Fiber Channel Introduction
Fiber Channel Introduction
Copyright 2007 EMC Corporation. All rights reserved. These materials may not be copied without EMC's written
consent. Use, copying, and distribution of any EMC software described in this publication requires an applicable software
license.
THE INFORMATION IN THIS PUBLICATION IS PROVIDED AS IS. EMC CORPORATION MAKES NO
REPRESENTATIONS OR WARRANTIES OF ANY KIND WITH RESPECT TO THE INFORMATION IN THIS
PUBLICATION, AND SPECIFICALLY DISCLAIMS IMPLIED WARRANTIES OF MERCHANTABILITY OR
FITNESS FOR A PARTICULAR PURPOSE.
Celerra, CLARalert, CLARiiON, Connectrix, Dantz, Documentum, EMC, EMC2, HighRoad, Legato, Navisphere,
PowerPath, ResourcePak, SnapView/IP, SRDF, Symmetrix, TimeFinder, VisualSAN, where information lives are
registered trademarks.
Access Logix, AutoAdvice, Automated Resource Manager, AutoSwap, AVALONidm, C-Clip, Celerra Replicator, Centera,
CentraStar, CLARevent, CopyCross, CopyPoint, DatabaseXtender, Direct Matrix, Direct Matrix Architecture, EDM, ELab, EMC Automated Networked Storage, EMC ControlCenter, EMC Developers Program, EMC OnCourse, EMC Proven,
EMC Snap, Enginuity, FarPoint, FLARE, GeoSpan, InfoMover, MirrorView, NetWin, OnAlert, OpenScale, Powerlink,
PowerVolume, RepliCare, SafeLine, SAN Architect, SAN Copy, SAN Manager, SDMS, SnapSure, SnapView,
StorageScope, SupportMate, SymmAPI, SymmEnabler, Symmetrix DMX, Universal Data Tone, VisualSRM are
trademarks of EMC Corporation. All other trademarks used herein are the property of their respective owners.
-1
Course Objectives
Upon completion of this course, you will be able to:
y Identify Fibre Channel:
Architecture
Layers
Topologies
The objectives for this course are shown here. Please take a moment to read them.
-2
The objectives for this lesson are shown here. Please take a moment to read them.
-3
Fibre Channel
y Fibre Channel is a serial data transfer interface intended
for connecting high-speed storage devices to computers
y The high-speed is obtained through the following
processes:
Networking and I/O protocols (such as SCSI commands) are
mapped to Fibre Channel constructs
And encapsulated and transported within
Fibre Channel frames
With this, the high-speed transfer of multiple
protocols over the same physical interface is
possible
Fibre Channel is a serial data transfer interface that operates over copper wire and/or optical fiber at
data rates up to 4 GB/s (gigabits per second) and up to 10 GB/s when used as ISL (E-ports) on
supported switches.
Networking and I/O protocols (such as SCSI commands) are mapped to Fibre Channel constructs, and
then encapsulated and transported within Fibre Channel frames. This process allows high-speed
transfer of multiple protocols over the same physical interface.
Fibre Channel systems are assembled from familiar types of components: adapters, hubs, switches and
storage devices.
Host bus adapters are installed in computers and servers in the same manner as a SCSI host bus adapter
or a network interface card (NIC).
Hubs link individual elements together to form a shared bandwidth loop.
Fibre Channel switches provide full bandwidth connections for highly scalable systems without a
practical limit to the number of connections supported (16 million addresses are possible).
The word fiber indicates the physical media. The word fibre indicates the Fibre Channel protocol and
standards.
-4
Channels
y Channel connections, such as parallel bus and tag,
ESCON, and SCSI provide fixed connections between
host systems and their peripheral devices
y Some characteristics of channel technologies are:
High performance
Low protocol overhead
Static configuration
Short distance (although ESCON is somewhat of an exception)
Connectivity within a single system
The next few slides explain how channels and network technologies help create Fibre Channel
technology.
Traditionally, host computer operating systems communicated with storage devices over channel
connections, such as parallel bus and tag, ESCON, and SCSI. These channel technologies provide
fixed connections between host systems and their peripheral devices.
Static connections are defined to the operating system in advance. Tight integration between the
transmission protocol and physical interface minimize the overhead required to establish
communication and transport large amounts of data to statically defined devices.
Some characteristics of channel technologies are:
y High performance
y Low protocol overhead
y Static configuration
y Short distance (although ESCON is somewhat of an exception)
y Connectivity within a single system
-5
Network
y Network technologies offer more flexibility and distance
capabilities than channel technologies
y Some characteristics of network technologies are:
Low performance
High protocol overhead
Dynamic configuration
Long distance
Connectivity among different systems
Network technologies are more flexible than channel technologies, and provide greater distance
capabilities. Most networks provide connectivity between client or host systems, and carry a variety of
data between the devices. A simple example is a network of desktop PCs within a company.
This type of setup can provide each PC with connectivity to file and print services, server-based
applications, and corporate intranets.
These PCs are networked to provide shared bandwidth and the ability to communicate with many
different systems. This flexibility results in greater protocol overhead and reduced performance.
Some characteristics of network technologies are:
y Low performance
y High protocol overhead
y Dynamic configuration
y Long distance
y Connectivity among different systems
-6
Channels
Networks
(benefits)
(benefits)
Fibre Channel
Fibre Channel captures some benefits of channels and networks. A Fibre Channel fabric is a switched
network, and provides a set of generic, low-level services onto which host channel and network
architectures can be mapped.
Fibre Channel standards define layered communications architecture similar to other networking
environments. Each level of the Fibre Channel protocol stack provides a specific set of functions.
