ACG HotSeat: Juniper Networks Sr. Director Ben Baker

0:00 [Music]

0:05 hello

0:05 this is ray motor from acg and welcome

0:08 to this edition of the hot sea

0:10 joining me today is ben baker from

0:11 juniper bend thanks for joining the hot

0:13 seat

0:13 thanks right good to see you and excited

0:15 to talk about this project

0:17 excellent excellent now maybe before we

0:19 get started could you tell our audience

0:21 a little bit about

0:22 what you do with uh juniper currently

0:24 today

0:25 hey i'm the senior director of cloud

0:27 solutions marketing

0:29 excellent now some time ago there was a

0:32 project that was involved you know

0:34 related to

0:35 uh cloud economics assessment model what

0:38 is the motivation behind this project

0:40 yeah i think it's one of the existential

0:43 business

0:44 questions of our time right it's you

0:45 know where do i run my applications in

0:48 my it workloads in in a public cloud or

0:51 my own data center

0:52 and virtually every company in the world

0:55 wrestles with this question you know

0:56 whether it's a company that we might put

0:58 a label on as

0:59 an enterprise or a cloud service

1:01 provider or a communication service

1:03 provider

1:04 every company out there is is running

1:06 software somewhere

1:08 so we've been working with our customers

1:10 over the world

1:11 for years to tackle this question then

1:13 you know i hear it all the time from

1:15 customers directly

1:16 hear it from our account teams non-stop

1:19 now you know the economics of

1:22 public private and hybrid clouds is

1:25 complicated

1:26 it's hard and we've found that it's not

1:28 broadly

1:29 well understood and you know there are

1:32 other

1:33 out there let's call them cloud

1:34 calculators um you know you can search

1:36 for the web right now and find them

1:38 but these are way too simplistic and

1:40 they're they're blunt

1:41 tools and you know they're just not

1:43 going to give you the insight that you

1:45 really need to make good decisions and

1:46 then

1:47 of course the public cloud providers

1:49 have cloud calculators too but

1:51 we found that people don't necessarily

1:53 trust these and

1:55 also again they don't have the

1:56 granularity and you know you can't look

1:58 under the hood of those models and

2:00 really understand the calculations um

2:03 so companies are making big time

2:06 strategic decisions that can

2:08 fundamentally affect the economics of

2:10 their business for years to come

2:12 and they're doing this with incomplete

2:13 information you know without sufficient

2:15 rigor behind the decision

2:17 and this is why we came to you guys this

2:19 is why we came to acg last year and said

2:22 we need to solve this problem hey help

2:24 us build a model for this

2:25 yeah no that's a good point because it's

2:28 it's interesting because sometimes some

2:29 of these trivial

2:31 roi calculators just isn't enough to

2:34 make a decision yet you're making

2:36 millions or in some cases billions of

2:37 dollar decision

2:39 without having some of the macro level

2:42 the major disconnect there

2:44 so so that being said what is the cloud

2:46 economics assessment model

2:49 yeah we call it uh you know as you say

2:50 the cloud economics assessment model or

2:53 cloud workload decision model and you

2:55 know to be really specific

2:56 it's a detailed total cost of ownership

2:59 model in excel

3:00 um it it compares the monthly expenses

3:02 of

3:04 10 different application areas if you

3:06 run these apps on

3:08 aws versus azure versus your own data

3:11 center

3:11 or any combination the model is very

3:14 comprehensive

3:16 it includes every cost imaginable you

3:18 guys did a great job with this

3:20 we've tested the model with a variety of

3:22 smes

3:23 and you know you might say well wait a

3:25 second if it's comparing monthly

3:26 expenses then

3:28 it doesn't include the capex that you

3:30 might need for your own data center um

3:32 but it actually does include capex and

3:35 you know we make

3:36 all costs variable so you can do that

3:39 side-by-side comparison between data

3:42 center

3:42 infrastructure you own and whether

3:44 that's on premises

3:46 or a few racks that you have in some

3:48 colo you know

3:49 versus a public cloud which of course is

3:52 is all opex and

3:53 all of a variable cost and you know of

3:56 course there are some things that you

3:58 just can't model

3:59 like security and compliance and

4:01 regulations

4:03 and some companies out there may for

4:05 example just want more control over

4:07 their it's

4:08 so there are some i guess non-cost

4:10 factors that go into this decision but

4:13 we think the model strikes a great

4:16 balance of having

4:17 an incredible amount of granularity and

4:20 flexibility

4:21 but you know shielding that complexity

4:23 from the user of the model

4:25 so you know it has literally hundreds of

4:28 knobs and dials that you can turn and

4:30 change the input variables but at the

4:32 same time

4:33 we do have well thought out default

4:35 values for for everything also

4:37 yeah that's great to have default values

4:40 so that people have a baseline

4:42 of of where to start with i remember

4:44 peter federoff working on this

4:46 uh and he was going over some of the

4:48 results and the complexity of the work

4:51 behind the scenes and this but um but

4:54 based on your experience and some of the

4:55 things you saw what are some of the

4:58 results or how would you interpret

5:00 some of the outcomes from those models

5:01 there

5:04 yeah at a high level the answer is the

5:07 multi-cloud

5:08 you know the optimal place to run some

5:10 applications

5:11 is public cloud other applications in

5:14 your own data center

5:15 the private cloud and and the model

