Information Technology Reference
In-Depth Information
1024-byte packets, 3 threads arrive at the line-rate forwarding of 512 bytes pack-
ets, 4 threads reach the highest forwarding rate, the maximum bandwidth of the
tester's ports, 20Gbps, for 256 bytes packets. 64 bytes packets forwarding rate
is relatively high, because in the case of the same internal packet buffer size,
the number of cached small packets is relatively high. For the system through-
put, under the same number of threads the larger packet size is, the higher the
throughput is, for the system throughput is proportional to the length of the
packet under the same packet forwarding rate. For the packet size of 1024 bytes
single-thread could achieve the maximum rate of the tester, and for other packets
sizes as the increasing of the number of threads (processing power), the system
throughput increases. The results show that the PacketDirect mechanism of the
prototype meets the performance requirements of the system throughput.
14
25
64B
128B
256B
512B
1024B
64B
128B
256B
512B
1024B
12
20
10
15
8
6
10
4
5
2
0
0
1
2
3
4
5
1
2
3
4
5
7KUHDGV
7KUHDGV
(a)Forwarding
(b)Throughput
Fig. 5.
PacketDirect IO Performance of Prototype
6Con lu on
Labelcast provides and abstracts computing and storing resource within net-
work devices for researches, and support running user-defined service based on
extensible resource container, which provides a reliable and programmable data
plane and support new network architectures. By dividing the data plane of the
network into fast forwarding plane and intelligence service plane based on se-
mantics of processing in networks, Label tables and Cast tables are proposed to
abstract forwarding resources and the computation and storage resources sepa-
rately. Action instructions based on fixed-length labels in Labelcast is easily to
be implemented in hardware and service atoms of special processing could be
dynamically extended compared with Openflow. NDN are designed on the NPE
platform with Labelcast model, which support a large scale of parallel requests
on the same content and packet by packet processing. The implementation of
the prototype promotes the evolution of network architectures.
