|
[ Up ] [ AppleTalk ] [ Cisco IOS ] [ IPX ] [ Network ] [ Windows ]
Novell's IPX
- Proprietary protocol derived from the XNS protocol.
- Connectionless datagram based, similar to IP.
- Added connection-oriented communication through SPX, similar to TCP.
- Used to transport the Novell Netware protocol suite.
- The original network-transport layer protocol that defines internetwork
and internode addresses.
- Used to route packets through an IPX internetwork.
- To communicate with devices on different networks, IPX routes information
through intermediate networks.
- Address (effectively a service address that addresses the packet to its
final network node destination) comprises three parts
- 32-bit network
- 48-bit host field that uses the physical hardware device address
- 16-bit socket that identifies the specific process
- Uses RIP to route packets in an internetwork.
- Uses SAP to advertise network services, broadcasted by default every 60
seconds.
- Uses the upper-layer SPX or NCP to provide sequence and session control.
- When being bridged, all possible encapsulation must to be supported (raw
Token Ring and raw FDDI are supported).
- When being routed, each network uses only one encapsulation.
- IPX RIP
- Similar with IPX IP in that its routing metric is delay as measured in
ticks and hops.
- Sends routing updates by default every 60 seconds and it flushes
routes after 180 seconds if no update has been received.
- Uses split horizon, but only directly connected peers in NBMA,
packet-switched WAN configurations can be connected.
- If the router discovers that a route to a network has failed, it sends
a RIP broadcast that flags the network as unreachable.
- The router waits for 10 ticks and then checks the routing table for an
alternative route.
- If necessary, the router can use the lost route algorithm to find an
alternative route. This lost route algorithm can prevent routing loops.
- By default, LAN hops are counted as 1 tick & WAN links 6 ticks. An
E1/T1 circuit has a value of 6/7 ticks.
- In case of 2 paths with equal tick counts, it uses a hop count as a
tiebreaker.
- If the hop numbers are equal too and the number of routes in the
routing table has reached the value determined by the IPX maximum-paths
command. In this case the most recent entry in the routing table will be
used.
IPX encapsulation format
| Novell Term |
Cisco Term |
Encapsulation |
Remarks |
| ETHERNET_802.3 |
novell-ether |
802.3 with FFFF |
- Also called as Raw Ethernet.
- The original default IPX encapsulation for Ethernet.
- A nonstandard encapsulation that resembles IEEE 802.3.
- Has FFFF instead of a usual 802.3 DSAP and SSAP.
- Has no LLC layer field.
|
| ETHERNET_SNAP |
snap |
802.2 SNAP with 8137 |
- Supported since NetWare 4.0
- Extends the IEEE 802.2 header by providing a type
code, which resembles the type code defined in the Ethernet Version
2 specification.
|
| ETHERNET_802.2 |
sap or iso1 |
802.2 with E0E0 (D)SAP |
- Default since Netware 4.x
- Is standard IEEE 802.2
|
| ETHERNET_II |
arpa |
arpa with 8137 type |
- Includes the standard Ethernet Version 2 header.
|
| TOKEN-RING |
novell-tr |
802.2 with E0E0 saps |
|
| TOKEN-RING_SNAP |
snap |
802.2 snap with 8137 |
|
| FDDI_SNAP |
snap |
802.2 snap with 8137 |
|
| FDDI_802.2 |
sap or iso1 |
802.2 with E0E0 sap |
|
SAP
- Allows nodes such as file or print server to advertise the services they
provide and their addresses.
- A packet of 480 bytes and there is a maximum of 7 SAPs each.
RIP
- A packet of 432 bytes and there is a maximum of 50 networks each.
GNS
- A request sent by a workstation when it comes online.
- Cisco router default GNS response delay on serverless networks is zero
second. Configuration should be updated if the response is too fast for slow
clients.
- Since Cisco IOS 9.21, router can be configured to respond to GNS queries
in a "round robin" way if there are several equidistant servers in
the SAP table.
- If no local NetWare server is available on the local network, the router
will forward the request to the nearest appropriate remote NetWare server.
Novell servers support numerous protocol stacks
- NFS servers
- AppleShare servers
- IBM SNA gateway
Novell's RIP and SAP
- Native protocols active on all IPX interfaces.
- Higher-lever protocols relying on MAC protocols and IPX to handle
lower-level communications such as node addressing.
- Size could be increased if all nodes in the network are capable of
receiving the large packet size.
IPXWAN
- A handshake protocol that negotiates end-to-end options for new links.
- It establishes an accurate routing metric whenever a dial-up link is
established.
- Should be implemented on PPP links if an accurate tick count is required.
- If you let the IPXWAN determine an accurate value for delay, you will get
better routing decisions.
- Ticks also can be adjusted with the ipx delay command based on the
bandwidth of the interface.
