Download Proposal-talk

Survey
yes no Was this document useful for you?
   Thank you for your participation!

* Your assessment is very important for improving the workof artificial intelligence, which forms the content of this project

page
1
page
2
page
3
page
4
page
5
page
6
page
7
page
8
page
9
page
10
page
11
page
12
page
13
page
14
page
15
page
16
page
17
page
18
page
19
page
20
page
21
page
22
page
23
page
24
page
25
page
26
page
27
page
28
page
29
page
30
page
31
page
32
page
33
page
34
page
35
page
36
Document related concepts
no text concepts found
Transcript 
Control and Data Networks
Architecture Proposal
S. Stancu, C. Meirosu, B. Martin
31 May, 2006
Stefan Stancu
1
Outline
• Overview of the TDAQ system and networks
• Technology and equipment
• TDAQ networks:
Control network - no special bandwidth requirement
Dedicated data networks:
DataFlow network – high bandwidth (~100Gbit/s cross-sectional
bw.) and minimal loss
BackEnd network – high bandwidth (~ 50Gbit/s cross-sectional
bw.)
Switch management network
• Conclusions
31 May, 2006
Stefan Stancu
2
ATLAS TDAQ System/Networks
Network root
(to ATCN)
Control Links
Data Links
Management links
Online Racks
High availability
Management
Control Network
24/7 when the accelerator is Servers
running
ROB
ROB ROB
ROB ROB
ROB
ROS PC
ROS PC
ROB
ROB ROB
ROB ROB
ROB
DataFlow Network
SVs
Management
Network
SFIs
SFO
Switch
L2PUs
BackEnd Network
Mass
storage
SFOs
EFPs
31 May, 2006
EFPs
EFPs
Stefan Stancu
3
Technology and equipment
• Ethernet is the dominant technology for LANs
 TDAQ’s choice for networks (see [1])
 multi-vendor, long term support, commodity (on-board GE adapters), etc.
 Gigabit and TenGigabit Ethernet
 Use GE for end-nodes
 10GE whenever the bandwidth requirements exceed 1Gbit/s
• Multi-vendor Ethernet switches/routers available on the market:
 Chassis-based devices ( ~320 Gbit/s switching)
 GE line-cards: typically ~40 ports (1000BaseT)
 10GE line-cards: typically 4 ports (10GBaseSR)
 Pizza-box devices (~60 Gbit/s switching)
 24/48 GE ports (1000BaseT)
 Optional 10GE module with 2 up-links (10GBaseSR)
31 May, 2006
Stefan Stancu
4
Architecture and management principles
•
The TDAQ networks will typically employ a multi-layer architecture:
 Aggregation layer. The computers installed in the TDAQ system at Point 1 are
grouped in racks, therefore it seems natural to use a rack level concentrator switch
where the bandwidth constraints allow this.
 Core layer. The network core is composed of chassis-based devices, receiving GE or
10GE up-links from the concentrator switches, and direct GE connections from the
applications with high bandwidth requirements.
•
The performance of the networks themselves will be continuously monitored for
availability, status, traffic loss or other errors.
 In-band management – The “in-band” approach allows intelligent network monitoring
programs not only to indicate the exact place of a failure (root cause analysis) but also
what are the repercussions of a failure to the rest of the network..
 Out-of-band management – A dedicated switch management network (shown in blue
in Figure 1) provides the communication between management servers (performing
management and monitoring functions) and the “out-of-band” interface of all network
devices from the control and data networks.
31 May, 2006
Stefan Stancu
5
Resilient Ethernet networks
• What happens if a switch or link fails?
 Phone call, but nothing critical should happen after a single failure.
• Networks are made resilient by introducing redundancy:
 Component-level redundancy: deployment of devices with built-in
redundancy (PSU, supervision modules, switching fabric)
 Network-level redundancy: deployment of additional devices/links in order
to provide alternate paths between communicating nodes.
