Computer Networks Lecture wired networks LAN MAN SAN

Computer Networks Lecture wired networks LAN, MAN, SAN Materials Laboratories: CORE NET ACCESS © Michael Turek

SAN

SAN - Storage Area Network Mainly uses medium fiber (Fiber Channel), but also include copper cables or coaxial It has some topology (point-to-point, so. Loop arbitration, star) Seen as an extension of the bus network storage Typical implementation: to foster communication between servers and storage. In this application - Scalable Allows the use of multiple connections between the same devices to increase the capacity (no crashes when looping)

SAN - Fiber Channel High-speed serial interface for communication between servers and storage Layered model Fiber Channel: And physical layer signaling: FC-0 - characteristics of the medium FC-1 - transmission protocol FC-2 - framing and signaling protocol Higher layer 3 -4 FC-3 - shared services (base) FC-4 - the layer mapping of higher layer protocols (network: IP, ATM)

SAN - Fiber Channel Addressing framing FC-2: It is based on 24 -bit identifiers FC ports (equivalent physical interfaces) Services common FC-3: Hunt Groups - to run redundant connections between devices in a SAN, based on the use of the alias address via Striping - aggregation links to increase bandwidth Multicast - directing simultaneous transmission to multiple receiving devices

Types of Topology FC (I) Point-to-Point It is possible to occur when the full duplex connection with the passage between the two ports in the N-type servers Loop arbitration Applied when connected to the device in simplex mode (TX-> RX) It contains up to 127 devices in a loop through their ports NL (information runs through the loop in one direction). The loop can be managed by a device equipped with ports FL then the option for physical topology loops is the use of arbitration hub FC (FL ports are for the bypass function is not connected to these ports NL)

Types of Topology FC (II) Star Called for FC topology or ODF FC Switch Fabric. In this topology there is a switch (Switch Fabric) Having ports F, which can be connected to ports of the N devices using FC. Managed Switch Fabric is often implements VSANs (virtual network SAN) and other technology platforms (eg. To Ethernet). Its ports can also be configured as FL and connect it to the Hub - to add a loop topology managed by the port. E also has ports - for connecting other Switch Fabric (scaling star topology)

Fiber Channel types of ports (I) Port E (Expansion Port E_Port) - switch port for connection to another switch, used for the expansion of the FC switch fabric topology. The combined ports form a so-called E. Inter-Switch Link. Port F (Fabric Port, F_Port) - switch port, used to connect it to other devices N ports. Harbor FL (Fabric Loop Port FL_Port) - switch port that can operate in a loop topology, used to connect to the switch ports NL.

Fiber Channel types of ports (II) Port G (Generic Port G_Port) - the port can operate in E port mode or port F. Harbor U (Universal Port, U_Port) - universal port that can operate in the mode E, F or FL. N port (Port Node, N_Port) - device port can not operate in the loop arbitration, this port connects the switch F Harbor NL (Node Loop Port NL_Port) - port of the device, must be in loop mode, connected to FL port hub or a switch.

Fiber Channel - addressing WWNN / wwpn A unique device identifier in the environment is called Fiber Channel WWNN address (World Wide Node Name) Or wwpn (World Wide Port Name) In a situation where the devices have several FC ports (the notation is the same address) 64 -bit address. Components: Version ID (1: value 0 x 1000, 2: 0 x 2000, 5 value: the value of 0 x 5 or 6 value 0 x 6). In versions 5 and 6 are short identifiers (16 bits instead of 4) Equipment manufacturer identifier (24 bits) The device ID (24 bits in version 1 and version 2, the 36 bits in version 5. 6)

Fiber Channel addressing brief 24 bit addresses are mapped to brief WWNN / WWNP. They are used in a loop topology and PF Switch Fabric They contain only the address of the interface. There are special values of these addresses - identified with the function of the device. Addresses important values: 0 x. FFFFF 6 Clock Synchronization Server 0 x. FFFFF 8: Alias Server 0 xfffffe: Management Server 0 x. FFFFFB: Time Server, 0 x. FFFFFC: Directory Server 0 x. FFFFFD Fabric Controller 0 xfffffe Fabric F Port 0 x. FFFFFF: Broadcast Address