- IPX WAN suggested delay value
| Bandwidth |
Type |
Ticks |
| 2.04 Mbps |
E1 |
6 |
| 1.544 Mbps |
T1 |
6 |
| 254 kbps |
|
6 |
| 128 kbps |
|
12 |
| 56 kbps |
|
18 |
| 38.4 kbps |
|
24 |
| 19.2 kbps |
|
60 |
| 9600 bps |
|
108 |
Note
Based on a formula that Novell and Cisco use to calculate ticks with IPXWAN 2.0.
Since Cisco IOS release 10.0, both IP and IPX are supported over PPP between
Cisco routers and Novell servers.
NLSP
- Link-state routing protocol.
- Has a faster convergence than IPX RIP.
- Advertises routing updates incrementally. However, compared to IPX RIP,
the SPF algorithm [n*log(n) - n: the number of adjacency relationship
between routing nodes] is CPU-intensive. One NLSP process (area) should be
limited to 400 routing nodes.
- Uses cost (1 to 63) and hop count (1023) as routing metrics.
- Its redistribution with IPX RIP is handled altomatically by an algirithm.
The algorithm will prevent loops, but may bring about suboptimal routes. And
there is no need to configure route filtering with access lists.
- Runs with Netware 3.11 or higher.
- Similar to OSI's ISIS hierarchical link-state protocol.
- Version 1
- Only one area can be defined.
- Supported since Cisco IOS R 10.3.
- Version 1.1
- Supports multiple areas, allowing a routing hierarchy.
- Supported since Cisco IOS R 11.2.
EIGRP
- Can be used to significantly reduce bandwidth consumption.
- Achieves this by only transmitting changes to the routing table.
- Belongs in the center, not at the edge of the network.
- By default peers exchange only reliable updates over WAN interfaces. But
over LAN interfaces, the full RIP and SAP tables are transmitted by default.
And redistribution is automatic.
- Using IPX EIGRP in the WAN will get the benefits of low bandwidth
consumption and fast convergence. You may need to use the delay command when
you are configuring the routers at the edge of the WAN cloud, because the
cloud topology is transparent to the LAN environment.
- Tracks IPX routing metrics such as ticks and hops as external metrics.
- As the route passes from IPX RIP to EIGRP and back, the hop count is
increased by 2. But the tick metric is not increased so the EIGRP cloud
seems to be zero ticks. With original EIGRP routes, the metric in ticks is
obtained from the interface delay. Then this tick value is accurately
distributed to IPX RIP routers. And for LAN interfaces, Cisco keeps to the
Novell convention of one tick per hop.
- Switching mode for IPX and IP on the router don't correspond exactly. This
is because of the unique data structures involved.
- For IPX, process switching consists of load balancing packet by packet
- this is also how it works for IP.
- With autonomous / silicon switching, load is balanced destination by
destination, which is gain the same way it works for IP.
- However, when IPX in the fast switching mode, it employs load
balancing packet by packet, which is not the method that IP uses.
- For IPX routing, the default is no load balancing. However, you can
adjust the maximum paths on the Cisco router to allow load balancing on
two to four equal cost paths simultaneously. Increasing maximum paths
enables load balancing.
Abbreviations
- DSAP: destination service access point
- EIGRP: Enhanced IGRP
- GNS: get nearest server
- IGRP: Internet Gateway Routing Protocol
- IPX: Internetwork Packet Exchange
- ISIS: Intermediate System to Intermediate System
- LLC: Logical Link Control
- NBMA: nonbroadcast multiaccess
- NCP: Netware Core Protocol
- NFS: Network File System
- NLSP: Netware Link Services Protocol
- RIP: Routing Information Protocol
- SAP: Service Advertisement Protocol
- SNA: Systems Network Architecture
- SPF: shortest path first
- SPX: Sequenced Packet Exchange
- SSAP: source service access point
- XNS: Xerox Network System
Terms
- Encapsulation
- The process of packaging upper-level protocol information and data into a
frame.
- Chosen based on the kind of interface that is being used.
- Hop: Each route that receives and forwards a packet of data in its journey
from source to destination.
- Round robin: A method stopping one server from being accessed all the time
and consequently taking a performance hit. If the router is not configured
to respond in this way, it will automatically respond with the first service
in its SAP table.
- Routing metric: method by which a routing algorithm determines that one
route is better than another.
- Split horizon
- A technique where route information is not advertised on the interface
that the information was received on.
- Used to prevent routing loops and hop count to infinity.
- Can cause problems and it can't be disabled in IPX or AppleTalk
routing.
- Problem associated can be overcome by configuring subinterface.
- Cannot be disabled in IPX RIP.
- Spoofing: A scheme used by routers to cause a host to treat an interface
as if it were up and supporting a session. Spoofs reply to keepalive
messages from the host, to convince the host that the session still exists.
NetWare servers send keepalive messages to all connected clients at
five-minute intervals. This keepalive acts as an interesting packet causing
DDR circuits to remain up.
- Tick: 1/18th of a second.
|