+ All Categories
Home > Documents > BNet: further investigations

BNet: further investigations

Date post: 22-Jan-2016
Category:
Upload: minna
View: 22 times
Download: 0 times
Share this document with a friend
Description:
BNet: further investigations. • • • •. n. H. Currently investigated structure. n=4 : 16x16. n * (n - 1) / 2 bidirectional connections. TG/BC. n - 1 ports. n - 1 ports. BNet controller. switch n × n. switch n × n. • • • •. n. • • •. n - 1. H: histogrammer TG: event tagger - PowerPoint PPT Presentation
14
GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 1 DAQ BNet
Transcript
Page 1: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 1

DAQBNet

Page 2: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 2

DAQBNet Currently investigated structure

switch n × n • • • • switch n × nn

n * (n - 1) / 2 bidirectional connections

• • • • • • • n

CNet PNet

n - 1

n - 1 ports n - 1 ports

H: histogrammerTG: event taggerHC: histogram collectorBC: schedulerDD: data dispatcherED: event dispatcher

TG/BC

DD/ED

CNet PNet

DD/ED

HCNet

DD/HC active buffer

BNet controller

n=4 : 16x16

Page 3: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 3

DAQBNet Simulation with SystemC

• event generator

• data dispatcher (sender)

• histogram collector

• tag generator

• BNet controller (schedule)

• event dispatcher (receiver)

• transmitter (data rate, latency)

• switches (buffer capacity, max. # of package queue, 4K)

Running with 10 switches and 100 end nodes.

Simulation takes 1.5 *105 times longer than simulated time.

Various statistics (traffic, network load, etc.)

Modules:

Page 4: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 4

DAQBNet Some statistic examples

single buffers excluded!

Page 5: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 5

DAQBNet Scheduling strategy

• use data channels for meta data

• traffic shaping by schedule

• calculate schedule in advance

• data transfers are triggered at absolut times

• requires deterministic behavior of network

• delivery control by buffer monitoring

• no transfer acknowledge

Page 6: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 6

DAQBNet Factorisation of 32x32 switch

1-1

1-2

1-3

1-4

1-5

1-6

6-1

6-2

6-3

6-4

6-5

6-6

1-1

1-2

1-3

1-4

1-5

1-6

6-1

6-2

6-3

6-4

6-5

6-6

• • • •

2 layers of 6 6x6 switches

Variance in latency increases!

Page 7: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 7

DAQBNet Factorisation of 32x32 switch

Variance in latency increases!

factorized switch 1

factorized switch 2

2 switches

1 switch 4 switches

Page 8: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 8

DAQBNet Synchronization of senders

• Extra time clock distribution (~µs)

• or from concentrator (must then be synchronized)

The active buffers must have a common clock to start transfers at a given absolut time.

Page 9: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 9

DAQBNet Broadcast problematic

• Broadcast through factorized switches

• How to guarantee packet delivery?

Broadcast is needed to distribute event tags andschedule start time to all data dispatchers.

Page 10: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 10

DAQBNet Error detection

• How to ensure that all data arrived?

• How to recover if not?

• Recover from component fail (or removal)?

• When emergency?

BNet controller gets buffer status from all DD/EDs.These are sent according schedule.

Page 11: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 11

DAQBNet Network performance measurements

chassis IXIA 1600T

IXIA testbed components

10 GE load module (traffic generator/analyzer)

Page 12: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 12

DAQBNet Network performance measurements

TALLY: measurement of Force10 E1200

Point to point

Page 13: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 13

DAQBNet Topics for investigations

• Event shaping

• Separate meta data transfer system

• Addressing/routing schemes

• Broadcast

• Synchronization

• Determinism

• Fault tolerance

• Real test bed

Page 14: BNet: further investigations

GSI, Mar 2005 Hans G. Essel, Sergey Linev: CBM - DAQ BNet 14

DAQBNet


Recommended