techtonic summit nyc
TRANSCRIPT
Data Center Futures
Bob Wise Samsung SDS Research America
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved
2
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved
This presentation is intended to provide information concerning Samsung’s efforts around containers and container orchestration. We do our best to make sure that information presented is accurate and fully up-to-date. However, the presentation may be subject to technical inaccuracies, information that is not up-to-date or typographical errors. As a consequence, Samsung does not in any way guarantee the accuracy or completeness of information provided on this presentation. Samsung reserves the right to make improvements, corrections and/or changes to this presentation at any time.
The information in this presentation or accompanying oral statements may include forward-looking statements. These forward-looking statements include all matters that are not historical facts, statements regarding the Samsung Data System' intentions, beliefs or current expectations concerning, among other things, market prospects, growth, strategies, and the industry in which Samsung operates. By their nature, forward-looking statements involve risks and uncertainties, because they relate to events and depend on circumstances that may or may not occur in the future. Samsung cautions you that forward looking statements are not guarantees of future performance and that the actual developments of Samsung, the market, or industry in which Samsung operates may differ materially from those made or suggested by the forward-looking statements contained in this presentation or in the accompanying oral statements. In addition, even if the information contained herein or the oral statements are shown to be accurate, those developments may not be indicative developments in future periods.
Logos remain the property of their respective owners. So there.
3
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved
This presentation is intended to provide information concerning Samsung’s efforts around containers and container orchestration. We do our best to make sure that information presented is accurate and fully up-to-date. However, the presentation may be subject to technical inaccuracies, information that is not up-to-date or typographical errors. As a consequence, Samsung does not in any way guarantee the accuracy or completeness of information provided on this presentation. Samsung reserves the right to make improvements, corrections and/or changes to this presentation at any time.
The information in this presentation or accompanying oral statements may include forward-looking statements. These forward-looking statements include all matters that are not historical facts, statements regarding the Samsung Data System' intentions, beliefs or current expectations concerning, among other things, market prospects, growth, strategies, and the industry in which Samsung operates. By their nature, forward-looking statements involve risks and uncertainties, because they relate to events and depend on circumstances that may or may not occur in the future. Samsung cautions you that forward looking statements are not guarantees of future performance and that the actual developments of Samsung, the market, or industry in which Samsung operates may differ materially from those made or suggested by the forward-looking statements contained in this presentation or in the accompanying oral statements. In addition, even if the information contained herein or the oral statements are shown to be accurate, those developments may not be indicative developments in future periods. Logos remain the property of their respective owners. So there!
4
Here’s a future that sucks… • Deploy some Openstack • Deploy some VMware • Deploy some Mesos for Hadoop and Spark • Deploy some Cloud Foundry on Openstack • Deploy some Cloud Foundry on metal • Use AWS but get locked into their very nice services • Use some GCE • Deploy dedicated Unix/Windows/Linux • Throw swarm, kubernetes, GKE, and ECS into the mix for good measure. Run some on Mesos, some on Openstack.
• Run some Mesos on Openstack for good measure.
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 5
More suck… • Build experTse on all of those systems… $$$, if you can even do it.
• Pay lots for support since you have too many different distributed systems
• Try to figure out how to log, alarm, debug all of that.
• Try to figure out how secure all of that • Try to manage complexity by assigning each system it’s own dedicated cluster
• Enjoy < 10% system uTlizaTon • Have unsaTsfied customers due to reliability and performance issues.
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 6
• 70% Datacenter uTlizaTon
• MulTple releases a day / CD
• 10,000-‐40,000 servers per admin
• DestrucTve tesTng in producTon
• No-‐ops (Self service deployment)
• Go very fast and break nothing
• Commodity hardware
• Extremely low licensing spend (open source)
• Comprehensive AutomaTon
• Extensive AnalyTcs
High Bar -‐ Industrial Grade Compute
#GIFEE
Yeah, we want that!
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 8
Control and Risk Risk reducTon is about control Vendor lock-‐in is loss of control More control leads to lower cost
Open Source + Samsung Engineers = Control
Goal…
10
Standardize, Containerize, Deploy
…to Samsung Data Centers.
…to developer systems for agility and producTvity.
…to public virtual machine clouds. …to new and even more efficient public container clouds.