-7
SAN
Storage Area
Networks
Fibre Channel
NAS
Network-Attached
Storage
CAS
Content Addressed
Storage
Type of transport
(FCP, FICON)
IP (iSCSI, FCIP,
iFCP)
IP
Type of data
Block
File
Object,
fixed content
Key requirement
Deterministic
performance
Multi-protocol
sharing
Longevity,
integrity assurance
Typical
applications
OLTP, data
warehousing, ERP
The Fibre Channel can be used as a type of transport in SAN and NAS solutions.
-8
The networks of host and storage devices (called Storage Area Networks, or SANs) are often
connected over Fibre Channel fabrics. A common method to manage the variety of devices on a SAN
is SNMP (Simple Network Management Protocol), and is popular because its widely supported and
can be run out of band (which is advantageous because it does not rely on the Fibre Channel network).
FibreAlliance, an open industry consortium, is defining an SNMP MIB (Management Information
Base) to facilitate SAN management. The MIB is a group of parameters (variables) whose values
define and describe the status of a network and its components. The Fibre Channel Management
Integration (FCMGMT-INT) MIB provides a heterogeneous method to manage multiple devices
across a SAN.
-9
The objectives for this lesson are shown here. Please take a moment to read them.
- 10
Levels
y The Fibre Channel standards defines a Level protocol
stack
Fibre Channel
Level Function
ULP
FC-4
FC-3
FC-2
FC-1
FC-0
- 11
FC-0
y The FC-0 level of FC-PH describes the
Fibre Channel link
Physical Interface
Optical and Electrical Interfaces
Cables, Connectors, etc
The FC-0 level of FC-PH describes the Fibre Channel link. The FC-0 level covers a variety of media
and the associated drivers and receivers capable of operating at a wide range of speeds. The FC-0 level
is designed for maximum flexibility and allows the use of many technologies to meet the widest range
of system requirements.
Each fibre is attached to a transmitter of a Port at one end, and a receiver of another Port at the other
end. When a Fabric is present in the configuration, a fibre may attach to an N_Port and an F_Port.
Patch panels or portions of the active Fabric may function as repeaters, concentrators or fibre
converters.
- 12
Connectors
y The LC assembly design
incorporates an RJ-style latch
in a connector body with half
the footprint of a conventional
SC
y The SC connector is the
standard connector for fiber
optic cables. It is a push-pull
connector
The connectors, cables, and host Bus adapters (HBA) characteristics that are part of the FC-0 level are
reviewed next.
The LC assembly design incorporates an RJ-style latch in a connector body, with half the footprint of a
conventional SC. High-precision, 1.25 mm ceramic ferrules and engineered composites provide a
durable package with consistent, repeatable performance.
The SC connector is the standard connector for fiber optic cables. It is a push-pull connector and is
favored over the ST connector (commonly used in patch panels). If the cable is pulled, the tip of the
cable in the connector does not move out, which would result in loss of signal quality.
- 13
Multi-Mode Cable
y Multi-mode transmitters send multiple short-wavelength
signals through the same fiber
y The angle of entry is high and the signals tend to cancel
each other out
y Multi-mode fiber has a larger diameter (62.5 or 50
microns) core than single-mode fiber
The multi-mode cable is dominant for distances of 500 meters or less. Multi-mode has an inner
diameter of 62.5 or 50 microns, and allows light to enter the cable in multiple modes, including
straight and at different angles. The multiple light beams tend to lose shape as they move down the
cable. This loss of shape is called modal dispersion and limits the distance.
- 14
Single-Mode Cable
y Single-mode transmitters send one longer-wavelength
signal down a much thinner-cored fiber
y The angle of entry is low (less bouncing) and there is little
to muddy the signal, hence the much greater distances
y Single mode cable has an inner diameter of 9 microns
The single mode cable is used for long distance cable runs, and is limited only by the power of the
laser at the transmitter and the receiver sensitivity. Single mode cable has an inner diameter of 9
microns and is always used with a long wave laser, which limits the effects of modal dispersion.
Therefore, with single mode cables, the distance is greatly increased.
- 15
An HBA is an I/O adapter that sits between the host computer's bus and Fibre Channel loop and
manages the transfer of information between the two channels. In order to minimize the impact on the
host processor performance, the host bus adapter performs many low-level interface functions
automatically or with minimal processor involvement.
In simple terms, a host bus adapter (HBA) provides I/O processing and physical connectivity between
a server and storage. The storage may be attached using a variety of direct attached or storage
networking technologies, including Fibre Channel, iSCSI, VI/IP, FICON, or SCSI. Host bus adapters
provide critical server CPU off-load that frees servers to perform application processing. As the only
part of a storage area network that resides in a server, HBAs also provide a critical link between the
SAN and operating system and application software. In this role, the HBA enables a range of highavailability and storage management capabilities, including load balancing, fail-over, SAN
administration, and storage management.
- 16
FC-1
y Defines the method used to Encode data
prior to transmission and decode the data
upon reception
y A benefit of using 8b/10b encoding is that it defines a
number of Special Characters
Whenever the encoder creates a character with more ones than zeros, it remembers this by setting a
single-bit variable, called the Current Running Disparity (CRD), to positive. Whenever the encoder
creates a character with more zeros than ones, it sets the CRD to negative.
The CRD is fed back to the encoder to select the appropriate encoding in order to balance the number
of ones and zero bits transmitted.
As a series of characters are processed, the output alternates between positive and negative disparity.
- 17
Encoding Process
y The encoding process transforms 8-bit input characters
into 10-bit transmission characters having the desired
attributes
The encoding process transforms 8-bit input characters into 10-bit transmission characters having the
desired attributes. Not all of the possible 10-bit patterns are used. 10 bits allow 1024 different
combinations.
