The document discusses strategies for successful implementation of hybrid cloud environments at both the organizational and application levels. At the organizational level, key strategies include defining the cloud model for the organization, enabling users to transition workloads between clouds, and providing incentives for users to place workloads where costs are lowest. At the application level, important strategies involve understanding infrastructure differences between clouds, being aware of limitations of abstraction layers, and avoiding vendor lock-in through use of open source software. The document provides examples and considerations for hybrid cloud adoption from various companies.
3. WHAT’S SCALR?
Why Talk About Hybrid Cloud?
Open-source Enterprise
Cloud Management
Platform.
All our enterprise customers
run on hybrid cloud.
• CloudStack + AWS at
Samsung
• OpenStack + AWS at
Expedia, NASA
4. And you
Who are you?
What do you do?
Image Credit: Bybzee from the Noun Project
6. DEFINE: HYBRID CLOUD
A system with components deployed
across multiple cloud platforms is said
to be using hybrid cloud.
Oftentimes, it’s one public cloud and one
private cloud (but that consideration
actually has limited implications).
Image Credit: greeblie (Flickr)
7. App Organization
At the app level, “operating on a hybrid
cloud” means you have interdependent
services that make up your app
distributed across clouds.
At the organizational level, “operating a
hybrid cloud” means you have some
resources, possibly independent,
distributed across clouds.
WHAT “SYSTEM”?
Challenges and success strategies largely
depend on what you are looking at
Image Credit: Dane Hetteix, Auda Samora from the Noun Project
8. 25% of enterprises will be
deploying IT workloads
and applications in a
hybrid cloud setup by 2015.
23% think that hybrid
cloud is the best execution
venue for cloud-native
apps (i.e. greenfield).
Private: 11%, Public: 66%.
43% think that hybrid
cloud is the best
execution venue for e-business
applications.
Private: 18%, Public: 39%.
TIME FOR STATISTICS!
Adoption exists both at the organization and app level
25% 23% 43%
Source: Forrester Research, “Cloud Wave 6” Study, 2014
9. COST OPTIMIZATION
Deploy ephemeral apps (e.g.
ad campaigns) and risky
ventures to a public cloud.
Deploy long term workloads to
a private cloud.
Or, “buy the base and rent the
spike.”
FAILOVER, DR, BACKUPS
Host failover resources and
data backups in a separate
cloud for redundancy and
possibly cost reduction.
CUSTOM SKUS
Deploy workloads that require
unusual CPU / RAM / Disk
ratios to a private cloud with
custom SKUs.
Deploy more typical
workloads to a public cloud.
LEGACY SYSTEMS
Deploy backend resources
that integrate with legacy on-premises
systems in a private
cloud (latency compliance).
Deploy frontend resources in
a public cloud.
PERFORMANCE
Achieve greater performance
by deploying to a private
cloud (e.g. smaller DC implies
lower latencies, and using
bare metal cloud is an option).
VENDOR LEVERAGE
Get a better deal with your
public cloud provider by
having an alternative.
WHY USE HYBRID CLOUD?
More reasons than you can think of! The same apply at both levels
COMPLIANCE
Deploy regulated workloads
and store regulated data in a
gold-plated secure private
cloud, use public cloud for the
rest.
AND MORE!
There are literally innumerable
use cases for hybrid cloud.
10. OPTIMIZATION
Hybrid cloud means you deploy
workloads to the cloud that is the
best suited to fit their requirements.
Key considerations often revolve
around SLAs and value / cost ratios.
Image Credit: Sabine Wollender from the Noun Project
12. USE CASE
What are we talking about here?
One application, multiple tiers distributed
across different clouds.
3 Key Success Strategies:
1. Understand infrastructure differences
2.Be mindful of leaky abstractions
3.Avoid vendor lock-in
Image Credit: Auda Samora, Callum Egan from the Noun Project
“Hybrid cloud? Everyone
was talking about it, but
no one was really doing it.”
Mark Williams, CTO of Redapt,
formerly at Zynga where he built zCloud
13. 1. INFRASTRUCTURE
“I’ll just use libcloud”
Abstraction libraries like libcloud or
fog resolve syntactic API
incompatibilities.
But don’t miss the forest for the
trees! Semantic incompatibilities
matter the most. Account for them!
E.g. provisioning times can vary
across clouds, so scaling parameters
that work well on one cloud may not
be appropriate for another one.
Image Credit: Edward Boatman, Richard de Vos from the
Noun Project
“APIs are tiny windows
into massive, complicated
systems”
Randy Bias, CTO of CloudScaling,
built Korea Telecom’s uCloud
14. 2. ABSTRACTIONS
“I’ll just use Docker”
Abstractions like Weave (a virtual network
for Docker) or a VPN let you simplify
application operations, but they are leaky
abstractions.
E.g. latency across clouds can be quite
high. Latency within a public cloud may
already be too high for your workload.
Test it!
Backend latency can quickly slow a
typical Django / Rails app to a crawl.
Don’t forget bandwidth costs, too.
Image Credit: André Raphael, Richard de Vos, Icomatic
from the Noun Project, Michael Wyszomierski (Flickr)
15. 3. LOCK-IN
“I wish I hadn’t used DynamoDB”
Managed services are the bread and
butter of cloud lock-in (e.g. DynamoDB,
SQS…)
Consider relying on trusted OSS
software instead if you intend to use
hybrid cloud (e.g. Redis, RabbitMQ)
Image Credit: Marwa Boukarim from the Noun Project,
Vinoth Chandar (Flickr)
17. USE CASE
What are we talking about here?
Entire organization. Numerous
independent apps deployed across
multiple clouds.
3 Key Success Strategies:
1. Define cloud for your organization
2.Enable users to transition across clouds
3.Encourage them with the right incentives
Image Credit: Dane Hetteix, Callum Egan from the Noun Project
18. 1. DEFINE
What cloud? Cloud-native, or just cloud?
Cloud can have many meanings, but to
freely distribute across clouds, your
organization must agree on one.
Different clouds encourage different
architectures (e.g. cloud-native “cattle
vs pets”, traditional “scale-up”).
The “big public clouds” are “elastic”,
or “scale-out” clouds.
Image Credit: Martin Delin, Icomatic from the Noun Project
Users of CloudStack should
be especially careful.
CloudStack accommodates
scale-up workloads
differently than most cloud
platforms.
19. 2. ENABLE
Support end-users that want to migrate from a cloud to
another
Most organisations start on a
public cloud and migrate back to a
private cloud later on.
End users expect a solid
experience that matches what
public clouds offer.
Image Credit: Simple Icons, Apache Software Foundation
In private cloud,
CloudStack “just works”
20. 3. ENCOURAGE
Provide the right incentives for users to place their workloads
where you want them to
To achieve organizational hybrid cloud
objectives (e.g. a cost reduction), end
users are the ones that must migrate.
Chargeback is an efficient way to
incentivize users to move to the cloud
where your costs are the lowest.
But private clouds can also typically
achieve greater performance than
public clouds.
Image Credit: TukTuk Design, hunotika from the Noun Project
Zynga was able to cut
cloud costs by 2/3 by
moving to zCloud.
Studios started moving
when staying on AWS
would negatively affect
their P&L.
22. At the organization level
CONCLUSION
What can you do?
Image Credit: Dane Hetteix, Callum Egan from the Noun Project
1. Define cloud for your organization
2.Enable users to transition across clouds
3.Encourage them with the right incentives
At the application level
1. Understand infrastructure differences
2.Be mindful of leaky abstractions
3.Avoid vendor lock-in