 Protocols are needed to correctly (and efficiently) deal with multiple paths in
the network [2]:
 Layer 2 protocols: Link aggregation (trunking), spanning trees (STP, RSTP, MSTP)
 Layer 3 protocols: virtual router redundancy (VRRP) for static environments,
dynamic routing protocols (e.g. RIP, OSPF).
31 May, 2006
Stefan Stancu
6
Control network
10 GE Control Links
Network root
(to ATCN)
•
~3000 end nodes.Design assumption:
4
Online racks
 No end-node will generate control traffic
in excess to the capacity of a GE line.
 The aggregated control traffic external to a
rack of PCs (with the exception of the
Online Services rack) will not exceed the
capacity of a GE line.
•
•
•
The network core is implemented with
two chassis devices redundantly
interconnected by two 10GE lines.
A rack level concentration switch can be
deployed for all units except for critical
services.
Protocols/Fault tolerance:
Conc.
Core 2
Conc.
Conc.
Conc.
17
LVL2 racks
1
DC Control
rack
 Resiliency:
 VRRP in case of Router/up-link failures
 Interface bonding for the infrastructure
servers.
Management
Servers
Core 1
 Layer 3 routed network
 Static routing at the core should suffice.
 One sub-net per concentrator switch
 Small broadcast domains  potential layer
2 problems remain local.
GE Control Links
4
SFI racks
57
EF racks
Conc.
3
SFO racks
DataFlow Network
BackEnd Network
SFO
Switch
31 May, 2006
Stefan Stancu
7
DataFlow network
~160 ROSs
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
A
~20 ROS
conc.
switches
B
ROS PC
ROS PC
A
ROS conc.
ROB
ROS PC
ROS PC
A
B
ROB
ROB
ROB
ROB
ROB
B
ROS conc.
ROS conc.
B (A)
B (A)
B (A)
A (B)
A (B)
Control
Network
A (B)
A,B
A,B
A
A,B
Central 1
B
A,B
Central 2
B
A
17 L2PU
racks
B
SVs
L2PU conc.
L2PU conc.
B
A
L2PUs
B
A
A
A
Control
Network
SFIs
~100 SFIs
A
SFIs
B
L2PUs
B
BackEnd Network
~550 L2PUs
L2PU conc.
A
A
L2PUs
L2PU conc.
GE Data Link
10GE Data Link
B
B
L2PUs
GE Control Link
31 May, 2006
Stefan Stancu
8
DataFlow network
~160 ROSs
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
A
~20 ROS
conc.
switches
B
ROS PC
ROS PC
A
ROS conc.
ROB
ROS PC
ROS PC
A
B
ROB
ROB
ROB
ROB
ROB
B
ROS conc.
ROS conc.
B (A)
B (A)
B (A)
A (B)
A (B)
Control
Network
A (B)
~100 Gbit/s
A,B
A,B
A
A,B
Central 1
B
A,B
Central 2
B
A
17 L2PU
racks
B
SVs
L2PU conc.
L2PU conc.
B
A
L2PUs
B
A
A
A
Control
Network
SFIs
~100 SFIs
A
SFIs
B
L2PUs
B
BackEnd Network
~550 L2PUs
L2PU conc.
A
A
L2PUs
L2PU conc.
GE Data Link
10GE Data Link
B
B
L2PUs
GE Control Link
31 May, 2006
Stefan Stancu
9
DataFlow network
~160 ROSs
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
A
~20 ROS
conc.
switches
B
ROS PC
ROS PC
A
ROS conc.
ROB
ROS PC
ROS PC
A
B
ROB
ROB
ROB
ROB
ROB
B
ROS conc.
ROS conc.
B (A)
B (A)
B (A)
A (B)
A (B)
Control
Network
A (B)
A,B
A,B
A
A,B
Central 1
B
A,B
Central 2
B
A
17 L2PU
racks
B
SVs
L2PU conc.
L2PU conc.
B
A
L2PUs
B
A
A
A
Control
Network
SFIs
A
~100 SFIs
SFIs
Two central switchesB
B
L2PUs
•One VLAN per switch (A and B)
•Fault tolerant
BackEnd(tolerate
Networkone switch failure)
~550 L2PUs
L2PU conc.