To prevent excessive DC (direct current) components and run length problems, only characters
containing 6 ones and 4 zeros, 5 ones and 5 zeros, 4 ones and 6 zeros are used. Any other weighting of
bits is invalid.
y 5 ones and 5 zeros are considered to have neutral disparity
y 6 ones and 4 zeros are considered to have positive disparity
y 4 ones and 6 zeros are considered to have negative disparity
y All possible 8-bit characters have 2 possible 10-bit encodings
This is the same method used to transmit a data stream in ESCON. The purpose is to break up the data
stream in the serial environment and, at the same time, allow control characters to be embedded into
the steam to speed-up communications.
- 18
Ordered Sets
y Two types of Transmission Characters (Data and
Special) are defined
IDLE
Data Field
Header
Payload
Optional Optional
(information being
Header Header
Transported
Fill
Bytes
Certain combinations of Transmission Characters, referred to as Ordered Sets, are designated to have
special meaning. Ordered Sets are used to identify frame boundaries, transmit primitive function
requests, and maintain proper link transmission characteristics during periods of inactivity.
- 19
FC-2
y Defines the structure of the Fibre Channel
frame
y The transported data is transparent to
FC-2 and visible to FC-3 and above
The FC-2 level serves as the transport mechanism of the Fibre Channel. The transported data is
transparent to FC-2 and visible to FC-3 and above.
- 20
Frames
y Frames are the basic building blocks of a Fibre Channel
connection
Header
Optional
Header
Optional
Header
Payload (information
Fill Bytes
being Transported
End-of-Frame
Data Field
CRC
Start-of-Frame
Frames are the basic building blocks of a Fibre Channel connection. The frames contain the
information to transmit, the address of the source and destination ports, and link control information.
All information in Fibre Channel is passed in frames. The maximum amount of data carried in a frame
is 2112 bytes; the total frame size is 2148 bytes. The general structure of a Frame is specific.
- 21
Header Contents
CRC
Data Field
Header
Optional
Header
R_CTL
Optional
Header
Payload (information
Fill Bytes
being Transported
End-of-Frame
Start-of-Frame
CS_CTL
TYPE
SEQ_ID
DF_CTL
OX.ID
DF_CTL
OX_ID
The header contains the Source and Destination Addresses, which allow the frame to be routed to the
correct port. The Type field interpretation is dependent on whether the frame is a link control or Fibre
Channel data frame. For example, if the frame is a data frame, a 08 in the type field indicates SCSI
FCP information in the Data field.
- 22
- 23
N_Port
N_Port
Data Frame
Data Frame
R_RDY
R_RDY
The FC-2 provides flow control for buffer management. When nodes initialize on the fabric, they agree
on operational parameters such as the number of buffers available (Buffer Credits). Transmitting nodes
can continue to transmit as long as there are buffer credits. R_RDYs are sent to a transmitting node as
buffers are cleared. An FC switch also performs buffering and flow control internally and externally.
- 24
Classes of Service
Classes of Service
Functions
Class 1
Class 2
Class 3
Class 4
Class 5
1:1
1 to Many
Optional
Multicast support
Broadcast support
Stacked connect Request
Optional
Optional
Dedicated Simplex
Camp-on
Buffered Class
Unidirectional Connection
End-to-End flow control
Optional
Optional
Optional
Optional
Optional
Optional
Optional
Optional
Optional
3
3
3
3
3
3
3
Optional
Optional
1:1
Classes of Service are different types of topology independent services provided by the Fabric and
used by the communicating N_Ports destination. The allocation and retention method between the
N_Ports and level of delivery integrity required for an application distinguish classes of service. If the
Fabric is not present, the service is provided as a special case of point-to-point. Fabrics and N_Ports
are not required to support all Classes of service.
The Classes of Service table is shown here. Please take a moment to review.
- 25
Source
FC Initiator:
FC Responder:
HBA
Symmetrix FA
Or
CLARiiON SP Ports
Fibre Channel addresses are used to designate the source and destination of frames in the Fibre
Channel network. The Fibre Channel address field is 24 bits/3 bytes in length. Unlike Ethernet, these
addresses are not burned in, but are assigned when the node enters the loop or is connected to the
switch. There are reserved addresses, which are used for services rather than interface addresses.
- 26
Addressing Layout
y Physical Address is switch specific and dynamically
generated in the Fabric Login (FLOGI)
y Each N_Port has a Fabric-unique identifier, with the
following layout
FC-SW
Switch
Port in Switch
FC-AL
AL_PA
02 1F 00
2007 EMC Corporation. All rights reserved.
The Physical Address is switch-specific and dynamically generated in the Fabric Login (FLOGI).
The Fabric Logon is discussed later in this lesson.
Address identifiers are three bytes in length. The Frame Header contains two three-byte fields for
address identifiers, the D_ID and S_ID fields. Each N_Port has a fabric-unique identifier, the N_Port
Identifier, by which it is known. The source and destination N_Port Identifiers and alias address
identifiers are used to route frames within the fabric.
- 27
Addressing layout
1. Domain ID: Identifies the source or target switch inside
the Fabric
2. Switch Port: Identifies the source or target port in the
switch
3. AL_PA - Used in Private Loop environment, identifying
the NL_Port (node loop port)
FC-SW
Switch
PortSwitch
in Switch
FC-AL
AL_PA
02 1F 12
2007 EMC Corporation. All rights reserved.
The most significant 8 bits of the Fibre Channel address contain the Domain ID, which basically
identifies the switch.
In the Fabric environment, this allows frames to be routed between switches.
