CoreWeave ARENA Demo

1

00:00:04,400 --> 00:00:05,066

Hi there.

2

00:00:05,066 --> 00:00:06,200

My name is Corey Sanders

3

00:00:06,200 --> 00:00:07,366

and I'm here to give you a quick

4

00:00:07,366 --> 00:00:08,466

walkthrough on the new

5

00:00:08,466 --> 00:00:11,233

ARENA experience on the CoreWeave cloud.

6

00:00:11,233 --> 00:00:12,033

Now, it started by me

7

00:00:12,033 --> 00:00:13,433

receiving an invitation,

8

00:00:13,433 --> 00:00:14,600

from the solution architects

9

00:00:14,600 --> 00:00:15,366

at CoreWeave

10

00:00:15,366 --> 00:00:17,266

who had set up an environment for me

11

00:00:17,266 --> 00:00:19,233

to be able to do some of these tests

12

00:00:19,233 --> 00:00:21,266

and an example walkthroughs on the,

13

00:00:21,266 --> 00:00:22,800

the cloud that's been

14

00:00:22,800 --> 00:00:23,933

that's been made for me.

15

00:00:23,933 --> 00:00:25,100

And so it started with me

16

00:00:25,100 --> 00:00:26,066

to be able to log

17

00:00:26,066 --> 00:00:27,600

right in here into the console.

18

00:00:27,600 --> 00:00:28,466

And so I logged in,

19

00:00:28,466 --> 00:00:31,133

and you can see here

20

00:00:31,133 --> 00:00:32,433

I have a cluster already up and running

21

00:00:32,433 --> 00:00:33,300

that's been set up for me,

22

00:00:33,300 --> 00:00:34,066

which is lovely.

23

00:00:34,066 --> 00:00:36,866

This is a, Kubernetes cluster running

24

00:00:36,866 --> 00:00:38,633

at the CoreWeave Kubernetes Service.

25

00:00:38,633 --> 00:00:40,300

And on top of this,

26

00:00:40,300 --> 00:00:41,500

they've also set up

27

00:00:41,500 --> 00:00:43,033

a SUNK environment for me,

28

00:00:43,033 --> 00:00:43,900

which we'll go through

29

00:00:43,900 --> 00:00:44,533

a little bit later,

30

00:00:44,533 --> 00:00:46,633

which is a Slurm solution.

31

00:00:46,633 --> 00:00:48,133

It's very unique on CoreWeave

32

00:00:48,133 --> 00:00:49,300

that allows me to do

33

00:00:49,300 --> 00:00:50,466

Slurm operations

34

00:00:50,466 --> 00:00:52,900

on top of a Kubernetes based

35

00:00:52,900 --> 00:00:55,100

infrastructure environment.

36

00:00:55,100 --> 00:00:56,433

And so the only really thing

37

00:00:56,433 --> 00:00:57,200

I need to do to be able

38

00:00:57,200 --> 00:00:58,233

to get access here

39

00:00:58,233 --> 00:01:00,500

is to bring a public key

40

00:01:00,500 --> 00:01:01,800

in with the private key,

41

00:01:01,800 --> 00:01:03,200

on my local machine.

42

00:01:03,200 --> 00:01:05,066

So that I can access the environment.

43

00:01:05,066 --> 00:01:07,000

And so I've already created a key.

44

00:01:07,000 --> 00:01:09,100

And so I just want to,

45

00:01:09,100 --> 00:01:11,600

get the public key out, here.

46

00:01:11,600 --> 00:01:12,633

And so I'm going to copy

47

00:01:12,633 --> 00:01:13,233

and paste it,

48

00:01:13,233 --> 00:01:16,733

and go right back into the console.

49

00:01:16,966 --> 00:01:18,100

It's just down here.