5:17 shows for example that you can save

5:18 up to 90 if you take this hybrid

5:21 approach

5:22 versus if you just put everything in in

5:24 the public cloud

5:25 right and you know so so everyone knows

5:28 that you use public cloud for speed and

5:30 flexibility right well

5:32 well you know well kind of i mean public

5:34 cloud is is very useful for bursts

5:36 so flexibility and agility you know

5:39 particularly in the context of short

5:42 time horizons so you know let's say a

5:45 pandemic hits

5:46 right you know just just hypothetically

5:47 of course and you know the demand for

5:49 your

5:50 video conferencing app goes through the

5:52 roof you know

5:53 you can really quickly ramp up on public

5:56 cloud

5:56 um but it would take you longer to

5:59 install your own

6:00 dc infrastructure so public cloud for

6:02 bursts

6:03 and then your own dc for you know

6:06 long-term

6:07 sustain sustainable growth and scale and

6:10 efficiency

6:12 um now going back to the model uh you

6:15 know in terms of

6:16 different applications so in general you

6:19 want to use

6:20 public clouds for apps that have modest

6:23 compute

6:23 memory and network requirements like web

6:26 servers

6:27 development environments some testing

6:29 environments

6:30 low priority business apps or apps that

6:34 only require

6:35 intermittent access to compute resources

6:38 you know so the compute resources would

6:39 be idle

6:40 for a large percentage of the time uh

6:42 also

6:43 for public cloud it fits applications

6:46 that run in batch mode

6:48 and can be scheduled um applications

6:51 that you could run

6:53 at very flexible times so these apps

6:55 could use spot instances

6:57 and aws and azure with spot instances

7:00 they're essentially

7:02 spare capacity that uh they price very

7:04 very cheaply

7:06 now on the other side uh the apps and

7:09 use cases where a privately owned data

7:12 center makes sense

7:13 are use cases where you have high

7:15 network traffic

7:16 so nfv you know vnf's virtual network

7:19 functions

7:20 uh gaming um video you know as long as

7:23 it's not pushed out to the edge

7:26 so the model shows that applications

7:28 with

7:29 large data transfer requirements can be

7:32 up to 10 times more

7:34 expensive on public cloud um

7:37 some other apps that uh make sense to

7:40 run

7:40 in your private dc apps with

7:43 non-traditional

7:44 compute requirements like ai ml

7:46 workloads on gpus

7:48 so gpu instances very expensive on

7:52 public cloud

7:53 uh and then finally use cases that have

7:55 consistently high compute requirements

7:58 you know the business case for these

7:59 types of use cases

8:00 is going to point you to investing in

8:02 your your own data center so

8:04 you know i i went through a bunch of

8:07 general statements there about the

8:08 results but the important point is that

8:10 we

8:11 tweak the model for a customer's

8:12 specific environment

8:14 to give you detailed quantitative

8:15 answers yeah no i actually i think you

8:17 did a good job of explaining not that

8:20 people should get rid of public cloud is

8:21 that there's a need and a fit

8:23 and understand those inflection points

8:26 based on your capacity demand because it

8:28 impacts a tremendous amount

8:30 of the economics with that so so that

8:33 being said is uh where do people go

8:35 and how do people find out more about

8:38 this

8:40 uh yeah so one of the first things i do

8:41 is uh you know i'd recommend that you

8:43 read the white paper

8:44 that you that acg has written um okay

8:46 you should be able to find this

8:48 juniper acg cloud economics white paper

8:51 and it should pop up

8:52 and this paper should reinforce

8:55 everything that we've talked about here

8:57 um and and then usually the best thing

8:59 to do is

9:00 contact your juniper account manager and

9:02 we can start the conversation

9:04 and you know ultimately what you want to

9:06 have happen with a model like this

9:08 is that we work closely with you to

9:11 customize the model for your own

9:12 environment

9:13 because the answer is always some form

9:15 of it depends right

9:17 and so we sit down with you yeah you

9:20 know um

9:22 it wouldn't it be simple if we just said

9:23 it depends and walk away

9:25 the answer is it depends and we have a

9:27 model so

9:28 you know so we sit down with you uh

9:30 virtually of course

9:32 go through the model look at the

9:34 assumptions uh we say hey we think

9:36 xyz and you say okay yeah we

9:40 we agree let's run the model or you say

9:42 no

9:43 you know our web server infrastructure

9:45 we use more vm resources than you're

9:47 saying or

9:48 or less and then we you know make the

9:49 changes and you just go down through the

9:52 assumptions and the inputs

9:53 for all the applications and the

9:56 infrastructure and

9:57 and hey maybe we we changed some of the

9:58 underlying calculations too if

10:01 we think that's best and the end goal

10:04 here is that

10:05 you eventually come up with a really

10:07 sound game plan for your business that

10:09 has

10:09 discipline and formality and analytical

10:12 rigor behind it and

10:14 you know now you is one of the people

10:16 responsible at your company for making

10:18 these decisions

10:19 you have the confidence and and you have

10:21 the information necessary to execute

10:24 on that question that we started off

10:25 with and you know it's when do i use

10:27 public cloud and

10:28 when do i use my own private data center

10:30 infrastructure yeah

10:32 no excellent excellent well ben you're

10:34 officially off the hot seat

10:35 thanks for joining all right thanks ray

10:38 yeah well

10:39 with ben i'd like to thank everyone for

10:41 joining

10:42 take care bye-bye

10:46 [Music]

10:48 you