Fiber Channel - zoning Zoning allows for segmentation of networks using Fiber Channel switches Two kinds: Zoning at the hardware level (Hardware zoning) Zoning in the software (Software zoning) Zoning is the same mechanism in relation to the virtual LAN (Virtual LAN, VLAN), because it determines the Cisco abbreviation VSANs (virtual SANs) Zoning is based on the addresses WWNN / wwpn - in the Switch Fabric is an array of classifying addresses WWNN / wwpn zones

Infini. Band (IB) Communication technology supercomputers assuming simplify switching system to maximize throughput and reduce latency It is evolving since 1999, successively multiplying bandwidth in future versions Topology based on hierarchical star without contention for bandwidth (switched fabric) Standards SDR (Single Data Rate), DDR (Double Data Rate), QDR (. . . ), the FDR-10, FDR, EDR, HDR, NDR Multiple bandwidth generally 2 Gbps obtained, with any another technology, but at the same time by multiplying the number of transmission lines within of the technology (1 x, 4 x, 12 x)

Infini. Band - okabowanie The Infini. Band cables are used copper and fiber. For each variant of this technology copper cables (including connectors) can be individually adapted. In addition, there are links aggregate (available through the multiplication of the number of lines) Length limits: 7 meters for SDR, DDR, QDR, 3 meters for the next technology 300 m for optical fiber connections The technology does not exist Ethernet crossover cable It is possible to connect directly to the host adapter (HCA) without the mediation switch x 12 X 4 X 1

Infini. Band - segmentation Technology introduces a subnet (Infini. Band subnets). The subnets devices are assigned 16 -bit identifiers LID It is possible routing between subnets using dedicated for this purpose routers There also bridges Infini. Band for Ethernet All devices Infini. Band They are backwards compatible subnet It is managed by Subnet Manager (SA) - Software allocating IDs by registering devices (ED - End Nodes), Subnet for blocking unauthorized access, management switches (switches in one subnetwork may be many), eliminating other unauthorized SA. SA may reside in. End Node or in the switch

Infini. Band - addresses GUID full addresses the role of the MAC, are fixed for the device and assigned by the manufacturer (composed of 24 bit identifier for the manufacturer and 40 -bit serial number). One device can be assigned many GUID used to identify devices on the subnet They are classified as: Node GUID - determines the Host Channel Adapter (HCA, the equivalent NIC), a router or a switch, GUID port - specifies the physical slot in the device The GUID - GUID also allows you to assign to the functionality provided by HCA (in practice so many GUID assignment to one HCA)

Infini. Band - addressing, summary Types of addresses: Lids (Local Identifiers), unique subnet, 16 -bit, unstable GUIDs (Global Unique Ids) 64 -bitthose compatible with the EUI 64, is intended to identify the equipment, and equipment used by the switches (Infini. Band. Fabric) Gids (Global Ids) 128 -bitthose intended for routing between subnets, volatile, composed of GID-prefix (64 bity) and GUID (64 bity) Multicast GID has 8 bits of the set to a value of 1 (obtained from the same convergence IPv 6)

Infini. Band - types of connections UD (Unreliable Datagram) - the default MTU assumes equal 2044 B and the maximum possible throughput without flow control RC (Reliable Connected) - it is assumed compatibility with IP MTU equal 65520 B (similar maximum length of an IP packet) and performing transmission control Both techniques allow you to run IPo. IB (IP over Infin. Band) And the use of IP addresses Defined is also called. statement. Infini. Band (Regardless of the call) with different meanings, for example: RDMA - Remote Direct Memory Access - it is possible to automatically copy entire blocks of memory Channel send or receive - an analogy for the transmission of data streams multicast transmission

Infini. Band - tools Recording software is required End Point Subnet Manager there ibping who use the addresses GUID Switches are the network identity, so there are collections ibswitches analogous to ibhost maintained by the operating system (Infini. Band End Point) The switches introduced the ability to share the so-called physical link. Virtual Lanes - identical to the traffic classes (permanent virtual links shared between the physical bandwidth). It is possible to embed Subnet Manager in the switch

