Juniper Apstra Demo: Make network changes with reliability guardrails
Experience More
Transcript
0:01 what if you could deploy to production
0:03 on a Friday night and still make it home
0:05 for dinner what if you could push
0:07 changes without waiting for a
0:08 maintenance window
0:09 did you allow for time to troubleshoot
0:11 and find mistakes when the deployment
0:12 didn't go as well as planned
0:15 the number one cause of outages in most
0:17 networks is human error the Pokemon
0:18 Institute says the average cost per
0:21 outage is seven thousand six hundred
0:23 dollars per minute and that the average
0:25 outage is 86 minutes long
0:27 did you budget an extra 645k for this
0:30 new deployment just in case something
0:32 goes wrong
0:33 we all have the Need for Speed to stay
0:35 ahead of business demands but automation
0:37 alone is only an amplifier good becomes
0:40 more good but bad becomes more bad
0:43 turns out that merely push and config
0:45 even a golden config doesn't guarantee
0:47 correctness if you don't believe that
0:49 then why do you use maintenance windows
0:51 it's time to roll out that big new
0:53 application is your network ready
0:55 did you budget for mistakes
0:58 the business is moving forward with a
1:00 new Commerce platform to be ready for
1:01 the holiday shopping rush and teams have
1:03 been working hard to get ready then
1:05 surprise marketing decided to run a big
1:07 pre-holiday sale in the timeline for
1:09 rollout just shortened by several weeks
1:11 the app team wants to push production
1:14 now on a Friday afternoon without a
1:16 maintenance window
1:18 do you even know if it'll work did you
1:20 budget for time to clean up the mistakes
1:22 that come when you're rushed this
1:24 deployment must be fast and it must be
1:26 right with abstra it will be
1:29 the requirement from the app team is
1:30 three vlans across two racks but the
1:33 process would be the same if this app
1:34 was spread across dozens or even
1:36 hundreds of racks
1:37 we'll be creating a virtual routing Zone
1:39 with three VX land-based virtual
1:41 networks to support this requirement
1:44 first we'll create the routing Zone
1:46 we'll assign IP addresses and vnis from
1:48 predefined pools and enable a DHCP
1:51 server for this routing Zone
2:09 then we'll commit that routing Zone
2:10 configuration to the network
2:19 next we'll update a previously created
2:20 connectivity template with this new
2:22 routing Zone to enable external routing
2:24 for the zone
2:48 foreign
2:50 networks and assign these virtual
2:52 networks to our new routing Zone
2:57 we'll assign vxlan as a network type
3:03 let the system choose a vni from a pool
3:07 enable DHCP on the virtual Network
3:12 assign a subnet in Virtual Gateway IP
3:14 address
3:19 ask App Store to create a new
3:21 connectivity template for the tag server
3:23 facing interfaces
3:25 and we'll choose which switches this
3:26 virtual network will be deployed on
3:31 oh didn't expect to see that
3:39 oops looks like I fat fingered that
3:42 virtual Gateway IP address nice catch
3:44 abstra
3:46 okay fixed and fixed before I deployed
3:49 so I didn't cause any problems on the
3:51 running Network
3:54 we'll repeat these steps for the other
3:56 two virtual Networks
4:17 we'll identify vnis to use from
4:19 predefined pools
4:24 and we'll assign these virtual networks
4:26 to the server-facing interfaces where
4:28 the networks will be deployed
4:56 notice that we didn't enter any
4:57 vendor-specific configuration details
5:00 apps rule render configs for us and we
5:02 don't need to log into network devices
5:04 and type into a CLI so there's zero risk
5:07 of making a manual error
5:09 so let's take a look at rendered configs
5:12 and see what we're adding we just added
5:14 hundreds of lines of config in just a
5:16 few minutes and we did this for every
5:18 device involved in the new network
5:19 service in this demo we're only
5:22 configuring a few devices but the
5:24 process is identical even if we did this
5:26 across hundreds of devices
5:28 how long would this have taken if we
5:29 created these configurations manually
5:32 and what could have gone wrong if we
5:34 made a mistake it didn't catch it before
5:36 we deployed
5:43 last step is to commit the configuration
5:45 and wait for the dashboard to turn green
5:55 we just rolled out that new network
5:57 service without mistakes in minutes
5:59 instead of days or weeks and we know
6:01 that it works because apps are validated
6:03 the configuration and didn't push
6:05 mistakes
6:06 we are assured that the network is
6:08 operating as intended
6:11 with app store we can be confident that
6:13 the deployments just work every time and
6:15 that will make a home for dinner
6:18 foreign
6:21 [Music]