D* < Dmax select the path P otherwise reject the request

WC Algorithm Z Requests accepted Z=N?

Yes

EXIT

No ERROR

14


Wang-Crowcroft Algorithm B

3 - 30

4 - 10

4 - 20

Bmin= 3 Mbps

X

2 - 20

A

Dmax=100 ms

C

X 2 – 30 4 - 20

F

X 6 - 10

1 - 30

Bmin= 3 Mbps Dmax=98 ms

Rejected!

1) Prune the links with Bav < Bmin

2) Find minimum delay path 3) Check if D < Dmax 15

D

4 – 20

E 1.

A_B_C_F

2.

A_D_E_C_F

1.

D = 96.43ms

2.

D = 102.49ms

96.43ms < 100ms Davide Adami – S.Margherita Ligure, Italy, April, 16-18, 2007

NS2 Software Modules Old Modules

MNS - MPLS Network Simulator

RSVP-TE\ns with Reservation Styles OSPF-TE\ns New Modules MPLS Recovery Strategies Path Computation Algorithm

16


New MPLS Node Architecture in NS2

OSPF-TE OSPF-TE module module

17

RSVP-TE RSVP-TE module module


Network Topology MPLS Backbone (1, 50) (0.3, 100)

Node0

(2, 10)

Network 0 LSR4 Node1

Node17

(2, 10) LSR9

LSR10

Network 1 LSR16

(2, 20)

Node18

(2, 10) LSR8

Node19

(2, 100)

(1, 10) LSR5

(1, 10)

(1, 30) LSR14 (2.5, 15)

LSR11

LSR15

(2, 20)

Node2

Network 2

(2.5, 10)

LSR6 (2.5, 10)

(2.5, 10) LSR7

(2.5, 10) Node20

(2.5, 10)

(2.5, 10)

LSR13

Node3 LSR12

(Bandwidth x Mbps, Delay y ms)

18


First Scenario Node17

Node0 LSR9 Network 0

LSR10

LSR4

Network 1 LSR16

Node18

LSR8

Node1

Node19

LSR5 LSR6

LSR14 LSR15

LSR11

Node2 LSR7

MPLS Backbone

Network 2

Node20

LSR13

LSR12

Node3

Ingress LER

19

Egress LER

Bandwidth (Kbps)

Delay (ms)

Path

Time (ms)

4

16

600

100

4-5-8-10-16

84

16

15

100

20

16-15

11

15

4

1800 200 15-14-11-6-4 Davide Adami – S.Margherita Ligure, Italy, April, 16-18, 2007

110

Second Scenario Node17


LSR10

LSR4

Network 1 LSR16

Node18

LSR8

Node1

Node19

LSR5 LSR6

LSR14 LSR15

LSR11

Node2

MPLS Backbone

LSR7

Network 2

Node20

LSR13

LSR12

Node3

Ingress LER

20

Egress LER

Bandwidth (Kbps)

Delay (ms)

Path

Time (ms)

4

16

600

100

4-5-8-10-16

84

16

15

100

20

16-15

11

15

4 Davide Adami – 2400 200 15-13-12-7-6-4 S.Margherita Ligure, Italy, April, 16-18, 2007

95

Third Scenario Node17


LSR10

LSR4

Network 1 LSR16

Node18

LSR8

Node1

Node19

LSR5 LSR6

LSR14 LSR15

LSR11

Node2 LSR7

Traffic Load 75% on the path 15_14_11_6_4

Ingress LER

21

MPLS Backbone

Network 2

Node20

LSR13

LSR12

Node3

Egress LER

Bandwidth (Kbps)

Delay (ms)

Path

Time (ms)

4

16

600

100

4-5-8-10-16

84

16

15

100

20

16-15

11

4

1800

200

15-14-11-6-4

1287

15


Fourth Scenario Node17


LSR10

LSR4

Network 1 LSR16

Node18

LSR8

Node1

Node19

LSR5 LSR6

LSR14 LSR15

LSR11

Node2 LSR7

Traffic Load 75% on the path 15_14_11_6_4

Ingress LER

22

MPLS Backbone

Network 2

Node20

LSR13

LSR12

Node3

Egress LER

Bandwidth (Kbps)

Delay (ms)

Path

Time (ms)

4

16

600

100

4-5-8-10-16

84

16

15

100

20

16-15

11

15

4 2400 15-13-12-7-6-4 Davide Adami – S.Margherita Ligure,200 Italy, April, 16-18, 2007

1287

Conclusion • The design and deployment of grids for remote instrumentation services require the introduction of new control plane mechanisms to dynamically allocate resources in high-speed (G)-MPLS networks • A centralized approach, based on a GNRB, has been designed and developed • A new algorithm for the computation of path with bandwidth and delay constraints has been proposed • Preliminary simulation results are promising • Next step: implementation and testing in a real grid environment

23
