What’s Beyond Virtualization?
Derek Collison - @derekcollison
Apcera Inc.
QCon London - March 7, 2014
What does the future of
Enterprise IT look like?
What does it look like
today?
It Depends!
Could be very old school - Just physical machines!
Could be virtualized: CPU/MEM, Storage and Network
Could be IaaS or Iaa...
Why do we care?
What do these things
really buy us?
They remove
undifferentiated heavy lifting!
They speed up
the slow and mundane
Transparent value-add!
What about PaaS?
What about PaaS?
Tries to speed up deployment
Very opinionated, still is..
Is only a piece of the larger puzzle
Carrying c...
So..
What do we want?
What do we want..
Self Service
Faster iterative development and deployments
Fault Tolerance, High Availability
Higher and ...
Software Defined Everything!
But..
What do we really want?
What do we really want?
Meta-data driven description of my system
Extreme Agility
Transparent Compliance
Fluid and Abstrac...
What do we want really?
Meta-data driven description of my system
Extreme Agility
Transparent Compliance
Fluid and Abstrac...
Meta-Data Driven
My app A needs to talk to B and C
I need 4 instances of A, 2 of B and 3 of C
It needs X memory and Y CPU
...
Is this
Autonomic Computing?
How would we do this?
How would we even start?
Getting Started - App A
What does App A need?
Where will App A be run?
How will App A find B and C?
How do others find my ...
Getting Started - App A
What does App A need? - Packaging and Dependencies
Where will App A be run? - Provision and Schedu...
Packaging and Dependency
What does the job need to run?
What runtimes, OS, libraries?
What tools can I use for consistency...
Packaging and Dependency
Challenges
Do these change when I change from Dev to
Production?
What runtimes, OS, libraries the...
Provision and Schedule
How fast can I provision?
Can my workload run anywhere and be compliant?
How do network perimeter s...
Addressing and Discovery
Is DNS sufficient?
Do we need to change our applications?
When things get moved, how does the sys...
Monitoring and Management
What happens when something fails?
Is this a manual process?
Who determines failure? Can we trus...
SO..
How do we get here?
Is it a Bolt-On Solution?
Or is it Bolt-In?
Bolt-On got us into this
mess in the first place!
What we need is a
Platform OS!
Programmable, pluggable
and composeable..
From the inside out..
The OS for the datacenter
The OS for the datacenter
Multiple Datacenters
Secure, Trusted, and Hybrid
Multi-Datacenter OS
Treat all resources as a common pool
Handle all networking access, addressing and discovery in
realtim...
Multi-Datacenter OS
Virtualization
SDN - Software-Defined Networking
Management and Resource Pooling
Intelligent and Compl...
Multi-Datacenter OS
Virtualization
SDN - Software-Defined Networking
Management and Resource Pooling
Intelligent and Compl...
Virtualization?
What about speed and weight?
Google chargeback diversion
What about containers, e.g. Docker?
Is there a co...
Multi-Datacenter OS
Virtualization
SDN - Software-Defined Networking
Management and Resource Pooling
Intelligent and Compl...
SDN?
Solve network perimeter security?
Does it involve application level changes?
What about layer 7 semantics?
How many I...
Multi-Datacenter OS
Virtualization
SDN - Software-Defined Networking
Management and Resource Pooling
Intelligent and Compl...
intelligent and Compliant
Job Scheduling
Pick the best place to run for a given job and policy
How does a system rebalance...
Multi-Datacenter OS
Virtualization
SDN - Software-Defined Networking
Management and Resource Pooling
Intelligent and Compl...
Intelligent Canarying
Want to roll out a new version of App A
Do we know what App A - v2 success looks like?
How do we do ...
Intelligent Canarying
What data is needed to say if it is ok?
resource utilizations - CPU, Mem, Storage
communication patt...
Summary
Summary
Intelligent, holistic platform technologies - Pluggable and Composeable
Transparent value add to jobs/workloads - ...
Some Resources
Docker - https:/
/www.docker.io/
Mesos - http:/
/mesos.apache.org/
CoreOS - https:/
/coreos.com/
Fleet, Etc...
Thank You
Upcoming SlideShare
Loading in...5
×

What's Beyond Virtualization? - QCon London 2014

2,337

Published on

What lies beyond virtualization? How do we start the journey to a secure, trusted hybrid platform that truly delivers the business value and velocity we all want?