A
A
L2PUs
L2PU conc.
GE Data Link
10GE Data Link
B
B
L2PUs
GE Control Link
31 May, 2006
Stefan Stancu
10
DataFlow network
~160 ROSs
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
A
~20 ROS
conc.
switches
ROB
ROB
ROS PC
ROS PC
ROB
ROB
ROS PC
10G ROS “concentration”
ROS PC
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
•Motivated by the need to use fibre transmission (distance > 100m)
A
A
B provided by aggregating
•Full bandwidth
10x BGE into 1x10GE
•Requires the use of VLANs (and MST) to maintain a loop-free topology
B
ROS conc.
ROB
ROS conc.
ROS conc.
B (A)
B (A)
B (A)
A (B)
A (B)
Control
Network
A (B)
A,B
A,B
A
A,B
Central 1
B
A,B
Central 2
B
A
17 L2PU
racks
B
SVs
L2PU conc.
L2PU conc.
B
A
L2PUs
B
A
A
A
Control
Network
SFIs
~100 SFIs
A
SFIs
B
L2PUs
B
BackEnd Network
~550 L2PUs
L2PU conc.
A
A
L2PUs
L2PU conc.
GE Data Link
10GE Data Link
B
B
L2PUs
GE Control Link
31 May, 2006
Stefan Stancu
11
DataFlow network
~160 ROSs
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
A
~20 ROS
conc.
switches
B
ROS PC
ROS PC
A
ROS conc.
ROB
ROS PC
ROS PC
A
B
ROB
ROB
ROB
ROB
ROB
B
ROS conc.
ROS conc.
B (A)
B (A)
B (A)
A (B)
A (B)
Control
Network
A (B)
A,B
A,B
A
A,B
Central 1
B
A,B
Central 2
B
A
17 L2PU
racks
B
SVs
L2PU conc.
L2PU conc.
B
A
L2PUs
B
A
A
A
Control
Network
SFIs
~100 SFIs
A
SFIs
B
L2PUs
B
BackEnd Network
~550 L2PUs
L2PU conc.
A
A
10G L2PUL2PUs
concentration
•One switch per rack
31 May, 2006
L2PU conc.
GE Data Link
10GE Data Link
B
B
L2PUs
GE Control Link
Stefan Stancu
12
DataFlow network
~160 ROSs
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
A
~20 ROS
conc.
switches
B
ROS PC
ROS PC
A
ROS conc.
ROB
ROS PC
ROS PC
A
B
ROB
ROB
ROB
ROB
ROB
B
ROS conc.
ROS conc.
B (A)
B (A)
B (A)
A (B)
A (B)
Control
Network
A (B)
A,B
A,B
A
A,B
Central 1
B
A,B
Central 2
B
A
17 L2PU
racks
B
SVs
L2PU conc.
L2PU conc.
B
A
L2PUs
B
A
A
A
Control
Network
SFIs
~100 SFIs
A
SFIs
B
L2PUs
B
BackEnd Network
~550 L2PUs
L2PU conc.
A
A
L2PUs
L2PU conc.
GE Data Link
10GE Data Link
B
B
L2PUs
GE Control Link
31 May, 2006
Stefan Stancu
13
BackEnd network
GE Data Link
10GE Data Link
DataFlow Network
GE Control Link
~100 SFIs
SFIs
Control Network
SFOs
Central 1
57 EF
racks
EF conc.
EF conc.
EF conc.
SFO conc.
~30 SFOs
•
•
•
EFs
EFs
EFs
•
31 May, 2006
Stefan Stancu
Mass
storage
~2000 end-nodes
One core device with built-in
redundancy (eventually two devices)
Rack level concentration with use of
link aggregation for redundant uplinks to the core.
Layer 3 routed network to restrict
broadcast domains size.
14
BackEnd network
GE Data Link
10GE Data Link
DataFlow Network
GE Control Link
~100 SFIs
SFIs
~30 SFOs
~2.5 Gbit/s
SFOs
Central 1
57 EF
racks
EF conc.