50

00:01:18,100 --> 00:01:19,400

If I click on my name and click

51

00:01:19,400 --> 00:01:21,200

settings and scroll down,

52

00:01:21,200 --> 00:01:22,266

you can see here

53

00:01:22,266 --> 00:01:23,866

SSH Public Key for Slurm.

54

00:01:23,866 --> 00:01:26,933

Type that in save it and away I go.

55

00:01:27,200 --> 00:01:28,700

And that's it for me

56

00:01:28,700 --> 00:01:30,566

in the console for now.

57

00:01:30,566 --> 00:01:33,733

Now once I get that key up and running,

58

00:01:33,733 --> 00:01:36,800

the next big step is I could go in

59

00:01:36,800 --> 00:01:38,966

and actually SSH in to the machine,

60

00:01:38,966 --> 00:01:41,366

into the login node here for my

61

00:01:41,366 --> 00:01:42,533

for my Slurm environment.

62

00:01:42,533 --> 00:01:43,766

So here you can see I'm logged

63

00:01:43,766 --> 00:01:44,400

in, gives me

64

00:01:44,400 --> 00:01:45,433

a little bit of detail of what

65

00:01:45,433 --> 00:01:46,166

the CoreWeave Slurm

66

00:01:46,166 --> 00:01:48,066

HPC cluster looks like,

67

00:01:48,066 --> 00:01:49,500

the things that have been pre-installed

68

00:01:49,500 --> 00:01:50,300

for me and set up,

69

00:01:50,300 --> 00:01:52,366

you can sort of see I'm running here,

70

00:01:52,366 --> 00:01:53,266

but instead of actually

71

00:01:53,266 --> 00:01:54,233

doing this through SSH,

72

00:01:54,233 --> 00:01:55,366

here's what I'm going to do.

73

00:01:55,366 --> 00:01:55,800

I'm actually

74

00:01:55,800 --> 00:01:57,466

going to run this through marimo

75

00:01:57,466 --> 00:01:58,866

because it helps me understand

76

00:01:58,866 --> 00:02:00,333

what I'm doing step by step

77

00:02:00,333 --> 00:02:01,566

and actually be able to,

78

00:02:02,933 --> 00:02:03,933

run the command

79

00:02:03,933 --> 00:02:04,966

directly from that environment.

80

00:02:04,966 --> 00:02:06,666

So let me spin up a marimo

81

00:02:06,666 --> 00:02:07,566

deployment here.

82

00:02:07,566 --> 00:02:08,566

It's going to be running

83

00:02:08,566 --> 00:02:11,266

on that login node. Okay.

84

00:02:11,266 --> 00:02:13,133

And so let me just copy that

85

00:02:13,133 --> 00:02:16,133

and let me open back up into,

86

00:02:16,233 --> 00:02:17,933

my console here.

87

00:02:17,933 --> 00:02:19,466

Bingo. Bango.

88

00:02:19,466 --> 00:02:22,100

And I have my marimo environment up

89

00:02:22,100 --> 00:02:23,266

and running

90

00:02:23,266 --> 00:02:25,366

as part of my SUNK deployment

91

00:02:25,366 --> 00:02:26,566

on the CoreWeave cloud.

92

00:02:26,566 --> 00:02:27,066

Okay,

93

00:02:27,066 --> 00:02:29,800

so let me just run my prerequisites here.

94

00:02:29,800 --> 00:02:30,633

I just have to make sure

95

00:02:30,633 --> 00:02:32,366

bash execution can happen.

96

00:02:32,366 --> 00:02:34,033

And so here we are. Welcome to SUNK.

97

00:02:34,033 --> 00:02:35,733

This allows me to play

98

00:02:35,733 --> 00:02:36,700

with the environment

99

00:02:36,700 --> 00:02:37,600

for SUNK deployment

100

00:02:37,600 --> 00:02:39,000

that's been set up for me.

101

00:02:39,000 --> 00:02:40,533

Within seconds,

102

00:02:40,533 --> 00:02:42,366

you can see I'm up and running

103

00:02:42,366 --> 00:02:43,966

and ready to run some commands.

