‍

0:00

Hi, I'm Cla Connor.

0:01

And I'm Gail.

0:02

And today we're going to dive into the

0:04

capabilities of the operations genie.

0:06

We're going to give it a really hard

0:07

task that normally takes mainframe

0:09

professionals

0:11

hours if not days to complete.

0:12

Oh, I know. I know it's going to take me

0:14

days to complete. I got the operations

0:17

genie chat interface running. As you can

0:19

see, it's connected to uh Google Gemini

0:22

3 flash. And all right, Claire, what

0:24

what would you like to do today?

0:26

Uh, well, everyone's always interested

0:27

in the monthly peak. So let's see if it

0:29

can find a rolling 4hour average for us.

0:32

All right. So we have some uh

0:34

pre-formatted SMF data here and let's

0:36

let's try to ask it. So I'm going to

0:38

type in locate the 4hour window with top

0:44

CPU consumption. All right. And here it

0:47

goes.

0:49

And as you can see I even made a small

0:51

typo and it still understood what I want

0:53

to do. And it's analyzing the CPU usage.

0:57

uh it's thinking how to extract the data

1:01

and it's performing the queries and it

1:04

came back and says that it found the

1:07

4hour window with high CPU consumption

1:09

on December 7 uh between uh noon and 400

1:13

p.m.

1:14

Oh, great. So, um what jobs are actually

1:16

running during that?

1:17

Right, let's try and ask it. uh list the

1:21

top 10 CPU

1:25

consuming jobs in the interval

1:30

and it's now querying the data to

1:33

identify the top CPU consuming jobs and

1:35

it came back with a list of jobs and uh

1:39

the total CPU time each consumed.

1:42

Oh, that's really cool. I don't think

1:44

I'm familiar with that top one. Um how

1:46

do we find out what it does? Let's try

1:49

and ask it. So I'm going to just type

1:52

what does job

1:55

do. As simple as that. All right. So

1:58

it's investigating the job uh and

2:01

extracting the data and it came back and

2:04

it identified uh this is an application

2:07

job managed by the software development

2:10

team. Uh, it's showing us the program

2:13

and the step name. And it says that this

2:17

is an IO inensive job. Apparently,

2:20

reading from tape um, and writing to

2:23

disk.

2:24

Oh, that's really cool. Um, and seeing

2:26

as it's a a high consumer, what would

2:28

happen if we were to move it out of the

2:31

rolling 4hour peak?

2:33

Yeah. So I guess some some people might

2:36

think it doesn't make sense to have a an

2:39

IO consuming job during the monthly

2:41

peak. So we can try and ask it to

2:44

estimate

2:45

uh savings if we

2:49

take this job out of the interval

2:57

and it came back with a wow 42% savings.

3:01

if we take this job out of the uh

3:05

monthly interval.

3:06

That's that's quite impressive. Um and

3:09

it it it definitely shows that it's got

3:11

other jobs related to it as well and it

3:13

links between them too. Um well, how

3:16

about we look at the top IO consumers

3:18

for that interval.

3:20

Yeah, let's see. Let's see what it says.

3:29

So we can simply ask it to list the top

3:31

IIO consumers in the interval

3:34

and it will query the SMF data again.

3:38

And it comes back with the same job and

3:40

a list of several other jobs that are

3:43

CPU consumers and how much uh uh blocks

3:46

count each of them moved during the

3:49

interval.

3:49

That I mean that's really impressive. So

3:51

as you say we can see that the top

3:53

consumers the same job there. Uh so

3:55

let's see if we can dive even deeper

3:56

like what DDs and devices does it use?

4:00

List the DDS and devices

4:04

job

4:06

uses.

4:09

It's going to analyze the DDS and

4:12

devices

4:14

and it came back and it's showing me the

4:17

DD names, the device numbers, uh number

4:20

of blocks transferred and the uh IO

4:23

connect time for each DD. So it seems

4:26

that in fact this job is reading from

4:29

tape and uh moving to disk. All right.

4:34

Well, that's that's really good. I can

4:36

we see a bit of a higher level system

4:38

view on the storage that it's it's

4:40

connected to?

4:41

Yeah. So let let's ask it to show us the

4:44

status of the SMS storage groups.

4:57

So as you can see this time it's getting

4:59

system information. This is a a live

5:02

system information not uh only data

5:05

based on the SMF.

5:08

And it came back with the list of the

5:09

storage groups and the pool types and

5:12

the total space in each storage group,

5:14

how much percent is used, the number of

5:16

volumes, uh and as well the description

5:19

of each storage group.

5:21

Wow, that's really impressive. It shows

5:23

a lot of detail in there. I mean, what

5:26

we've just done there in like 3 4

5:28

minutes would normally take a whole day

5:30

to do. So, yeah, it it's really

5:34

impressive how how quick it manages to

5:35

look through all of that.

5:37

Yeah. All right. Now, we've gone much

5:39

deeper and we've seen what jobs are in

5:41

that peak period, what they're doing,

5:44

and and some of the possible impacts of

5:46

moving them out. And we've managed to do

5:48

all of that in natural language rather

5:50

than having to flip between 3270 and

5:53

lots of other different screens. We've

5:55

managed to do all of that from the

5:57

operations Genie chat interface.

6:00

Yeah.

‍