EF conc.
EF conc.
SFO conc.
~50 Gbit/s
Control Network
•
•
•
EFs
EFs
EFs
•
31 May, 2006
Stefan Stancu
Mass
storage
~2000 end-nodes
One core device with built-in
redundancy (eventually two devices)
Rack level concentration with use of
link aggregation for redundant uplinks to the core.
Layer 3 routed network to restrict
broadcast domains size.
15
BackEnd network
GE Data Link
10GE Data Link
DataFlow Network
GE Control Link
~100 SFIs
SFIs
Control Network
SFOs
Central 1
57 EF
racks
EF conc.
EF conc.
EF conc.
SFO conc.
~30 SFOs
•
•
•
EFs
EFs
EFs
•
31 May, 2006
Stefan Stancu
Mass
storage
~2000 end-nodes
One core device with built-in
redundancy (eventually two devices)
Rack level concentration with use of
link aggregation for redundant uplinks to the core.
Layer 3 routed network to restrict
broadcast domains size.
16
Interchangeable processing power
•
•
Standard processor rack with up-links
to both DataFlow and BackEnd
networks.
The processing power migration
between L2 and EF is achieved by
software enabling/disabling of the
appropriate up-links.
ROB
ROB ROB
GE Data Link
10GE Data Link
ROB ROB
ROB
ROS PC
ROS PC
ROB
ROB ROB
ROB ROB
ROB
GE Control Link
DataFlow Network
Data conc.
Ctrl. Conc.
PUs
SVs
SFIs
L2PUs
BackEnd Network
XPU rack
EFPs
31 May, 2006
Stefan Stancu
EFPs
17
EFPs
Switch Management Network
Management
Servers
Control Network
• It is good practice not to mix
management traffic and
normal traffic.
 If the normal traffic has
abnormal patterns it may
overload some links and
potentially starve the in-band
management traffic
• The switch management
network provides the
management servers access the
out-of-band interface of all the
devices from the control and
data networks
• Flat Layer 2 Ethernet network
with no redundancy.
SDX Level 2
Ctrl. core
Ctrl. core
Mgmt root
(24 port)
24 port
48 port
DF core
DF core
BE core
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
SDX Level 1
24 port
48 port
48 port
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
Ctrl. Rack Conc.
Data Rack conc.
USA 15 Level 2
24 port
ROS Data conc.
ROS Data conc.
GE Mgmt Link
GE fibre Mgmt Link
GE Control Link
24 port
USA 15 Level 1
ROS Data conc.
ROS Data conc.
31 May, 2006
Stefan Stancu
18
Conclusions
• The ATLAS TDAQ system (approx. 3000 end-nodes)
relies on networks for both control and data acquisition
purposes.
• Ethernet technology (+IP)
• Networks architecture maps on multi-vendor devices
• Modular network design
• Resilient network design (high availability)
• Separate management path
31 May, 2006
Stefan Stancu
19
Back-up slides
31 May, 2006
Stefan Stancu
20
Option A
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
ROS PC
ROS PC
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
ROB
ROB
ROB
ROB
ROB
USA 15
SDX
Central 1
SVs
L2PU conc.
L2PUs
31 May, 2006
Central 2
SFIs
•1GE links
•Central Sw. in SDX
L2PU conc.
SFIs
Stefan Stancu
L2PUs
1000-BaseT (GE copper)
1000-BaseSX (GE fibre)
10G-BaseSR (10GE copper)
21
Option B
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROB
ROS PC
ROS PC
ROB
ROS PC
ROB
ROS conc.
ROS PC
ROB
ROS PC
ROB
ROB
ROS conc.
ROS conc.
USA 15
SDX
Central 1
SVs
L2PU conc.
L2PUs
31 May, 2006
SFIs
•10GE links
•Central Sw. in SDX
Central 2
L2PU conc.