In the era of software-defined everything, one goal is to reach a fluid infrastructure that has the level of plasticity needed to self heal itself and provide higher level SLAs for applications and services. Adding value to existing applications and services in a transparent fashion requires a rethinking of core technologies in the platform space. In this talk we will take a look at some low level technologies and approaches to achieving this goal. Topics will range from Intelligent layer 7 SDN with semantic awareness, distributed scheduling algorithms, policy distribution and invalidation, health monitoring and management, self healing techniques, and the role of unsupervised deep machine learning and anomaly detection.

Published in: Technology
0 Comments
8 Likes
Statistics
Notes
  • Be the first to comment

No Downloads
Views
Total Views
2,337
On Slideshare
0
From Embeds
0
Number of Embeds
4
Actions
Shares
0
Downloads
47
Comments
0
Likes
8
Embeds 0
No embeds

No notes for slide

What's Beyond Virtualization? - QCon London 2014

  1. 1. What’s Beyond Virtualization? Derek Collison - @derekcollison Apcera Inc. QCon London - March 7, 2014
  2. 2. What does the future of Enterprise IT look like?
  3. 3. What does it look like today?
  4. 4. It Depends!
  5. 5. Could be very old school - Just physical machines! Could be virtualized: CPU/MEM, Storage and Network Could be IaaS or IaaS++ Could be IaaS + PaaS Could be off-premise and in the Cloud
  6. 6. Why do we care?
  7. 7. What do these things really buy us?
  8. 8. They remove undifferentiated heavy lifting!
  9. 9. They speed up the slow and mundane
  10. 10. Transparent value-add!
  11. 11. What about PaaS?
  12. 12. What about PaaS? Tries to speed up deployment Very opinionated, still is.. Is only a piece of the larger puzzle Carrying costs in the delivery pipeline Pipeline is a biz requirement thru value delivered BUT PaaS as a standalone technology is not enough..
  13. 13. So..
  14. 14. What do we want?
  15. 15. What do we want.. Self Service Faster iterative development and deployments Fault Tolerance, High Availability Higher and guaranteed SLAs Composeable Systems - legos
  16. 16. Software Defined Everything!
  17. 17. But..
  18. 18. What do we really want?
  19. 19. What do we really want? Meta-data driven description of my system Extreme Agility Transparent Compliance Fluid and Abstracted Infrastructure and Services Multiple delivery models in a single system
  20. 20. What do we want really? Meta-data driven description of my system Extreme Agility Transparent Compliance Fluid and Abstracted Infrastructure and Services Multiple delivery models in a single system
  21. 21. Meta-Data Driven My app A needs to talk to B and C I need 4 instances of A, 2 of B and 3 of C It needs X memory and Y CPU It needs NNN storage It requires I/O SLAs for talking to B and C It needs to be available via a URL for trusted identities It needs to run on-premise and co-located near B
  22. 22. Is this Autonomic Computing?
  23. 23. How would we do this?
  24. 24. How would we even start?
  25. 25. Getting Started - App A What does App A need? Where will App A be run? How will App A find B and C? How do others find my App A? What happens on a failures?
  26. 26. Getting Started - App A What does App A need? - Packaging and Dependencies Where will App A be run? - Provision and Schedule How will App A find B and C? - Addressing/Discovery How do others find my App A? - External Mapping What happens on a failures? - Health Monitoring
  27. 27. Packaging and Dependency What does the job need to run? What runtimes, OS, libraries? What tools can I use for consistency, compliance, audit? SCCS and Chef / Puppet AMIs or VMDKs Docker Images
  28. 28. Packaging and Dependency Challenges Do these change when I change from Dev to Production? What runtimes, OS, libraries then? Who defines what these are? Are the existing tools and best practices still sufficient?
  29. 29. Provision and Schedule How fast can I provision? Can my workload run anywhere and be compliant? How do network perimeter security models effect placement? What is my unit or work? VM, App, Image? Can the system automatically handle compliance and policy? Can compliance and deployment be handled independently? What new tools exist? Mesos, Fleet?
