OpenStack is not product, it’s a process. Building an OpenStack cloud should never be the beginning nor the end of a cloud project. This is so because:
OpenStack is not a product that tries to solve any one business problem in particular, but a technology that aims to be usable by as many as possible
Since the beginning of computing, projects which include a culture shift require the for the enterprise processes to evolve too, and this is often the hardest part of a cloud project
OpenStack is a growing technology, as are the requirements of users, and if you do not plan to make your cloud continuously evolve, you are as good not doing an enterprise cloud project
Based on our experience delivering cloud project to enterprises, this presentation will give some key learnings on what are the required steps to make sure that your enterprise is ready for the OpenStack transformation.
3. About me
● VP Products & Pre-Sales @ eNovance
● Been working on OpenStack since Austin
● Founded OpenStack Telemetry (aka Ceilometer)
● Travelling the world doing OpenStack
● @nijaba on irc & twitter
4. Disclaimer
● No enterprises were harmed in the making of this
presentation
● This presentation may not only apply to OpenStack, but to
most tools that one can use that deliver an API based cloud
IaaS
6. Is OpenStack a Product?
❏ Yes
✓ No. It's a technology
framework.
7. OpenStack is not a product...
● Comparing products with technologies with products is like
comparing an engine with a car
● You have to keep focusing on your use cases
● You either need to integrate OpenStack with other
components or get a productized OpenStack from a vendor
8. OpenStack is not a product...
OpenStack will never be ready for the
enterprise!
9. OpenStack is not a product...
Yet enterprises may want an OpenStack
based cloud...
10. Why are enterprises interested in cloud?
● Save money
○ Reduce infrastructure costs
○ CAPEX <-> OPEX flexibility
● Make money
○ Reduce time to market
○ Enable new initiatives
○ Achieve bigger scale plans
11. Is free software a good enough reason?
“Free Software is gratis, right?
We’ll save on licensing costs!”
12. Is free software a good enough reason?
Are licensing costs a good enough reason to switch to
OpenStack?
❏ Yes
❏ No
13. VMWare ESX recurring costs
● Licensing costs are only a
small portion of overall cost
● Operation costs are linked to
human salaries to operate the
DC
⇒ Humans costs are the largest
part
Analysis done by a large customer using ESX on his recurring costs. This may greatly vary depending on location
14. Is free software a good enough reason?
● Free software still has costs
○ Maintenance (subscription or internalized)
○ Expertise (experts are not cheap)
○ Knowledge (your teams have to learn new skills)
15. Is free software a good enough reason?
Are licensing costs a good enough reason to switch to
OpenStack?
❏ Yes
✓ No, it’s only marginal
17. Good reasons for Free Software
● Real good reason for free software:
○ Adaptability (you can adapt or complete the software)
○ Security (you can control what’s in it)
○ Sharing (so you can split costs with others)
○ Standard (so you can be sure it interoperates)
○ Hybridation (your providers works the same way)
○ etc...
18. Good reasons for OpenStack
Benefits should not only come from the IaaS
itself….
… but from the transformation it enables!
19. Transformation enablement examples
● Enable business units to be more reactive in a faster market
● Enable development teams to be more productive and
autonomous
● Enable (cloud ready) applications to be more scalable
● Enable more creativity
● Enable finer grain cost analysis and responsibilization
⇒ Be quicker, better, smarter to success!
20. OpenStack is the transformation?
“So, I just have to install OpenStack and I
am done?”
21. OpenStack is the transformation?
● OpenStack is just a technology, a tool to enable the
transformation
● OpenStack is an agility enabler
○ For your business units
○ For your development teams
○ For your marketing
⇒ But only if you change the ways they interact with the
data center
22. Transforming the interactions: BUs
Business units before
● Tell IT what they need
● Wait for approval
● Wait for IT to complete the need
● Complain when it fails
● Rely on IT for application SLA
● Are restrained by global policies
that should not apply to them
Business units transformed
● Iterate on their needs
● Own the application they use
● Quickly launch MVPs to test
● Operate their applications
● Rely on IT for data center SLA
● Can adapt policies to match their
market needs
23. Transforming the interactions: Developers
Developers before
● Receive lengthy specification and
write code to match their
understanding of the problem
● Fight for hardware allocation
● Transfer operational burden as
quickly as possible
● Are not responsible of SLA
● Rely on global data store to be
globally available
Devops
● Closely interact with stakeholders
● Iterate quickly to validate/adapt
● Are responsible for hw allocation
● Same team operate the
applications they developed
● Build SLA into the application
● Build hybridation into the
application
24. Transforming the interactions: Marketing
Marketing before
● Tell IT what they need
● Subcontract wildly to agencies
● Complain when it fails
● Don’t understand delays
● Always buy advertising space and
create unrealistic hurdles
● Get tired of waiting and go AWS
Marketing transformed
● Allocate their own hardware
● Provide access to agencies
● Are responsible of their operations
● Fight with their own delays
● Find someone else to hurry for
their crazy plans
● Are finally traceable in their spend
25. Transforming the interactions
OpenStack enables you to transform:
● Complainers → Creators
● Consumers → Operators
● Developer divas → Devops
● Large mammoth → Lean organizations
But you still have to operate the transformation!
26. Raising the abstraction level
Computing could be summarized a tool which main purpose is
to raise abstraction levels
Binary → Assembly → Language A → Language B…
and is using it to fight against its own gravitational laws for
scaling
Mainframe → PC → Client Server → 3 Tier → Distributed → Hybrid
Single Tasking → Multi Tasking → Multi Threading → Multi Core → Multi Node
27. Cloud is just one more abstraction
Cloud Infrastructures are just one more abstraction
enabling the handling of massive number of nodes
as if they were a simple entities (aka pet vs. cattle)
28. Abstraction → Transformation
To benefit from the abstraction you have to transform your tools
● Central DBs → Distributed DB (noSQL)
● Vertical Scaling → Horizontal Scaling
● Filesystems → Blocks and Objects
● FIFO → Bus
● Specialized → Commoditized
● Unique → Pattern based
● Sequential → Parallelized
OpenStack is made for the abstraction to happen
29. Transformation vs Migrations
OpenStack is made for the abstraction to happen and this is
why it is not meant to handle your old workloads
● Migrations still have a cost
● Without adaptation it does not bring any benefits
● You need to accelerate, not just create more work
● Lean enablement cannot be done with bloated software
30. A possible way forward
● Start using your cloud with news apps or components
○ Convert your apps calmly, within their normal life cycles
○ You can mix and match computing models to proceed in
small shorter steps
○ Not all need to be cloud at once
⇒ then you should start seeing real benefits of OpenStack
31. A possible way forward
● Make hybrid apps
○ enable front ends to scale on multiple sites
○ enable data to be maintained on multiple locations
⇒ which in turn will provide the benefits of real hybrid
deployments that OpenStack enables
32. A possible way forward
● Change the way your IT teams work:
○ each group should be responsible of delivering its
resources globally
○ stop dividing those who operate from those who design
○ stop defining silos in terms of responsibility but in term of
APIs and measurable objectives to maintain
○ Stop the pre control craziness
■ trust is the first enabler of delegation
■ which leads to responsibility, autonomy and creativity
○ Measure end goals, not milestones
⇒ this is the only thing that matters
33. Summary
● Openstack is not a product
● It will never be “enterprise ready” by itself
● License costs are marginal
● Concentrate on enabling better processes
● and better tools
● Transforming your enterprise to produce more value
34. THANK YOU
Nick Barcet | nick@enovance.com | @nijaba | Juno OpenStack Summit | May 2014