Abhi Shamsundar, Product Management, AIDE, Juniper Networks

Demo - Day 2 Operations with Two New Juniper Marvis Actions

Demo Drop AI & ML
 Abhi Shamsundar Headshot
The still image shows presenter Abhi Shamsundar in the bottom left corner of the screen. The first slide is shown, featuring the title “Wired Stuck Client.” There is a picture of an EX switch with numerous ports. Two points are made: Number one: Mist AI determines a client in a bad state. Number two: Future: power cycle the port to bring it back.

Explained: Marvis’s two handy new Actions.

Two new features have been added to Marvis, the industry's first and only virtual network assistant driven by Mist AI™.  Watch this short demo video to learn about these handy new features and to discover how they further enhance the customer experience. They are, after all,  just a small part of the many amazing things Marvis can do

Show more

You’ll learn

  • Two use cases that aptly illustrate the new Mist AI Actions 

  • How the new Actions appear in the Marvis network assistant dashboard, making your job easier 

Who is this for?

Network Professionals Business Leaders


 Abhi Shamsundar Headshot
Abhi Shamsundar
Product Management, AIDE, Juniper Networks


0:08 right we want to do showcase some pieces

0:11 of marvelous actions uh a quick

0:14 introduction on marvelous actions itself

0:16 marvelous actions are are what we call

0:19 as cup of coffee view

0:21 um

0:22 to the world of saying uh how is your

0:26 network doing and are there actionable

0:28 things across your thousands of switches

0:30 that you can do as an administrator

0:32 I will show the dashboard uh just a

0:35 couple of slides uh to speak about the

0:37 two new marvelous actions that's been

0:39 introduced

0:39 so one of our Focus areas as we spoke to

0:42 our customers saying you know you're

0:43 focusing so much on uh the client side

0:45 of the house which is what we want we

0:47 want to see the client experience

0:49 but the most important use case that are

0:51 out there uh is you know some of these

0:53 applications that run on sometimes Poe

0:56 they just get stuck uh it happens and

0:59 usually the solution to it is just you

1:02 know bounce them and they come back come

1:03 back right back up the solution isn't

1:06 usually the same simplistic way I

1:08 defined it let's say for example there's

1:11 a camera and the camera is connected to

1:13 your switch and these cameras are

1:16 usually done they usually look at a ping

1:18 Loop to see whether they're pingable or

1:20 not and at some point a camera goes away

1:22 now the application theme which is the

1:25 camera team has to open a ticket with

1:28 the network team to say go as a first

1:30 order of defense let's just bounce the

1:31 port and see if it comes back up

1:34 is it uh an easy process not necessarily

1:37 because you have to open a ticket and

1:38 the ticket will be served in the order

1:39 of operations you're looking at camera

1:41 being down anywhere between an hour to

1:43 eight hours or however long it takes for

1:45 the network team to actually bounce the

1:47 port unless you have somebody on site to

1:49 just go on plug and plug

1:51 simple example of what we are driving

1:54 towards

1:55 can we as a system we get Telemetry all

1:58 the time from each every single port and

2:00 we as a system identify that this is a

2:03 camera first up and there are seven

2:05 other cameras on the same switch and 60

2:07 other cameras on the same side and they

2:09 are all working

2:10 and this same camera seasonality wise

2:13 was also working yesterday last week all

2:16 of these elements considered if that is

2:19 the case if you're true positive

2:21 can we actually bounce the port for you

2:23 instead of going through this cycle of

2:26 network theme application theme ticket

2:28 and bounce support

2:30 it's a simple use case it's a lot of

2:33 algorithms that run to meet all the

2:35 requirements that I mentioned right

2:37 above and that's that's where we're

2:39 getting into this point saying today

2:40 we're able to identify the court and we

2:43 will tell you that these are the ports

2:44 tomorrow if you give us permission we

2:47 will bounce the port for you

2:49 a simple use case but a whole lot of

2:52 backend that goes into making this

2:54 action possible so that's the uh the

2:57 temporary name for it is wired stock

2:59 client I think we should improve come up

3:01 with a smarter name uh but we will get

3:03 to that

3:04 so that's the camera use case another

3:07 use case that we have come about is uh

3:10 you know devices that are constantly