  30. 30. Addressing and Discovery Is DNS sufficient? Do we need to change our applications? When things get moved, how does the system react? Is load balancing handled or is this a manual process? What happens when we scale up or down? How do others find us?
  31. 31. Monitoring and Management What happens when something fails? Is this a manual process? Who determines failure? Can we trust the system? What if they are sick, not dead? Latency vs Chaos Do we know if the change even helped? Pluggable Health
  32. 32. SO.. How do we get here?
  33. 33. Is it a Bolt-On Solution?
  34. 34. Or is it Bolt-In?
  35. 35. Bolt-On got us into this mess in the first place!
  36. 36. What we need is a Platform OS!
  37. 37. Programmable, pluggable and composeable..
  38. 38. From the inside out..
  39. 39. The OS for the datacenter
  40. 40. The OS for the datacenter Multiple Datacenters
  41. 41. Secure, Trusted, and Hybrid
  42. 42. Multi-Datacenter OS Treat all resources as a common pool Handle all networking access, addressing and discovery in realtime, and at scale Be aware of ontologies and their communication semantics Be security and policy aware Be purposely built to accept and promote rapid change Provide policy compliant resource isolation, connectivity and SLAs
  43. 43. Multi-Datacenter OS Virtualization SDN - Software-Defined Networking Management and Resource Pooling Intelligent and Compliant Job Scheduling Intelligent canarying, A/B rollouts
  44. 44. Multi-Datacenter OS Virtualization SDN - Software-Defined Networking Management and Resource Pooling Intelligent and Compliant Job Scheduling Intelligent canarying, A/B rollouts
  45. 45. Virtualization? What about speed and weight? Google chargeback diversion What about containers, e.g. Docker? Is there a container equivalent for .NET? Micro-task Virtualization?
  46. 46. Multi-Datacenter OS Virtualization SDN - Software-Defined Networking Management and Resource Pooling Intelligent and Compliant Job Scheduling Intelligent canarying, A/B rollouts
  47. 47. SDN? Solve network perimeter security? Does it involve application level changes? What about layer 7 semantics? How many INSERTS per second from all of App A? Can I disallow DROP and DELETE calls between 1a-3a? Can the network be made compliant and transparent? It just works, e.g. mobile
  48. 48. Multi-Datacenter OS Virtualization SDN - Software-Defined Networking Management and Resource Pooling Intelligent and Compliant Job Scheduling Intelligent canarying, A/B rollouts
  49. 49. intelligent and Compliant Job Scheduling Pick the best place to run for a given job and policy How does a system rebalance, utilize new resources? Centralized or Distributed Algorithms? How does policy effect decision making? E.g Geo
  50. 50. Multi-Datacenter OS Virtualization SDN - Software-Defined Networking Management and Resource Pooling Intelligent and Compliant Job Scheduling Intelligent canarying, A/B rollouts
  51. 51. Intelligent Canarying Want to roll out a new version of App A Do we know what App A - v2 success looks like? How do we do roll in and roll back (if needed)? How do we avoid our fingers on the keyboard? What is needed for this process to be automated?
  52. 52. Intelligent Canarying What data is needed to say if it is ok? resource utilizations - CPU, Mem, Storage communication patterns - cascading effects temporal awareness All data feeds into anomaly detection services Utilizes unsupervised deep machine learning
  53. 53. Summary
  54. 54. Summary Intelligent, holistic platform technologies - Pluggable and Composeable Transparent value add to jobs/workloads - No code changes! Packaging and Dependency Management - Policy aware Job Scheduling and Provisioning - Also policy aware Addressing, Discovery, Networking - Policy again, theme developing Monitoring and Management Lifecycle Management and Intelligent Canarying
  55. 55. Some Resources Docker - https:/ /www.docker.io/ Mesos - http:/ /mesos.apache.org/ CoreOS - https:/ /coreos.com/ Fleet, Etcd - https:/ /coreos.com/using-coreos/etcd/ Continuum - http:/ /apcera.com/continuum/
  56. 56. Thank You
  1. A particular slide catching your eye?

    Clipping is a handy way to collect important slides you want to go back to later.

×