1
00:00:00,750 --> 00:00:05,040
Now Christian often asked about when and open flow is.

2
00:00:05,040 --> 00:00:06,990
Is this actually real?

3
00:00:07,140 --> 00:00:13,500
Are there any implementations of software defined networking or open flow in the real world?

4
00:00:13,800 --> 00:00:18,660
One of the earliest examples of an and open flow can be found at Google.

5
00:00:19,460 --> 00:00:25,460
Google started deploying open flow in their networks in 2010.

6
00:00:25,640 --> 00:00:29,390
Now, at the time of this recording, that is over five years ago.

7
00:00:29,420 --> 00:00:33,230
So sudden and open flow have been around for quite a while.

8
00:00:33,410 --> 00:00:40,160
I recommend that if you're interested in seeing examples of Zen and open flow, have a look at the Open

9
00:00:40,160 --> 00:00:43,250
Networking Summit videos on YouTube.

10
00:00:44,380 --> 00:00:48,250
This first example, which I've explained is open flow at Google.

11
00:00:51,000 --> 00:00:56,730
And the video explains how Google had problems with the existing infrastructure and then deployed their

12
00:00:56,730 --> 00:01:01,050
own switches running open flow and other protocols to scale.

13
00:01:01,710 --> 00:01:05,850
In this presentation from 2012.

14
00:01:06,360 --> 00:01:14,430
Google's infrastructure is explained and others in this video explains how Google have got to 100% network

15
00:01:14,430 --> 00:01:16,230
utilization in their network.

16
00:01:16,980 --> 00:01:21,810
He further explains why they used open flow in their core network.

17
00:01:21,960 --> 00:01:29,340
One of the reasons given is that networking costs do not go down as Google scaled their network.

18
00:01:29,610 --> 00:01:35,700
This is in contrast to compute and storage when running a data center at scale.

19
00:01:35,730 --> 00:01:40,560
Google expects massive savings compared to a small implementation.

20
00:01:40,800 --> 00:01:45,390
This was happening to compute and storage, but not to networking.

21
00:01:45,480 --> 00:01:52,380
The cost of networking stayed linear or increased, which is very different to servers and storage,

22
00:01:52,380 --> 00:01:57,360
where the per device cost went down as the implementation was scaled.

23
00:01:57,840 --> 00:02:03,430
In networking, it was hard for Google to save money even when there were large implementations.

24
00:02:03,450 --> 00:02:12,060
So by implementing open flow and sd-wan, Google were able to manage many networking devices as a single

25
00:02:12,060 --> 00:02:16,570
unit rather than lots of individual networking devices.

26
00:02:16,590 --> 00:02:19,740
Now they are updates to the Google presentation.

27
00:02:20,480 --> 00:02:25,460
Here's one from 2015 which provides a lot more detail.

28
00:02:26,660 --> 00:02:29,540
There's also presentation from 2014.

29
00:02:29,570 --> 00:02:32,540
I strongly recommend that you watch this video.

30
00:02:32,570 --> 00:02:37,130
The 2015 keynote presentation about Google's implementation.

31
00:02:37,520 --> 00:02:43,760
It explains the reason why and the advantages of implementing open flow and steam in Google's network.

32
00:02:46,070 --> 00:02:52,250
Google have also released more detailed information where they pull back the curtain on Google's network

33
00:02:52,250 --> 00:02:53,150
infrastructure.

34
00:02:53,330 --> 00:03:00,800
So a lot of detail is provided on how Google have implemented their network and can manage multiple

35
00:03:00,800 --> 00:03:03,680
devices using open flow and MSDN.

36
00:03:04,010 --> 00:03:07,520
The history is provided of different Google implementations.

37
00:03:08,910 --> 00:03:17,400
As well as detailed whitepaper information explaining scaling network devices, protocols used and so

38
00:03:17,400 --> 00:03:17,910
forth.

39
00:03:18,710 --> 00:03:21,530
If you're interested in the details, have a look at these documents.

40
00:03:22,320 --> 00:03:26,850
Otherwise, simply watch this presentation from the Open Networking Summit.