Software-Defined Networking: A Comprehensive Survey [pdf](arxiv.org)
arxiv.org
Software-Defined Networking: A Comprehensive Survey [pdf]
http://arxiv.org/pdf/1406.0440.pdf
5 comments
As someone who works in this area but is not aware of academic developments or terminology, this looks like a great paper. Thanks for sharing. My direction: http://stani.sh/walter/pfcts/
Here's some more info on SDN:
http://www.networkworld.com/article/2159545/software/what-is...
Can anyone explain in simple terms why SDN is the "Cisco killer"?
Traditionally the smarts that determine how the network is used (e.g. optimising routes) were provided by the network devices, bundling hardware and software. SDN allows the control to be separated from the hardware, rendering the hardware a basic commodity. Cisco traditionally sold devices/hardware, and would struggle to compete with commodity hardware. Cisco are trying to respond by strengthening their software.
It's a little more nuanced than that. There are two basic orthogonal aspects which scare the traditional network vendors.
1)The architectural separation of control and data planes, with the control plane having a unified and global view over the entire network, allows network functionality, currently primarily operational enhancements, that weren't possible before. One example would be segmenting your network into many logical networks, say one for each tier in your web application, and then being able to arbitrarily place and move around VMs to different hosts with the network automatically being reconfigured to make sure traffic goes where it needs to but not anywhere else.
2) By removing most of the smarts from the data plane elements like switches and routers, and with the rise of good merchant switching silicon, there is little reason to pay a "Cisco tax," even if one weren't using the SDN architectural model to do anything different from your legacy network. Whitebox switches from OEMs like Quanta are really cheap compared to Cisco switches. What's more, the majority of all Cisco revenue comes from switches and related products like SFPs; Cisco is very exposed here. Whitebox products have been available for some time, but the software side of these was very weak. The SDN ecosystem that now exists has given rise to good alternatives on the software side so that one can have a turnkey whitebox switching solution, even if running a full SDN data center is not the plan.
If you look at what has happened architecturally in the most advanced data centers (Google, facebook, amazon, etc), both of these advancements have the norm for some time. The promise of SDN products made available to the general public is that one can reap some of the benefits that Google, et al. have realized, without having to hire an army of developers to reinvent the wheel.
It is not just Cisco that must respond to SDN, but Cisco, as the dominant switch provider, has the most to lose.
1)The architectural separation of control and data planes, with the control plane having a unified and global view over the entire network, allows network functionality, currently primarily operational enhancements, that weren't possible before. One example would be segmenting your network into many logical networks, say one for each tier in your web application, and then being able to arbitrarily place and move around VMs to different hosts with the network automatically being reconfigured to make sure traffic goes where it needs to but not anywhere else.
2) By removing most of the smarts from the data plane elements like switches and routers, and with the rise of good merchant switching silicon, there is little reason to pay a "Cisco tax," even if one weren't using the SDN architectural model to do anything different from your legacy network. Whitebox switches from OEMs like Quanta are really cheap compared to Cisco switches. What's more, the majority of all Cisco revenue comes from switches and related products like SFPs; Cisco is very exposed here. Whitebox products have been available for some time, but the software side of these was very weak. The SDN ecosystem that now exists has given rise to good alternatives on the software side so that one can have a turnkey whitebox switching solution, even if running a full SDN data center is not the plan.
If you look at what has happened architecturally in the most advanced data centers (Google, facebook, amazon, etc), both of these advancements have the norm for some time. The promise of SDN products made available to the general public is that one can reap some of the benefits that Google, et al. have realized, without having to hire an army of developers to reinvent the wheel.
It is not just Cisco that must respond to SDN, but Cisco, as the dominant switch provider, has the most to lose.