Wired LAN

Historical background-wire LAN The first attempts: Arcnet - Attached Resource Computer Network the first standard local area networks, allow the transport speed of 2. 5 Mbps. The network was built based on the bus using coaxial cable. Ethernet technologies - the first standard local area network developed in 1976 by Xerox, Dec and Intel, specifying the technical principles of data transmission up to 10 Mbps, then Fast Ethernet a standard that enables transmission speeds of up to 100 Mbps, 10 G Ethernet, Gigabit Ethernet - continuation of Fast Ethernet - 1 Gbps, 10 Gbps , 40 Gbps, 100 Gbps Technologies based on the ring: Token. Ring - 4 -16 Mbps speed built in the topology of a ring, FDDI - transmission of 100 Mbps, the dual ring topology, the optical fiber. Technologies based on switching cubicles: ATM - Asynchronous Transfer Mode, then the solution for the ATM LAN (switches LANE)

More importantly IEEE 802 standards Technology on ISO OSI physical layer (based on it) and a data link network installations (including wireless). IEEE 802 data link layer is divided into two sub-layers: LLC and MAC Examples: 802. 2 - LLC (Logical Link Control) 802. 3, the specification of which was established on the basis of the Ethernet 802. 4 - token Bus 802. 5 - token Ring 802. 11 - Wireless LAN 802. 15 - PAN (Personal Area Network, eg. Bluetooth) 802. 16 - Wireless Metropolitan Area Networks (Wi. MAX)

- Ethernet standard assumptions (I) interframe spacing (IFG interframe gap, the gap interpacket) Stretch of time after the termination of the state of occupancy of the line. It results from the maximum distance between the end hosts and the signal propagation time in the medium. Equal to the transmission time of 96 bits (12 octets). Stations can transmit only when the link is free. In the case of a busy channel must wait until the end of the transmission wait time and further interframe spacing. If the transmitting station detects a collision, suitable for a forcing a collision time (transmission time of 32 bits). If a collision occurs during transmission of the preamble, the station continues to transmit the preamble, then gives even 32 bits of the same signal. Upon detection of a collision station adjusts the length of the time period Ti, during which will not attempt transmission.

- Ethernet standard assumptions (II) It is necessary to introduce an algorithm providing for the delay retransmission frame when it detects a collision. For Ti, and the number is the number of attempts undertaken. It is possible to a maximum of 16 attempts, after which the adapter returns an error. Time Ti is determined from the formula: Ti = R S where S - the width of the time slot, Ri - random number <0, 2 n-1>. wherein n = min (I, 10). Time Ti increases with the number of attempts made to transmit. This time must be a random number, so there was not. backoff (Coupling collision)

- Ethernet standard assumptions (II) time slot S (time slot) Transmission time is 512 bits for Ethernet 10 and 100 Mb / s and 4096 bits for network 1 Gb / s. It stems from two elements: The time required for reaching a signal from one end of the network the maximum size to the other end and back. The maximum time required to solve the problem resulting from the occurrence of the collision (collision detection and sending a signal by the time forcing collision)

- Ethernet standard assumptions (III) Length 512 bits of the time slot defines the smallest size Ethernet frame 64 bytes. All frames smaller than 64 B are consideredcollision fragments (collision fragment) Or so. weak frame (runt frame) And they are automatically discarded by receiving stations. Time slot ensures that if a collision occurs it will be detected during the transmission of the first 512 bits of the frame (for a 10 / 100 Mbit) In networks 10 GB problem of time slot does not exist - a variant of a half-duplex medium usage is not possible.