SFIs
Stefan Stancu
L2PUs
1000-BaseT (GE copper)
1000-BaseSX (GE fibre)
10G-BaseSR (10GE copper)
22
Option C
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS PC
ROS PC
ROS PC
ROS PC
Central 1
ROB
ROS PC
ROS PC
ROB
ROB
ROB
ROB
ROB
•10GE links
•Central Sw. inUSA
USA
15
Central 2
SDX
SV conc.
L2PU conc.
L2PUs
31 May, 2006
SFI conc.
SFI conc.
SFIs
SVs
SFIs
Stefan Stancu
L2PU conc.
L2PUs
1000-BaseT (GE copper)
1000-BaseSX (GE fibre)
10G-BaseSR (10GE copper)
23
20 ROSs trunking
ROB
ROB
ROB
ROB
ROB
ROB
10x
ROS
PCPC
ROS
10xGE
A
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
10xGE
B
ROS conc.
ROS conc.
Trunk
10x
ROS
PCPC
ROS
10xGE
B
10xGE
A
ROB
ROB
ROB
ROB
ROB
ROB
B B
Trunk
A A
A,B
A,B
A,B
Central 1
A,B
Central 2
B
A
A
SVs
L2PU conc.
A
B
A
L2PU conc.
B
A
L2PUs
B
SFIs
SFIs
B
L2PUs
GE Data Link
10GE Data Link
31 May, 2006
Stefan Stancu
24
20 ROSs VLANs
ROB
ROB
ROB
ROB
ROB
ROB
10x
ROS
PCPC
ROS
10xGE
A
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROB
10xGE
B
ROB
ROB
ROB
ROB
ROB
ROB
10x
ROS
PCPC
ROS
10xGE
A
10xGE
B
ROS conc.
ROS conc.
B (A)
A (B)
A (B)
B (A)
A,B
A,B
A,B
Central 1
A,B
Central 2
B
A
A
SVs
L2PU conc.
A
B
A
L2PU conc.
B
A
L2PUs
B
SFIs
SFIs
B
L2PUs
GE Data Link
10GE Data Link
31 May, 2006
Stefan Stancu
25
10 ROSs single switch
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS10x
PC
ROS PC
10xGE
A
ROB
ROB
ROB
ROB
ROB
10xGE
B
ROS conc.
B (A)
A (B)
A,B
A,B
A,B
Central 1
A,B
Central 2
B
A
SVs
L2PU conc.
A
B
A
A
L2PU conc.
B
A
L2PUs
B
SFIs
SFIs
B
L2PUs
GE Data Link
10GE Data Link
31 May, 2006
Stefan Stancu
26
10 ROSs double switch
ROB
ROB
ROB
ROB
ROB
ROB
ROB
ROS10x
PC
ROS PC
ROB
ROB
ROB
ROB
ROB
10xGE
B
10xGE
A
ROS conc.
ROS conc.
B
A
A,B
A,B
A,B
Central 1
A,B
Central 2
B
A
SVs
L2PU conc.
A
B
A
A
L2PU conc.
