Department of Teleinformatics Network Services Royal Institute of Technology

Download 290.3 Kb.

Page	4/10
Date	28.01.2017
Size	290.3 Kb.
	#8874

1 2 3 4 5 6 7 8 9 10

3.4Autoconfiguration

IPv6 introduces the term autoconfiguration – the capability of configuring a node without the help of a human. This is a very welcomed feature for network administrators, but also for non-experienced end users.

3.4.1Mechanisms

IPv6 delivers the autoconfiguration functionality using three mechanisms:

Neighbor Discovery (ND) [27] is actually a set of ICMPv6 messages, which replaces the services provided by ARP and Router Discovery defined in IPv4.
Stateless autoconfiguration [33] assigns a globally valid address to an interface by combining its link-local address with address prefix information advertised by nearby routers. No servers or human interaction is required for this operation.
Stateful autoconfiguration provides additional autoconfiguration parameters such as DNS servers by using the Dynamic Host Configuration Protocol for IPv6 (DHCPv6) [8]. This is the preferred method of administrating address assignments since it gives full control over the assignment process. However, DHCPv6 is still a work in progress and has not been fully implemented or tested yet.

These mechanisms may be used together or separately depending on the network topology and router parameters set by the network administrator as described in the next section.

3.4.2Procedure

The automatic configuration of a node is a multi-step process. A flow chart illustrating the steps involved can be found in Appendix A. The complete process is as follows:

The interface is activated.
A link-local address is generated (but not assigned to the interface) by concatenating the predefined prefix FE80::/10 with a 64-bit interface identifier as described in Section 3.3.4. The interface identifier can typically be the IEEE-802 address of the interface card (e.g. Ethernet, FDDI) or another unique number taken from other parts in the node (e.g. the main board serial number).
Neighbor discovery is then used to check if the newly constructed address is unique (on the link). This is done by sending ND solicitation messages with the destination address set to the address being tested, and source address set to the unspecified address (::). If a ND advertisement message is received, the address is not unique and has to be recreated either manually or randomly.
Once the link-local address is known to be unique, the address is assigned to the interface being configured.
Using the new link-local address as a source address, a ND router solicitation message is sent to the all-routers multicast group (FF02::2).
In response to the ND router solicitations, routers send a unicast ND router advertisement message back to the node. The advertisement specifies if the node should use stateless or stateful autoconfiguration by setting the managed configuration flag accordingly. If stateless autoconfiguration is to be used, a site-local or global address is constructed using an address prefix included in the advertisement and the current link-local address. The new address is then assigned to the interface (which now has two addresses). The host is now configured for communication inside the site or even on the Internet at large.
If there is no response from a router, or the advertisement specifies managed addressing, stateful autoconfiguration should be used. This is handled by DHCPv6, which defines message types for configuring all necessary parameters.

3.5Real-time Support

To meet the demands of today’s increase in real-time application such as streaming audio or video, IPv6 specifies the following new related features.

3.5.1Flows

In the IPv6 header, there is a 20-bit Flow Label field defined. A flow is implicitly defined as a set of packets that come from the same source to the same (unicast or multicast) destination bearing the same flow label. New flow labels are generated randomly in the range 1 to FFFFF hex. Packets are however not forced to use flow labels and then use a value of zero as flow label. In fact, most packets will probably have this unspecified flow label.
Flows may be used to indicate that some packets require special handling by the IPv6 routers in the network such as low delay or high bandwidth. Then may also be used in conjunction with the router header [15] to restrain all packets to the same path through the network. To provide resource allocation, a protocol such as the Resource Reservation Protocol (RSVP) [9] could be used. RSVP is based on the use of flows and is therefore well suited for IPv6 as described in [6].
The usage of the flow label is still experimental, and a final decision on the usage will be made when the needs come clearer.

3.5.2Traffic Class

The 8-bit Traffic Class field can also be found in the IPv6 header. Much as the Type of Service field in IPv4, this field provides the usage of differentiated services [28]. It also provides prioritized routing where packets sent with higher priority originating from the same source will be prioritized.

3.5.3Jumbograms

To support extreme high-speed traffic in real-time, IPv6 provides a possibility of sending very big packets; so called Jumbograms [7]. Usually, the 16-bit Payload length field in the IPv6 header limits the maximum payload length to 65,535 bytes. Using the Jumbo Payload option in a hop-by-hop extension header [15], the maximum length is extended to 4,294,967,295 bytes (using a 32-bit length field). However, the use of Jumbograms requires the underlying link layer having a link MTU of at least 65,575 bytes (65,535 + 40 for the IPv6 header).
W

hen using the Jumbo Payload options, the Payload length field in the IPv6 header must be set to zero. The Next header field in the header is also set to zero, indicating the following hop-by-hop extension header where the actual payload length can be retrieved as illustrated in Figure 3 .6.

Figure 3.6: The Jumbo Payload option

Download 290.3 Kb.

Share with your friends:

1 2 3 4 5 6 7 8 9 10