- Ethernet standard assumptions (IV) IEEE 802. 3 and derivatives defined by the following values of the times and the size of individual frames for Ethernet variants:

Ethernet over medium TP standard auto-negotiation LP (link Pulse) - a mechanism to auto-negotiate connection speed and consequently the medium variant of the standard for Ethernet The origin of the name - signaling (FLPFast Link Pulse). Formerly (10 Mbps network): NLP (Normal Link Pulse) The mechanism also adapted to the network 1000 Base-T and 10 GBase-T Negotiating mode is designed to determine the most common speed of interconnected devices. Priority is duplex (over half-duplex) Principle: at fixed time intervals (for example. 16 ms) pulse sequences are sent (16 bits) containing information about the capabilities of the interface (Technology Ability Field). The mechanism of enhanced technology, Auto-MDIX (Medium Dependent Interface Crossover) - detection of interlacing cable HP

The principle of operation auto-negotiation LP Mode of operation: inquiry-confirmation. Confirmation may be wrong (it is not possible to select the data transmission parameters). Format: 4 bits: ID LAN technology Bit 6: possible mode flag (10, 100, full duplex, half duplex, etc. ). acknowledgment bit (status information) bit error Other: reserved

Token Ring (I) Developed by IBM in the years 1970 -1980 Stations are connected in a physical ring: each station has a connection with the immediately preceding and next TCU is the physical interface (Trunk Coupling Unit) When the station is enabled processes the token and forwards When the station is off the token is passed from her without Live priorities have stations can order the token if necessary transmission Workstations is attached to hubs called unit MAU (Multistation Access Unit) - eg. Cisco 2517 or Cisco 3920.

Token Ring (III) Combinations MAU: Each unit has a MAU type sockets and ringin Ring. Out, serving to connect it with other units MAU You can combine with each other to 12 units MAU With the multiplication of connections between MAU - automatic service event of wire interruption

Token Ring Frame Types of frames: Date / Command Frame token Frame Abort Frame Date / Command Frame is similar to Ethernet and includes: SFD: 8 bits, Frame control field: 8 bits Field access control: 8 bits destination address: 48 bits source address: 48 bits The data (payload)max theoretical length: 18200 octets Suma control: 32 bits End stop 8 bits Status: 24 bits (error codes, etc. ).

FDDI / CDDI Fiber Distributed Data Interface / Copper Distributed Data Interface The protocol is based on the transfer token A standard developed with the assumption the use of optical fiber as a transmission medium. Transmission 100 MB / s A typical urban network, the LAN option Performed as a backbone network high reliability Max 500 stations Two counter-rings In case of failure (interruption of the rings) stations carry out transmission of feedback

FDDI / CDDI - equipment RYPE interfaces FDDI: (SASSingle-attached station) - has a single port duplex (can not perform the double-node ring) (SACSingle-attached concentrator ) - it has a single port duplex, located in the hub and is used to connect SAS (DASDual-attached station) - has two ports (A and B) is a node in a double ring topology physical FDDI: Podwójny ring participation DAS interface (optionally emergency single ring) Gwiazda of participation SAS interfaces.

FDDI Concentrators There are hubs or FDDI CDDI (eg. Cisco C 1400) for the creation of LAN nodes (cost reduction) They have a variety of interfaces and SAS the most one (output) the DAS. Interfere with two rings: Primary - involving SAS interfaces and an expansion of one of the rings connected to DAS Secondary - an expansion of a second ring connected to DAS They have two MAC addresses for the Primary and Secondary (addresses are required to register interfaces, processing token, etc. ). Often working with so-called. bypass Switch - a device that through a special connector bypass monitors the activity of the hub and in case of failure of the DAS cut off from the double ring (ring protection)

FDDI installations Option 1: Double ring of INTERFACE DAS Kveryone interface (eg. on the PCI module in the router, hub) has two ports. Marked symbols A and B next to device. Option 2: Physical star topology with SAS interfaces and concentratora. Ports (usually PCI) We combine FDDI fiber or cable CDDI (twisted pair RJ-12) Interfaces SAS huband. Highlighted hub interface (double DAS) Switches on the ring created in INariancie 1.

FDDI installations DAS interface consists of two ports, labeled A and B form a ring, port A, port B connected to the next device. Continuing a connection until physically close the ring (ring-forming speed to return to the first device) During normal operation, the port is able to DAS described by the word THRU (transfer two tokens). In case of failure or cable station connected to a port - able WRAP B. In case of failure B - able WRAP A. DAS Interface PCI cards can also operate as a SAS - SAS interfaces connected to the hub (takes precedence when port A second port remains inactive)

FDDI - frame format FDDI frame - It is also similar to Ethernet and contains the following fields: Preamble: 16 bits SFD: 8 bits Frame control field: 8 bits destination address: 48 bits source address: 48 bits The data (payload)max theoretical length: 9000 * 5 bits = 5625 octets (max - if the content is not formatted). Max. It is determined by the maximum time before putting the token. Suma control: 32 bits End stop 8 bits Status: 24 bits (error codes, etc. ).

