Information Technology Reference
In-Depth Information
2 Related Works
Openflow utilizes the three tuple
to denote
the abstraction layer of network forwarding. In the evolution progress, the match-
ing field extended from 12-tuples of Ethernet, IP and TCP in Openflow specifi-
cation 1.0 to 15-tuples in Openflow specification 1.1, and from fixed length rules
to the variable-length matching field based on TLV
Matching, Instructions, Statistics
<
>
in
Openflow specification 1.2. The complexity processing of matching implemented
in Openflow switches try to support new network architectures. However, just
enlarging the matching field would increase the diculty of the implementation
and could not satisfy unknown network protocols. The processing of Openflow
extends from one stage process to multiple stage pipelines in order to solve the
combination blast of multiple tables, which would increase the complexity of
mapping the special processing of the content-centric or service-centric network
architectures to the multiple pipelines of Openflow. The functions of Openflow
switches are continuing to develop, but they are limited to the processing of
packet basic options, such as modification and replacement, and still could not
support the requirements of new network applications, such as data cache.
To solve the complexity of the implementation of Openflow multi-stage
flow tables, based on current router functions of forwarding and controlling,
Openflow+[15] made an extension of Openflow on the aspects of the interface
between the forwarding and controlling layers and flow tables, and then pro-
posed the presentation of openflow rules based on TLV and the mechanism of
the mapping Openflow flow tables to ACL and FIB of routers. The scheme pro-
vides a simple method to implement Openflow on commodity routers, simplify
the implementation of Openflow and enhance the deployment of Openflow.
Juniper enlarges the forwarding layer by customization services, and then
proposes the service layer[16]. By integrating computing and storing resource
with network elements Juniper service layer provides the platform and SDK for
services from the third-party, which could support multiple services of network
managements and has good scalability. Due to the capability of SDK Juniper
service layer runs services mainly including network management, which could
not support packets processing in depth in the key data path. On the other hand,
Juniper service layer only admits authorized partners developing new network
applications on their platform, which lacks of openness.
Middleboxes[11] provide a novel method of patching to add new functions to
Internet, which could implement general packet forwarding and advanced packet
processing, such as NAT, firewall, proxy, IDS, etc., in order to enhance the
ability and security of Internet. On the other hand, the widespread deployment
of middleboxes and other network appliances has primarily resulted in some
challenges and criticism due to poor interaction with higher layer protocols.
Middleboxes also brought a lot of problems, which utilize the method of redi-
recting or DNS mechanism for targeting network trac to Middleboxes. The
method would increase the complexity of the network control, and add packet
processing delay. On the other hand, Middleboxes are short of scalability, which
could not meet the requirements of high-performance flows processing, and lacks
<
Type, Length, Values
>
 
Search WWH ::




Custom Search