3:11 stuck at high CPU

3:13 so super simple all of your nms's in the

3:16 world can probably tell you your CPU is

3:19 high what are you doing here right one

3:22 obviously we are able to tell you

3:24 amongst the thousands of devices in your

3:25 network which are consistently stuck

3:27 here are the people who are the switches

3:29 that that you would want to pay

3:30 attention to that's one that probably

3:32 your nms can do today

3:34 two we are telling you which are the top

3:37 three processes that are actually

3:39 causing this device to be in this state

3:41 that also tells you that

3:44 it all comes back to the Mist Cloud how

3:47 can we use this data to say there is

3:49 something common amongst all of this how

3:52 can we help you implement best practices

3:54 for example on your switches

3:56 if you are constantly addressing your

3:59 where something called this routing

4:00 engine that helps route packets makes

4:02 decisions all of that is in the software

4:04 it's consumes CPU Cycles

4:06 if you want to protect your routing

4:09 engine it is Juno's best practice to

4:11 always enable protection of rotting

4:12 engine where you say I will only answer

4:14 packets from these destinations my bgp

4:17 peer somebody who's doing an SNMP query

4:20 uh whatever those or SSH from five or

4:23 six clients which are trusted

4:26 but rest of them if there's an hsrp

4:28 packet I'm just an L2 device I don't

4:30 want to care about it I would I always

4:31 multicast for example ospf multicast if

4:34 you know or even any broadcast packet

4:36 that is there I don't want to just open

4:38 the packet drop it open the packet drop

4:40 it that's CPU Cycles can I save on that

4:42 that's protection of routing engine

4:44 another example that could be you know

4:47 us saying the there's a there are these

4:49 Optics that we certify but that you're

4:51 using other Optics in we from uh the

4:54 missing Universe perspective are able to

4:55 say whenever you're using these

4:57 uncertified Optics there's a problem

4:59 something for you to look at so there is

5:02 a lot of cloud angle that we are

5:04 bringing in to say we can help you if

5:07 there are devices that are stuck we can

5:08 help you provide suggestions to say you

5:11 know we're not only just telling you

5:12 high CPU go figure out you're telling

5:14 you how can we help make this better so

5:16 that's the intention of these two Marvis

5:19 actions

5:20 um I will introduce you introduce

5:22 marvelous actions to you uh this is the

5:24 cup of uh coffee view as we call call

5:26 them there's a whole bunch of actions

5:28 that are already in here

5:30 and again all of this is at the scale

5:34 of 50

5:36 000 60 000 ports one of our customers we

5:39 see uh 1.2 million ports

5:41 we see we have about 400 actions in

5:44 there the Juniper office

5:46 uh amongst the 52 000 ports they have

5:48 they run about 1200 switches all across

5:51 uh the the all of all of the offices

5:54 across the world uh we're talking about

5:56 close to 26 27 marvelous actions so we

5:59 are not killing you with you know things

6:01 that are just alerting we'd rather say

6:03 these are true positives that we are

6:05 super confident about and you can pay

6:07 attention and you can kill uh you can

6:09 take care of it for example bad cables

6:10 negotiation mismatches

6:13 um these are the two new uh ones that

6:15 we're talking about we said you know

6:17 there is uh stop for uh this is uh this

6:21 particular device uh what do you want to

6:23 do with it you could power cycle disable

6:25 the port and the second one is saying

6:28 you know there is a device that's stuck

6:30 at 100 perennially stuck at 100 and not

6:33 just a simple Spike these are again the

6:36 our intent is for us to say

6:38 from this dashboard to say if you look

6:40 at this every morning this should give

6:43 you a list of items that you can

6:45 immediately take care of because

6:46 obviously we can't fix a bad cable uh we

6:49 cannot uh you know uh you know add uh or

6:52 how do you deal with a port stack issue

6:54 how do you do with the CPU we'll give

6:56 you recommendations

6:57 then eventually if you trust us enough

7:00 say Auto uh bounce support for me we'll

7:03 get you there that's that's the

7:05 intention of as simple as they look

7:08 there's a lot of math course behind it

7:09 and also you know the power of cloud you

7:11 will bring in elements of saying

7:14 we have the ability to do this and Cloud

7:16 Source information to say this is best

7:18 practices you might it's good for you to

7:20 follow this

Show more