Information Technology Reference
In-Depth Information
F
1
11
F
1
13
F
1
34
F
1
14
F
1
46
(4)
W
1
=
•
•
•
•
,
,
,
,
,
,
Forwarding is the key to the trust by workflow — any meaningful workflow will
have forwarding in it; this ubiquity makes forwarding a universal source of informa-
tion on trustworthiness of network nodes. To speed up forwarding a routing cache
stores recently used entries, and is consulted before the forwarding tables. If the ker-
nel finds a matching entry during route cache look up, it will forward the packet im-
mediately and stop traversing the forwarding tables [14]. Fig. 2 demonstrates a case
of RC implemented as a compact binary trie data structure (only tails are shown) with
leaves, being destination IP addresses labelled by a timestamp of last usage, outgoing
port, error condition, if any was generated, and a source address of forwarded data-
gram, thus RC provides the evidence of performed forwarding. The rightmost icon
shows how a compact tree is being expanded when a new entry is cached.
...
...
11
FIB
1
0000
0
195.37.76.195
195.37.76.203
RC
11
11
$
$
.195
.
203
Age
OutPort
Err Src
Forwarding
Fig. 2.
Fast forwarding path (FIB — Forwarding Information Base; RC — Routing Cache;
Age
— time to expire for an RC entry;
Err
- error condition; $- leaf of a trie structure)
2.3 Network Model
The Internet of today handles media and media signalling flows, orthogonal to them
management and control flows, when and if midcom infrastructure is available [15] as
well as some in-network generated and consumed house-keeping flows, e.g. routing. We
describe another house-keeping information exchange — autonomic communication.
An autonomic network is an in-lay of a media network; each node is a first-class
citizen; no distinction is being done intentionally between user, access, edge, back-
bone, etc. node types. We assume that autonomic communication is done by exchange
of messages with no visible relation between message sequences, i.e. there is no no-
tion of autonomic communication flow. Message source and destination are not nec-
essarily always applications, the model allows nodes to exchange messages on the
discretion of protocol stack entities, like it is done in ICMP [16] and IGMP [17]. This
might look similar to exchange of messages between roles as proposed in role-based
architecture [18], however we do not want to hard-code in role-specific headers
sources and destinations of our messages.
A message has a payload and a header: payload represents sender's media behav-
iour, expressed in workflows, header has message source and indirects the destination
by what is called here a concern — a predicate on the behaviour — to a concerned