The middle 8 bits contain the area address, which is implemented as the port address within the switch.
In the Fabric environment, this allows frames to be routed between switches.
In the Private Loop environment, the Domain and Area fields contain zeros, and the port field contains
the AL_PA (Arbitrated Loop Physical Address) for the NL_Port.
Note:
With a McDATA switch, the third byte is always 13.
- 28
Company OUI
Vendor Assigned
12 bits
23 bits
24 bits
1 0 0 0 0 0 0 0 C 9 2 0 DC 4 0
Example: Emulex HBAs World Wide Name
A World Wide Name, or WWN, is a 64-bit address used in Fibre Channel networks to uniquely
identify each element in the network.
Assigned to a host bus adapter (HBA) or switch port by the vendor at the time of manufacture, it is
similar to the MAC address in an Ethernet network.
There are two designations of WWN; World Wide Port Name and World Wide Node Name. Both are
globally unique 64-bit identifiers. The difference lies in where each value is physically assigned.
For example, a server may have dual HBAs installed, thus having multiple ports or connections to a
SAN. A WWPN is assigned to each physical port.
The WWNN represents the entire server, which can be referred to as the node or node process, and is
derived from one of the WWPNs. EMC uses the WWPN for all configurations
Fibre Channel specifications allow for multiple formats of the World Wide Name. The example shown
is that of the IEEE Registered Name Format.
NAA is the Name Assignment Authority, which assigns the 24 bits (IEEE Company ID) to the specific
vendor (i.e. EMC).
Values for World Wide Name formats are based on the IEEE company_id. More information on these
formats can be found at www.standards.ieee.org.
- 29
5 0 0 6 0 4 8 2E8 9 1 2B9 0
Example: Symmetrix FA World Wide Name
5 0 0 6 0 1 6 0 0 0 6 0 0 1B2
Example: CLARiiON SP World Wide Name
The WWN for Symmetrix FAs and Clariion SPs ports are soft-assigned, so they remain unchanged
after a component replacement.
- 30
5 0 0 6 0 4 8 2E8 9 1 2B
0101
0000
1000
0010
9
0
Port
0
1
1110
Side FA 4 bits
011110
195183790
15bA
y Symmetrix DMX
5 0 0 6 0 4 8 A CC C 8 3 2
0101
0000
1000
A
1
Port
0
0001
Side FA 4 bits
100001
2CB
Because the WWN of the FA is dependent on the Symmetrix Serial Number and the Slot rather than being burned in, the
WWN stays constant if the FA fails and must be replaced. To determine the WWN of an FA depends on the version of the
Symmetrix being used. The original calculation was that the last 6 bits of the WWN were used. Bit position 6 specified the
port (0 or 1), Bit position 5 the processor ( 0-A/1-B ) and Bit position 1-4 the slot (add 1 to get the adapter).
On the DMX series, the ports are laid out as follows, from the top down:
DB - processor 4, port 1
DA - processor 4, port 0
CB - processor 3, port 1
CA - processor 3, port 0
BB - processor 2, port 1
BA - processor 2, port 0
AB - processor 1, port 1
AA - processor 1, port 0
The new calculation had to be both backward compatible with previous Symmetrix versions, and at the same time allow for
more addresses. The new calculation borrows a bit from the serial number. It was determined that we do not need 30 bits
for a serial number, 29 gives us all the range we need. So, bit position 30 is now dubbed a "half bit" The half bit
determines which pair of processor we are working with. It is set to 0 for processors A & B, and 1 for C & D. Then, as
above, we make use of the processor" bit. This is used along with the half bit to determine exactly what processor. It
breaks down like this:
Half bit 0 / Processor bit 0 A
Half bit 0 / Processor bit 1 B
Half bit 1 / Processor bit 0 C
Half bit 1 / Processor bit 1 - D
Then, like the original calculation, we use the Port bit to determine the 0/A port or the 1/B port.
- 31
5 0 0 6 0 1 6 0 0 0 6 0 0 1 B2 5 0
0101
0000
0110
0000
Port
0000
0000
0110
0101
0000
CLARiiON - CX600
WWN seed - 00:60:01:b2
Therefore, the resulting WWNN is 50:06:01:60:80:60:01:b2
The resulting WWPNs are:
Storage Processor A
Port 0 - 50:06:01:60:00:60:01:b2
Port 1 - 50:06:01:61:00:60:01:b2
Port 2 - 50:06:01:62:00:60:01:b2
Port 3 - 50:06:01:63:00:60:01:b2
Note:
On the FC4500, we have the only real exception as the FC ports are in fact connected to a hub and
share their WWPN. To get around this, only one port from each SP would be connected to the Fabric.
Two ports existed per SP to allow for a dual cluster direct attach.
- 32
Frames Routing
y When N_Ports are
connected to F_Ports,
Fabric
the Fabric Login
begins, associating
Sw 01
physical address and
World Wide Name
Port 07
Target
Source
Physical cable connected
ISL
Port 05
FA WWN:
50060482E8912B90
HBA WWN:
10000000C920DC40
Sw 02
Introduction to Fibre Channel - 33
When N_Ports are connected to F_Ports, the Fabric Login begins associating the physical address and
worldwide name.
- 33
Port 05
Target
FA WWN:
50060482E8912B90
HBA WWN:
10000000C920DC40
Sw 02
Introduction to Fibre Channel - 34
The WWN and Physical Address association is done by the switch and stored using internal tables.
The frames are routed using their Physical Address, as happens in an Arbitrated Loop environment
using just the Arbitrated Loop Physical Address.