FDDI - Frame Types Since there is the second version of FDDI (as in Ethernet) - a definition field identifying the protocol encapsulated - to this purpose the sub-protocol (LLC or SNAP), which frame is arranged as a payload FDDI You can distinguish the following frame: Basic data frame - without a defined protocol data formatting, 4500 octets with MAC addresses, without them 4478 LLC frame data - basic frame formatted according to the protocol LLC SNAP data frame basic frame formatted according LLC, then according to SNAP frame Token Frames for station management

FDDI - LLC - Logical Link Control The protocol IEEE 802. 2 Use - control link for services, addressing, identification transmitted content LLC frames are placed as an additional header that describes the data (payload) Transmitted by other protocols The maximum length of the frame (including the MAC address 4500 octets, excluding addresses: 4478 octets) Format data fields LLC frame: DSAP: 8 bits - service access point target SSAP: 8 bits - source service access point Field test: 8 bits Data: max. 4475 octets

FDDI - SNAP LLC SNAP - Sub-Network Access Protocol LLC SNAP allows you to specify content transmitted in the frame FDDI - defining a field identifying the protocol Field format data frames LLC SNAP (first 3 fields come from the LLC): DSAP: 8 bits - service access point target SSAP: 8 bits - source service access point Field test: 8 bits Protocol ID: 16 bits Organizational unique identifier: 24 bits Data: max 4470 octets

FDDI, Token Ring, - Token frame Each FDDI interface (SAS port or hub) is isolated ring segments - so the token frame does not propagate throughout the medium, but is passed consistently referred to as The station having token can send a different frame Format the whole frame for frame FDDI token (much simplified, since the frame is transmitted at high frequency: Preamble: 16 bits SFD: 8 bits Frame control field: 1 octet The end stop 4 or 8 bits

A ring initialization procedure The problem: which station has a frame to generate a first token? Procedure: After passing the test station initialization state and generates a TRT (Token Rotation Timer) - defining the minimum token rotation time at which it is able to handle it The frame application is sent TRT If a station receives a different frame application compares TRT - if it is worse than the TRT received frame is discarded and the station promotes further its own frame When a station receives its own frame (no one else did not reject it) - the TRT is valid for the whole of the ring, and this station emits a token.

Mo. CA (Multimedia over Coax Alliance) - the standard for building a home LAN based on existing coaxial cabling, used for television signal distribution CABLE modem may have Mo. CA interfaces from which a composite signal (TV + data) is run to Mo. CA adapters other rooms. The adapter extracts the channel that contains the data and converts it - usually for Ethernet Available Mo. CA specifications: 1. 0, 1. 1, 2. 0 (2010) Mo. CA channels (they must be set up in accordance) The frequencies from 500 MHz to 100 are arranged at 25 MHz and have reference numbers: E 1 -E 5, A 1, B 1, C 1 -C 5 The frequencies of 1150 MHz to 1500 are arranged at 50 MHz and are numbered D 1 -D 8

Rapid. IO Local network of high bandwidth and low latency, designed for data centers Bandwidth determined based on the multiplication value of 2. 5 Gpbs (Gen 1 1 x) Eg. Successively with 5 Gpbs (Gen 2 1 x), 10 Gpbs (Gen 1 x 4) or (Gen 2 2 x), 80 Gpbs (Gen 2 16 x) and so on. Specifications: 1. 1, 1. 2, 2. 0, 3. 0 (2013) The technology originally developed for embedded systems (communication between modules) - but by adaptaję wiring and the introduction of hubs (named Switch box ) Can be used to build the LAN. Hubs are most commonly used inserts optic SFP + or QSFP + (similar to those used eg. In Ethernet). There NIC Rapid. IO for servers (PCI-X, PCIe, etc. ).