104

00:02:43,966 --> 00:02:45,733

So the first thing you know, I can see.

105

00:02:45,733 --> 00:02:47,400

Hey, what's my storage look like?

106

00:02:47,400 --> 00:02:48,866

Running these commands

107

00:02:48,866 --> 00:02:50,000

give me exactly storage

108

00:02:50,000 --> 00:02:50,833

that's been set up for me

109

00:02:50,833 --> 00:02:51,833

in the environment.

110

00:02:51,833 --> 00:02:54,933

I can see what resources do I have?

111

00:02:55,066 --> 00:02:56,566

Right. And what's the defaults here?

112

00:02:56,566 --> 00:02:57,766

So I've got the partition

113

00:02:57,766 --> 00:02:58,833

be all partition.

114

00:02:58,833 --> 00:03:01,200

And it gives me four nodes.

115

00:03:01,200 --> 00:03:02,900

These are H100s

116

00:03:02,900 --> 00:03:03,800

which will be, you know, pretty

117

00:03:03,800 --> 00:03:05,866

nice to me to run a couple of jobs on.

118

00:03:05,866 --> 00:03:07,433

And so I can even go into some details

119

00:03:07,433 --> 00:03:08,700

on those partitions

120

00:03:08,700 --> 00:03:09,333

and take a look

121

00:03:09,333 --> 00:03:11,466

and see things like default. Yes.

122

00:03:11,466 --> 00:03:13,166

And so what other other

123

00:03:13,166 --> 00:03:14,366

sort of configurations

124

00:03:14,366 --> 00:03:16,866

have been set for me ahead of time?

125

00:03:16,866 --> 00:03:17,933

And so this makes,

126

00:03:17,933 --> 00:03:19,200

of course, the running of these jobs

127

00:03:19,200 --> 00:03:20,200

so much easier.

128

00:03:20,200 --> 00:03:22,466

I'm interacting with Slurm right now.

129

00:03:22,466 --> 00:03:24,833

And so I can go in and take a deeper

130

00:03:24,833 --> 00:03:26,866

look here at some of the configuration.

131

00:03:26,866 --> 00:03:28,266

And so this is sort of one

132

00:03:28,266 --> 00:03:29,400

of the many configuration files,

133

00:03:29,400 --> 00:03:30,600

but this is a key one for Slurm.

134

00:03:30,600 --> 00:03:34,066

And so you can see things like, preempt

135

00:03:34,066 --> 00:03:35,666

type es for policy.

136

00:03:35,666 --> 00:03:37,000

So you can understand

137

00:03:37,000 --> 00:03:37,466

sort of

138

00:03:37,466 --> 00:03:38,700

what, you know,

139

00:03:38,700 --> 00:03:41,166

what's expected for preemption.

140

00:03:41,166 --> 00:03:43,833

You can, as jobs are sort of prioritized.

141

00:03:43,833 --> 00:03:45,200

A whole bunch of different settings

142

00:03:45,200 --> 00:03:45,766

and controls

143

00:03:45,766 --> 00:03:46,666

here around

144

00:03:46,666 --> 00:03:47,433

sort of how you're going

145

00:03:47,433 --> 00:03:49,033

to run your Slurm job

146

00:03:49,033 --> 00:03:49,900

on this environment.

147

00:03:49,900 --> 00:03:50,766

I'm going to leave it all

148

00:03:50,766 --> 00:03:52,966

as it has been set for me.

149

00:03:52,966 --> 00:03:55,366

But I could go in and modify, right?

150

00:03:55,366 --> 00:03:56,500

I could even create,

151

00:03:56,500 --> 00:03:58,033

you know, in my own little,

152

00:03:58,033 --> 00:03:59,833

window here and cat out,

153

00:03:59,833 --> 00:04:01,600

do it, you know, do an SSH

154

‍