- 34
Fabric
Target
Sw 01
Port 07
Physical cable connected
Source
ISL
WWN -> Switch/Port/AL_PA
Port 05
0600
10000000C920DC40 -> 02
Port 06
FA WWN:
50060482E8912B90
HBA WWN:
10000000C920DC40
Sw 02
Introduction to Fibre Channel - 35
When nodes generate frames, they are routed by their addresses, not their WWNs. However, tables can
be built which can associate WWNs to the destination addresses. A World Wide Name is a 64 bit value
(16 characters).
- 35
Fabric
Process
Process aa
F_Port
N_Port
F_Port
N_Port
Fabric Login
11
Accept
F_Port
F_Port
Port Login
Fabric Login
Accept
N_Port
N_Port
22
Process
Process xx
Process
Process bb
Accept
Process
Process yy
Process
Process cc
Process Login
Accept
Process
Process zz
There are three types of login supported in Fibre Channel: Fabric, Port and Process. All node ports
must attempt to log in with the Fabric. This is typically done right after the link or the Loop is
initialized. It consists of the node port transmitting a Fabric Login (FLOGI) frame to the well-known
Fabric address hex'FFFFFE'. The normal response is an Accept (ACC) frame from the Fabric back to
the node port. Fabric Login accomplishes the following things:
y It determines the presence or absence of a Fabric.
y If a Fabric is present, it provides a specific set of operating characteristics associated with the entire
Fabric, including which Classes of service are supported.
y If a Fabric is present, it will optionally assign or will confirm the native N_Port Identifier of the
N_Port that initiated the Login.
y If a Fabric is present, it initializes the buffer-to-buffer credit.
Before a node port can communicate with another node port, it must first perform N_Port Login with
that node port. Similar to Fabric Login, the N_Port transmits a PLOGI frame to the destination node
port. Again, the normal response is an ACC frame. Port Login accomplishes the following things:
y It provides a specific set of operating characteristics associated with the destination N_Port,
including Class of service.
y With Class 3 services, buffer-to-buffer credit is initialized.
PRLI is an acronym for Process login. Process logins establish sessions between related processes on a
source and target N_Port. The processes are typically FC-4 layer applications.
- 36
Fabric Login
y Fabric Login (FLOGI) is used by an N_Port to determine
the presence of a Fabric, then exchange service
parameters
y N_Port performs a login to address FFFFFE (F_Port
Server) using a source address of 000000
y The Fabric Login service returns a frame which assigns
an address to the port (24 bits)
When a Fibre Channel device is attached to a fabric, it begins a fabric login (FLOGI). FLOGI is an
extended link service command that sets up a session between two participants. With FLOGI, a session
is created between an N_Port or NL_Port and an F_Port or FL_Port. An N_Port sends a FLOGI frame
that contains its Node Name, N_Port Name, and service parameters to a well-known address.
- 37
FLOGI
HBA
N_Port
N_Port
Fabric Server
FA
N_Port
N_Port
(xFF FF FE)
Registration Request
Registration Request
Directory Server
(NameServer)
Query Request
(xFF FF FC)
Port Identifier (SID)
Query Request
When the N_Port logs in, it uses a 24-bit port address of 0. Because of this, the fabric is allowed to
assign the appropriate port address to that device, based on the Domain-Area-Port address format. The
newly assigned address is contained in the ACC response frame.
- 38
An NL_Port first opens the destination AL_PA 0 before issuing the FLOGI request. In both cases, the
switch accepts the login and returns an accept (ACC) frame to the sender. If some of the service
parameters requested by the N_Port or NL_Port are not supported, the switch sets the appropriate bits
in the ACC frame to indicate this.
When the NL_Port logs in, a similar process starts, except that the least significant byte is used to
assign AL_PA and the upper two bytes constitute a fabric loop identifier. Before an NL_Port logs in, it
goes through the LIP on the loop, which is started by the FL_Port. From this process, an AL_PA has
already been derived. The switch then decides if it will accept this AL_PA or not. If not, a new AL_PA
is assigned to the NL_Port, which then causes the start of another LIP. This ensures that the switch
assigned AL_PA does not conflict with any previously selected AL_PAs on the loop.
- 39
y Once a port has logged in, it can query the Name Service
database for information about all other logged in ports
After the node gets its Fabric address from FLOGI, it must register with the SNS with port login
(PLOGI) on address 0xFFFFFC. The device may register values for all or just some database objects,
but the most useful are:
y Port Identifier (ID) = S_ID
y Port Name(PN) = WWN of the N_Port
y Classes of Service (CS) = Class 3 currently
y FC-4 Types Supported (FT) = SCSI-3
y Port Type (PT) = N_Port
Once a port has logged in, it can query the Name Service database for information about all other
logged in ports.
- 40
When an N_Port logs in to another N_Port, a table can be built which will keep track of the WWN of
the logged in port along with its Fibre Channel address. For example, when the NT and SUNs login to
the Symmetrix, a table is created. This speeds data transfers and node-to-node communications.
Some Host types (for example, HP-UX) may not maintain a connection while idle and would need to
re-login.
Port login is used to establish a session between two nodes, swapping service parameters and making
themselves known to each other.
Port logins exchange information such as:
y host address (SID)
y frame size (receive buffer size)
y flow control & version information (TOVs)
y port name (WWPN)
- 41
Process login is used to set up the environment between related processes on node ports.