B
A
L2PUs
B
SFIs
SFIs
B
L2PUs
GE Data Link
10GE Data Link
31 May, 2006
Stefan Stancu
27
Sample resiliency test (see [4])
MST resiliency test
10
9
VLAN2
8
Throughput (Gbit/s)
VLAN1
Core
MST
7
6
5
4
3
2
Aggregation
1
10
VLAN1 throughput
VLAN2 throughput
20
30
40
50
60
70
80
90
Time (s)
VLAN1
31 May, 2006
VLAN2
Stefan Stancu
28
100
Sample resiliency test
MST resiliency test
10
9
VLAN2
8
Throughput (Gbit/s)
VLAN1
Core
MST
7
6
5
4
3
2
Aggregation
1
10
VLAN1 throughput
VLAN2 throughput
20
30
40
50
60
70
80
90
Time (s)
VLAN1
31 May, 2006
VLAN2
Stefan Stancu
29
100
Sample resiliency test
MST resiliency test
10
9
VLAN2
8
Throughput (Gbit/s)
VLAN1
Core
MST
7
6
5
4
3
2
Aggregation
1
10
VLAN1 throughput
VLAN2 throughput
20
30
40
50
60
70
80
90
Time (s)
VLAN1
31 May, 2006
VLAN2
Stefan Stancu
30
100
Sample resiliency test
MST resiliency test
10
9
VLAN2
8
Throughput (Gbit/s)
VLAN1
Core
MST
7
6
5
4
3
2
Aggregation
1
10
VLAN1 throughput
VLAN2 throughput
20
30
40
50
60
70
80
90
Time (s)
VLAN1
31 May, 2006
VLAN2
Stefan Stancu
31
100
Sample resiliency test
MST resiliency test
10
9
VLAN2
8
Throughput (Gbit/s)
VLAN1
Core
MST
7
6
5
4
3
2
Aggregation
1
10
VLAN1 throughput
VLAN2 throughput
20
30
40
50
60
70
80
90
Time (s)
VLAN1
31 May, 2006
VLAN2
Stefan Stancu
32
100
Control network – Online services
Control - Core
C1
C2
C3
C4
Control - Core
GE ctrl
10GE ctrl
GE mgmt
M5
C1
C2
C3
C4
M5
GE ctrl
GE mgmt
Control - conc sw
C1
C2
C3
C4
M5
CPU -
CPU -
C1
C1
CPU -
C1
CPU -
C1
CPU -
C1
CPU -
C1
(a) 10Gbps Concatenated Control
traffic
31 May, 2006
(b) 1Gbps Direct connection to
Control Core
Stefan Stancu
33
Control network – rack connectivity
GE ctrl/mgmt
GE ctrl
Control - conc sw
C1
C2
C3
C4
C5
GE ctrl
GE mgmt
M6
Control - conc sw
C1
C2
C3
C4
CPU CPU -
C1
D1
C1
D1
CPU CPU -
C1
D1
C1
D1
CPU -
D1
CPU -
C1
D1
C1
FILE - SERVER
FILE - SERVER
C1
C1
(a) Shared
Control and
Management
31 May, 2006
(b) Separate
Stefan Stancu
Control and
Management
34
C5
M6
Control network – rack connectivity
GE data
10GE data
GE ctrl
GE mgmt
GE data
GE ctrl
GE mgmt
DATA - DF Core
D1
D3
D2
Control - conc sw
Mn
C1
C2
C3
C4
C5
CPU D1
C1
D1
D2
D3
Dn
DATA - conc sw
Control - conc sw
DATA - DF Core
C1
Mn
C2
C3
C4
C5
M6
D1
D2
D3
D4
Control - conc sw
M5
C1
C2
C3
C4
C5
D2
CPU -
M6
D1
C1
D2
CPU D1
CPU -
CPU -
FILE - SERVER
C1
C1
D1
C1
CPU -
CPU -
D1
D2
CPU -
CPU -
D1
C1
C1
D1
C1
D1
D1
C1
C1
D1
D2
D3
FILE - SERVER
FILE - SERVER
C1
C1
DC control
31 May, 2006
L2PU
Stefan Stancu
DATA - BE Core
Mn
GE data
GE ctrl
GE mgmt
SFIs
35
M6
Control network – rack connectivity
DATA - BE Core
D1
D3
D2
Control - conc sw
Mn
C1
C2
C3
C4
C5
GE data
GE ctrl
GE mgmt
M6
Control/Data - conc sw
CPU D1
GE ctrl/data
GE mgmt
C1
D2
CPU -
DATA - conc sw
C1
C2
C3
C4
C5
D1
D1
Control - conc sw
M6
D2
D3
D4
M5
C1
C2
C3
C1
D2
CPU -
CPU -
CD1
D1
C1
CPU -
C1
D1
D2
CPU FILE - SERVER
CD1
C1
CPU -
C1
D1
CPU CPU -
CD1
D1
D1
D2
D3
Mn
DATA - SFO out
SFO
31 May, 2006
GE data
GE ctrl
GE mgmt
FILE - SERVER
C1
FILE - SERVER
C1
EF shared
ctrl/data
Stefan Stancu
C1
EF separated
cltr/data
36
C4
C5
M6
Related documents