Edward Wustenhoff, VP of Infrastructure and Platform Services, Juniper Networks

Should Workloads Run On-Prem or In-Cloud?

I See IT Like IT Is Operations
Edward Wustenhoff Headshot
Should Workloads Run On-Prem or In-Cloud?

Got questions about workload deployments? Start here for answers.

Juniper’s Edward Wustenhoff is often asked whether to deploy workloads on-premises or in the cloud. In this video, part two of a series on IT services (see parts one and three), he takes a deep dive into the factors to consider when deciding how to successfully run a workload. Thanks to real-world, easy-to-understand examples, you’ll learn useful information to help you deliver high-quality infrastructure services that are easy to consume, delivered fast and in a timely fashion, and add value to your business. 

Show more

You’ll learn

  • The two main vectors that drive the decision as to whether to deploy on-prem or in-cloud

  • How Juniper was able to treat every home office like a branch office during the COVID-19 pandemic

Who is this for?

Network Professionals Business Leaders

Host

Edward Wustenhoff Headshot
Edward Wustenhoff
VP of Infrastructure and Platform Services, Juniper Networks

Transcript

0:00 [Music]

0:00 foreign

0:00 [Music]

0:07 when it comes to workloads I'm often

0:09 asked whether to deploy it on premise or

0:12 in the cloud

0:16 there are two main factors that drive

0:18 the decision as to where to deploy

0:20 commitment and control commitment refers

0:23 to the duration of how long the workload

0:26 will exist where control refers to how

0:28 much control you need over the compute

0:30 storage data center Network to

0:33 successfully run the particular workload

0:35 the longer you can commit and the more

0:37 control you need will drive the workload

0:40 closer to you here are a few examples of

0:42 commitment if you know a particular

0:45 workload will exist for 20 to 30 years

0:47 you could think about it like a real

0:49 estate transaction but very few

0:51 Enterprises can predict so far ahead

0:54 maybe it's not 30 years but five to ten

0:57 years at those Horizons managed services

0:59 or co-location Services might be an

1:01 option

1:02 often in these cases the decision is

1:05 more Finance driven because you can

1:07 consider buying the hardware Computing

1:09 stores and move costs from Opex to capex

1:12 however those longer term scenarios

1:15 present some drawbacks for example you

1:18 are committed losing the option to

1:20 quickly adjust up or even more

1:22 importantly down

1:24 businesses often underestimate what it

1:27 will take to operate manage the

1:29 availability security performance

1:31 economics and change of this type of

1:33 infrastructure and often not recognized

1:36 how many in-house compute and storage

1:38 teams do you know of that deliver new

1:40 features every three months like other

1:43 service providers

1:45 on the other side of this argument what

1:48 if the workload is going to exist only

1:50 for a few days or even less

1:52 it is clear that those types of

1:54 workloads can be run further away from

1:56 you in a cloud model where you consume

1:59 compute storage Data Center and to a

2:00 large extent Network as a service

2:03 some of the key advantages are agility

2:07 real example at some point I wanted to

2:10 understand what the user experience

2:12 would look like in the asia-pacific

2:14 region if I would host my services in

2:16 Singapore

2:17 we spun up a clone of the service in the

2:20 cloud in a couple of weeks costing about

2:22 ten thousand dollars

2:25 to find out it was a total fire failure

2:29 in the old days it would have taken 6 to

2:31 12 months at least to get a data center

2:33 buy and install the equipment and get

2:35 the services up and running all to find

2:38 out it would fail

2:40 another Financial aspect I'd like to

2:42 call out is that if I don't use cloud

2:45 services my cost is literally zero this

2:48 cannot be said if you are in a 10-year

2:50 contract and a five-year depreciation

2:52 cycle

2:54 the others say

2:56 control the level of control you need

2:58 over the underlying compute storage data

3:01 center Network Services also drives

3:03 where you put the workload here are some

3:06 examples of control that would drive the

3:09 infrastructure closer towards you

3:11 the most common reason is because the

3:13 workload can't be moved due to technical

3:16 limitations this is why enabling your

3:19 applications to be cloud ready is so

3:22 important because even when you decide

3:24 not to run it there at least you have

3:26 the option to do so

3:28 security or compliance regulations is

3:31 another key driver if you cannot share

3:33 your workload an Associated data with

3:36 anybody

3:37 you will end up having to control a lot

3:39 more of the stack

3:41 performance I once had to keep a

3:44 workload in-house as the code was so

3:46 performance sensitive it would only run

3:48 in a specific type of Chip and the scale

3:51 made it cost prohibitive to run

3:53 sub-optimized

3:55 other considerations might be

3:57 competitive or IP related that would

4:00 disqualify having the workload and its

4:03 data with a certain vendor or outside of

4:05 a building that you control

4:08 the good news is that with today's

4:10 diversity of service providers products

4:12 and models you can find almost always

4:15 exactly what you need

4:16 there are many options that support

4:18 different levels of commitment and

4:20 control

4:21 at Juniper we use a combination of these

4:24 models and we enable ourselves to do so

4:27 by building the network infrastructure

4:29 to support the many different scenarios

4:32 so let me tell you a little bit more

4:33 about what we did

4:35 we deployed several points of presence

4:37 across the world

4:39 independently but in strategic proximity

4:42 with strong network access capabilities

4:44 to Juniper's corporate Network local

4:47 networks and cloud service providers

4:50 in our Branch offices we use jssr SRX

4:54 and ap12s and 45s to leverage local

4:57 connections at low cost using our sd-wan

5:00 Technologies

5:02 over the last five years we almost got

5:04 our Telco spending in half by doing so

5:07 during covet this helped us a lot as we

5:10 could treat every home office like it

5:12 was a branch office

5:13 and like many of us we did have to

5:16 manage our VPN handling capacity at the

5:19 Pops of course

5:20 we provide Network proximity to cloud

5:22 service providers in these pops

5:24 leveraging our qfx SRX and vsrx product

5:28 capabilities

5:30 this has enabled us with a whole new

5:32 level of agility

5:34 for example we built a dedicated lab

5:36 connection securely with one of our

5:39 cloud service providers in a matter of

5:41 days

5:42 I hope this information is useful to you

5:44 in your quest for delivering

5:46 infrastructure services that add value

5:48 to your business are of high quality

5:50 easy to consume and delivered Fast And

5:53 Timely

5:54 if you want to learn more please reach

5:57 out to us by contacting sales let's make

6:00 it happen

6:00 [Music]

Show more