The Fibre Channel Protocol for SCSI-3 standard specifies that each SCSI task corresponds to Fibre
Channel exchange. A Fibre Channel exchange consists of a single SCSI command or a group of linked
SCSI commands. The FCP mapping of SCSI-3 to Fibre Channel defines four information sets that are
transferred between SCSI initiator and target. The information sets are modeled after the SCSI-3
architecture defined protocols services.
y FCP_CMND: Corresponds to Send SCSI Command protocol service.
y FCP_XFER_RDY: Transports the offset and request byte count objects of the Send Data-In and
Receive Data-Out protocol services.
y FCP_DATA: Transports the data object of the Send Data-In and Receive Data-Out protocol
services.
y FCP_RSP: Corresponds to the Send Command Complete protocol service.
- 42
FC-3
y FC-3 was put into Fibre Channel as a
placeholder
y In concept, FC-4 would pass requests
to FC-3 that would perform the desired service and then
pass onto FC-2
The FC-3 was put into Fibre Channel as a placeholder. In concept, FC-4 would pass requests to FC-3
that would perform the desired service and then pass onto FC-2.
Some things that have been identified as probably fitting into FC-3 are:
y Data Striping
y Multipathing
y Mirroring
y RAID
y Data encryption
y Data compression
y Data Translation
- 43
FC-4
y The FC-4 level of Fibre Channel is
designed to hand off to another protocol,
such as SCSI
y Fundamentally, the commands at FC-4 for SCSI allow
the SCSI initiator and target to communicate over Fibre
Channel
The FC-4 level consists of several standards documents that describe how different upper-level
protocols use the transport services provided by levels FC-2, FC-1 and FC-0. The purpose of an FC-4
protocol mapping is to make a logical connection between the ULPs and Fibre Channel's transport
facilities.
- 44
- 45
The objectives for this lesson are shown here. Please take a moment to read them.
- 46
Logical
Topology
Physical
Topology
Windows
Server
Fibre Channel
Switch
Storage
Sun
Server
The Fibre Channel environment consists of a physical and logical topology. The physical topology
describes the physical interconnects among devices like servers, storage, and switches. The logical
topology describes the logical paths established between the operating system device names and their
associated storage ports and volumes.
Logical topologies in the EMC/Fibre Channel switch environment can generally be described in terms
of fan-in (into the EMC storage array) and fan-out (out of the EMC storage array).
- 47
The ANSI Fibre Channel Standards define three topologies: Direct Connect, Arbitrated loop, and
Switched fabric.
This slide displays a Fibre Channel topology where two devices are directly connected to each other.
EMC specifies a 2 node Fibre Channel Arbitrated Loop connection called Direct Connect. This
Topology is the same as the standard Arbitrated Loop, except there are two nodes in the Loop and the
Arbitrated Loop Switch is not used.
- 48
Fibre Channel
Hub
Arbitrated Loop (FC-AL) is a daisy-chain connecting up to 126 devices in a loop configuration over
attachment points called L_Ports (loop ports). FC-AL is a low-cost connectivity solution because it
does not require switches.
FC-AL is a good choice for small to medium-sized configurations, and provides a growth path by
allowing connection of a loop to a switched Fabric.
Efficiency and connectivity is enhanced by incorporating one or more hubs into the loop. Routing
traffic through a hub on each leg of a loop eliminates the loss of the entire loop, as happens in a
hubless loop.
- 49
Fibre Channel
Switch
A Switched Fabric is one or more Fibre Channel switches that connect multiple devices. Rather than
travel around an entire loop, frames are routed between source and destination by the fabric. This
topology allows the most number of connectivity, with a theoretical 16 million devices per fabric.
- 50
Fibre Channel hubs are used with Fibre Channel Arbitrated Loops to increase server and storage
connectivity. This way, multiple servers can access multiple storage devices simultaneously.
A hub can only have one Full Duplex connection between port pairs at any one time. Although every
port is capable of full speed operation, only one pair of ports can be active at any one time. This is
because only two nodes in a FC-AL environment can be active at any one time and they must
communicate with each other.
EMC Symmetrix storage systems are qualified with hubs in limited configurations with HP-UX, Sun
Solaris, Windows NT, Windows 2000, Windows 2003, and Siemens servers.
- 51
Switches
y A Fibre Channel switch is a device that routes data
between host bus adapters and fibre adapters on storage
systems
y A Fabric is a single switch or multiple switches that
interconnect various N_Ports, and is capable of routing
frames by using only the D_ID information in a Fibre
Channel frame header
y Larger Fabric are built by connecting
multiple switches through E_Ports
Switched Fabrics offer strong advantages over hubs, in both performance and manageability. For
example, a hub can only have one full-duplex connection between port pairs at one time. Although
every port in a 10-port hub is capable of full-speed operation, only one pair of ports can be active at
any one time. A switched Fabric, however, can have every port pair running full duplex simultaneously
at 200 megabytes per second. In a switched topology, aggregate bandwidth of 200 megabytes per
second times the Fabric port count is possible.
The aggregate bandwidth of a hub is fixed, but the aggregate bandwidth of a Fabric scales as additional
port pairs are added. This scaling capability is an important management feature, because it allows the
administrator of the enterprise storage network to reallocate the bandwidth of the network among
N_Ports without moving cables and reconfiguring the environment.
A second management benefit of switched Fabrics over hubs is a Fabric-based service called the name
server, which maintains a table of all logged-in devices.
Used by the N_Ports for device discovery, the table is maintained by the Name Server during Fabric
reconfigurations.
- 52
Directors are
y Availability features
y Highest availability
y High performance
y Highest performance
y Web-based management
features
y Web-based and/or
console-based
management features
In a SAN, implementation is necessary to select the director, switch or hybrid technology solution.
