Coursera SDN M1.2.1 SDN History: Programmable Networks 2

NOTE

Capsules Example

1.Type Field: excute forwarding routine => carry program

2.Previous address: If the node now is not an active network node, it can get the preivous address and forwarding routine from the field.

3.Dependent Fields: necessary! It's the args of the forwarding code.

4.Payload.

Some Previous Notable Projects

1.ANTS(MIT) => carry Java Code => lower QoS guarantee but better Real-Time performance.

2.SwitchWare(Penn) => support invocation(调用) of switchlets

3.Smart Packets(BBN) => network management

4.Open Signaling(Columbia) => programmable processing of packet streams

5.Tempest(Cambridge) => Switchlets

Why disappear?

1.Timing was off.
-1.No killer applications.

SDN's killer application => Data Centers of the Clouds.

-2.Hardware doesn't support.

2.Some missteps.
-1.focus too much on sercurity aspect, but not how do we provide programmablity.
-2.focus too much on programmer, but not our network operator!
-3.focus too much on Interoperability(协同).

On the contrast, OpenFlow did a good job grappling(i.e.fix) with backwards compatible(兼容) with switch hardware.
1.Simple firmware(固件) upgrades.
2.Switch hardware already supported the basics.

The legacy of active networks for SDN

1.Programmable functions in network to enable innovations.

2.Demultiplexing(多路处理) programs on packet headers. eg. Flowvisor, P4.

3.Paying attention to middleboxes and how these functions are composed.

It's worth to looking back to review the project of integrating middleboxes with active networks and see how SDN can refer to it.

2017/2/23