The switch/director decision is often financial. Technically, directors are preferable to switches. If we
consider the differences between them:
y Directors are easier to manage because the relationship between ports is equal and, therefore, it
does not need to be considered in performance planning. A single high-availability Fabric is
possible using directors, though switches require two Fabrics.
y Directors have scaled failure localization. This means the GBIC, power supply, etc. can each be
replaced without bringing down the entire device, limiting risk of catastrophic failure.
y Switches offer some localization, at the GBIC or power supply, but in general, a physical failure
results in whole device replacement.
Directors are considered to be more highly available than switches. To be fully redundant, there is a
requirement for multiple connections to more than one network, with complete routes to all devices
throughout those networks. This improves resiliency of the Fabric, allowing the SAN to maintain its
functionality in the even of failures. Because fault isolation in a director is at a more granular
component level and the components are already dual redundant, directors have the least adverse
impact on availability.
- 53
y Data-center deployment
MDS-9509-V2
y Maximum scalability
y Large Fabrics
ED-140M ED-10000M
COST
AP-7420B
ED-48000B
MDS-9506-V2
MDS-9216i/A
MP-2640M
DS-4900B
MDS-9140
MDS-9120
MDS-9020
MP-1620M
DS-4700M
Connectrix Switches
DS-4100B
DS-4400M
DS-220B
SERVICE LEVEL
2007 EMC Corporation. All rights reserved.
Only EMC offers a complete range of SAN productsfrom Connectrix directors for data-center
deployments to Connectrix switches for data-center and departmental deployment.
The Connectrix Family gives customers:
y The widest range of SAN service levels
y More choice
y Additional Fibre Channel functionality
y A pathway to IP SANs
y An additional platform for future network-hosted applications
- 54
DESCRIPTION
N_Port
NL_Port
F_Port
FL_Port
E_Port
G_Port
U_Port
In an environment using Host Bus Adapters and Symmetrix FC Directors, ports are configured as
either N-Ports or NL-Ports:
y N_Port: Node port, a port at the end of a point-to-point link
y NL_Port: Node Loop port, a port which supports the arbitrated loop topology
Fibre Channel Switch ports are also configured for specific applications:
y F_Port: Fabric port, the access point of the Fabric which connects to a N_Port
y FL_Port: Fabric Loop port, a Fabric port which connects to a NL_Port
y E_Port: Expansion port on a switch. Links multiple switches
y G_Port: General port, a Connectix McDATA switch port with the ability to function as either an
F_Port or an E_port
y U_Port: Universal port, Connectrix B series equivalent to a G_Port
- 55
Fabric
N_Port
N_Port
N_Port
N_Port
N_Port
N_Port
F_Port
F_Port
F_Port
F_Port
F_Port
F_Port
Fabric
E_Port
E_Port
E_Port
E_Port
F_Port
F_Port
ISL
G_Port
G_Port
N_Port
N_Port
- 56
Zoning
y Partitions a Fibre Channel switched Fabric into subsets
of logical devices
y Zones contain a set of members that are permitted to
access each other
y A member can be identified by its Source ID (SID), its
World Wide Name (WWN), or a combination of both
Zoning is used to partition a Fibre Channel switched Fabric into subsets of logical devices. Each zone
contains a set of members that are permitted to access each other. Members can be switch ports, HBAs,
or storage ports. When zoning is enabled, members in the same zone can see and communicate with
each other, but members in separate zones cannot. Ports and devices distributed across multiple
switches in a Fabric may be grouped into the same zone.
A zone member can be identified by its Source ID (SID), its World Wide Name (WWN), or a
combination of both. Zones can be created using each of these forms of member identification. Thus,
there are essentially three types of zoning: hard zoning, soft zoning, and mixed zoning.
- 57
What is Zoning?
10000000C920C4E4
50060482B8912B8E
10000000C920C321
50060482B8912B9E
50060482B8912B8F
200000E069000CFE
50060482B8912B9F
200000E0690014B8
200000E06900156D
500601684003491C
- 58
Under single-HBA zoning, each HBA is configured with its own zone. The members of the zone
consist of the HBA and one or more storage ports, such as a Symmetrix Fibre Adapter port, with
volumes the HBA will use.
This zoning practice provides a fast, efficient, and reliable means to control the HBA discovery/login
process. Without zoning, the HBA attempts to log in to all ports on the Fabric during discovery and
during the HBAs response to a state change notification. With single-HBA zoning, the time and Fibre
Channel bandwidth required to process discovery and the state change notification are minimized.
Here are two very good reasons for Single HBA Zoning: First, it cuts down on the reset time for any
change made in the state of the Fabric. And secondly, only the nodes within the same zone are forced
to log back into the Fabric after a Registered State Change Notification.
When a nodes state has changed in a Fabric, it has to perform the Fabric Login process again before
resuming normal communication with the other nodes it is zoned with. If there is only one SCSI
Initiator in the zone, then the amount of disrupted communication is reduced.
If a zone has two HBAs and one of them had a state change, then BOTH are forced to log in again,
causing disruption to the other HBA that did not have any change in its Fabric state. Performance can
be severely impacted by this.
- 59
FC
Switch
WWPN = 10:00:00:60:69:40:DD:A1
Domain ID = 25
Port = 3
Host
Fabric
FC
Switch
Storage
WWPN = 50:06:04:82:E8:91:2B:9E
In general, zoning can be divided into two categories: Port zoning and WWN zoning.
- 60
WWN Zoning
WWPN = 10:00:00:60:69:40:8E:41
Domain ID = 21
Port = 1
WWPN = 10:00:00:00:C9:20:DC:40
FC
Switch
WWPN = 10:00:00:60:69:40:DD:A1
Domain ID = 25
Port = 3
Host
Fabric
FC
Switch
Storage
WWPN = 50:06:04:82:E8:91:2B:9E
WWN zoning creates zone sets using the WWNs of the attached nodes. This allows you to move nodes
from one switch port to another without having to change the zone configuration. The zone set
becomes independent of individual switch ports.
- 61
Port Zoning
WWPN = 10:00:00:60:69:40:8E:41
Domain ID = 21
Port = 1
WWPN = 10:00:00:00:C9:20:DC:40
FC
Switch
WWPN = 10:00:00:60:69:40:DD:A1
Domain ID = 25
Port = 3
Host
Fabric
FC
Switch
Storage
WWPN = 50:06:04:82:E8:91:2B:9E
Port zoning is based on physical topology and uses the domain ID and port number of the switch. This
method allows you to change the HBA without changing any zoning information. However, it does not
give you the flexibility to physically move attached nodes between switch ports without having to
redefine the port number in the zone set.
- 62
Zoning Components
Member #1
Zone #1
Member #2
Active Zone Set
Member #3
Zone #2
Member #4
Member #5
Zone #3
Members are attached adapters which can be included in a zone. They include HBAs and Storage Fibre
adapter ports. A Storage Fibre adapter is a front end port on a storage system. Examples include the
Symmetrix FA, CLARiiON SP, and Compaq adapter ports. A zone contains a set of members that can
access each other. A member can be a Storage Fibre or Host Bus adapter that is logged into a switch
device. Devices spread throughout multiple switches in a multi-switch fabric may be grouped into the
same zone. Multiple zones can be assigned membership in zone set. Zones are built and then included
into a specific zone set based on the customers requirement to activate multiple zones simultaneously.
Zone sets are then activated to simultaneously activate the zones in the set.
Zoning a hosts HBA with a Symmetrix or CLARiiON Storage Fibre adapter port allows
communication between those members. Volumes visible on that storage port are now available to the
host. Members of a zone can see each other; members in different zones cannot.
A zone set is a group of zones that you can activate or deactivate as a single entity in a single or multiunit fabric. Only one zone set can be active at one time per fabric.
The active zone set is a single zone set that is currently enabled. When a zone set is active, all zones
that are members of that zone set are active. Only one zone set can be active for the fabric at one time.
- 63
VSAN
y VSANs are logical SANs providing isolation among
physically connected devices
VSANs are logical SANs over a common physical fabric allowing logical segments in the network.
Multiple independent SANs over a common physical infrastructure provide isolation among devices
that are physically connected.
- 64
y SONET
Long-distance solutions (1000s km)
Smaller sites feeding into WDM
y IP
Native IP for simplicity
Bridged IP when WDM and SONET not available
Reuse part of existing infrastructure
WDM
y Metro region (100s km)
y Shared infrastructure for many applications
y High bandwidth needs
y Limitation: Availability of dark fiber; Can be high cost
SONET
y Long-distance solutions (1000s km)
y Smaller sites feeding into WDM
y Limitation: Cost varies (IP can be less expensive); Can be difficult to provision
IP
y Native IP for simplicity
y Bridged IP when WDM and SONET not available
y Reuse part of existing infrastructure
y Limitation: Need a well-designed network; Limited interoperability between IP bridges
- 65
SAN Routing is a new technology development of the FC protocol. A study group within T11 is
dedicated to the standards development to accommodate the solution.
- 66
SAN Virtualization uses the storage network to redirect IOs on-the-fly, thereby allowing free space
that can be used from any array in the network and makes data migration a real-time activity.
- 67
iSCSI
y Native TCP/IP protocol
y An IP-based protocol for
establishing and managing
connections between IPbased storage devices,
hosts, and clients
y No Fibre Channel content
Today, there are three block storage over IP approaches: iSCSI, FCIP, and iFCP.
iSCSI is a native IP-based protocol for establishing and managing Connections between IP-based
storage devices, hosts, and clients. There is no Fibre Channel content.
- 68
FCIP
y TCP/IP based tunneling/encapsulating
protocol for connecting/extending Fibre
Channel SANs
y More IP content, little Fibre Channel
content
- 69
iFCP
y Gateway to gateway protocol for FC over IP
y Mapping natively in IP across Fibre Channel
and IP
y An IP-based tunneling protocol for
interconnecting Fibre Channel devices
together in place of Fibre Channel switches
y More Fibre Channel content; when iFCP
creates the IP packets, it inserts information
that is readable by network devices and
routable within the IP network.
y IFCP wraps Fibre Channel data in IP packets
but maps IP addresses to individual Fibre
Channel devices
2007 EMC Corporation. All rights reserved.
iFCP is a gateway-to-gateway protocol for FC over IP and maps natively in IP across Fibre Channel
and IP. It is an IP-based tunneling protocol for interconnecting Fibre Channel devices together in place
of Fibre Channel switches.
When iFCP creates IP packets, it inserts information that is readable by network devices and routable
within the IP network. iFCP wraps Fibre Channel data in IP packets, but maps IP addresses to
individual Fibre Channel devices.
- 70
LAN
Native iSCSI allows for all communications using Ethernet. Initiators may be directly attached to
iSCSI Targets or may be connected using standard Ethernet routers and switches.
- 71
SAN
Bridging architectures allow for the Initiators to exist in an Ethernet environment while the storage
remains in a Fibre Channel SAN.
- 72
SAN
WAN
Extension architectures are most often used to provide connectivity across large distances. Either
FCIP or iFCP bring the long distance benefits of IP to Fibre Channel.
- 73
Course Summary
Key points covered in this course:
y Fibre Channel Architecture
y Fibre Channel Layers
y Fibre Channel Topologies
y SAN concepts
These are the key points covered in this training. Please take a moment to review them.
This concludes the training.
- 74