Open Source ContainerizaTon • Breaking vendor lock to expensive proprietary soluTons • More efficiency in data center uTlizaTon
• Easier path to commodity compute / white box infrastructure
• Greater portability to and from public cloud systems
• More agility in development and deployment
• Rapid development and deployment
• Deployment freedom/portability
• Scaleup/Scale down • Lowest possible cost deployments (capex and opex)
Strategic Elements • The dedicated cluster is dead. • Collorary: We need large shared clusters. • Only one kind of turtle at the bolom
– AWS Turtles or Google Turtles? – StandardizaTon – CommodiTzaTon – AutomaTon – Do one thing well, keep it as simple as possible
• Need to support legacy applicaTon porTng and cloud naTve
• Control: Open source, no lock-‐in
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 12
ExecuTon
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 13
Target Data Center Stack
Bare Metal
Container-‐opTmized host
Container Mgr
Container OpTmized Clustering and Resource AllocaTon
PaaS apps Java/Scala
Ruby Web Dev
Non-‐PaaS Apps Back end processing
Legacy Apps 3rd Party
Public Cloud Bare Metal
Container-‐opTmized host
Container Mgr
Public Cloud Bare Metal
Container-‐opTmized host
Container Mgr
Public Cloud
PaaS Cluster
PaaS Cluster
RelaTonal DB
Big Data DB
AnalyTcs Cluster
Big Data DB
AnalyTcs Cluster
Apps
Apps
Databases and Analy>cs Hadoop
Non-‐SQL DB (Cassandra, Mongo, etc) SQL DB (MySQL, Postgress, etc)
Container OpTmized Network
Target Data Center Stack
Bare Metal
Container-‐opTmized host
Container Mgr
Container OpTmized Clustering and Resource AllocaTon
PaaS apps Java/Scala
Ruby Web Dev
Non-‐PaaS Apps Back end processing
Legacy Apps 3rd Party
Public Cloud Bare Metal
Container-‐opTmized host
Container Mgr
Public Cloud Bare Metal
Container-‐opTmized host
Container Mgr
Public Cloud
PaaS Cluster
PaaS Cluster
RelaTonal DB
Big Data DB
AnalyTcs Cluster
Big Data DB
AnalyTcs Cluster
Apps
Apps
Databases and Analy>cs Hadoop
Non-‐SQL DB (Cassandra, Mongo, etc) SQL DB (MySQL, Postgress, etc)
Container OpTmized Network
Container OpTmized • Container OpTmized Hosts
– Very small footprint – only what is needed to run containers – Very fast booTng – Very fast update – security requires up-‐to-‐date infrastructure – Examples: – CoreOS – Atomic (Redhat) – Snappy (Ubuntu) – RancherOS
• Container OpTmized Network – Explosion of IP addresses: IP Address per container – IPV6
• Container OpTmized OrchestraTon – High rate of container start/stop – Higher density per server
“Opinionated” Cats Ahead…
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 17
Wrong Turtles at the Bolom
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 18
Openstack
Kubernetes
Upside Down
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 19
Kubernetes
Swarm
Much More Sensible
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 20
Swarm
Kubernetes
Turtles Stacked Too Deep
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 21
Openstack
Kubernetes
Mesos
Kubernetes is not a PaaS.
Type I Kubernetes User: Wants to manage a single app Type II Kubernetes User: Wants to manage data centers
Beler
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 22
Openstack
Kubernetes Mesos
Beler sTll…
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 23
Mesos
Openstack Kubernetes
Where we are headed…
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 24
The VMs, too. In containers.
Mesos Openstack Swarm
Kubernetes
PaaS
Strategic Conclusion
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 25
There Can Be Only One*
* For each enterprise. IF you really want GIFEE.
OpTon: Openstack • MulTple support opTons – no support lockin. • Wrong type of turtle • Burdened by featuriTs
– Collorary: Not keeping it simple
• Lack of strong technical leadership • Known issues at scale • Too much religious fervor
– Nobody expects the Pythonistas! • Have used this in produc3on, opinion is qualified.
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 26
OpTon: Mesos • A decent choice, proven track record at scale • Single vendor contributor ecosystem
– Feels like lock-‐in to Mesosphere if you want commercial support
– Confusing posiTon from Mesosphere vs DCOS • Complex and needing significant experTse to configure
operate – …but beler than openstack!
• Good ecosystem around big data (Spark, Hadoop) • Roots in scienTfic compuTng, not modern service
development…. – Container networking – Service discovery
• Have not used this in produc3on.
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 27
OpTon: Swarm • Not really in the same category as these other systems, but Docker seems headed that way
• Really a deployment tool, not a cluster management system
• Not suitable for large scale • Single vendor dominated • Support for APCC, RKT, etc seems challenged. • Very small contributor base • Wouldn’t consider using this in produc3on at scale, so opinion is unlikely to change for a while.
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 28
Choice: Kubernetes • Good design by a vibrant, healthy community
– MulT-‐contributor • Built from scratch for modern service develop/deploy
– Pods are the right abstracTon • Rapid pace of improvement
– Behind Mesos at point in Tme, but higher velocity – Our “Best Future” bet
• MulT-‐vendor support opTons (no support lock-‐in) – STll nascent.
• Great contributors with the right experience • Best high scale public cloud container opTon
– Google Container Engine • Supports mulTple container specs: Docker and APCC
29
Kubernetes Concerns • CNCF is the right direcTon, but slow gevng off the ground. • Featuri3s risk is ever present. • We are concerned and will con3nue to push as priori3es:
– HA – Scaling – Reliability – OperaTons tooling
• Accessibility for new users is criTcal to ecosystem health: – Quick start – DocumentaTon – UIs
• Looking forward to more companies in the direct support ecosystem.
30
A Few Words About PaaS
• We need to support mulTple PaaS frameworks.
• We do not want to dedicate clusters to each PaaS system. – Cloud Foundry / Diego/ BOSH
• PaaS workload needs to shared with other workloads
• A PaaS should not be in the container scheduling business!
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 31
A Few More Words About PaaS
• Cloud Foundry needs to rebase to Kubernetes (and Mesos) to live in this world. – Ditch BOSH – Ditch Diego – Focus on value up the stack rather than down.
• Kudos to Deis, Apache Stratos, and OpenShiw
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 32
SDSA Ref Architecture EvoluTon
CoreOS
Docker
Kubernetes 1.1+
AWS
Flannel
Apps
Terraform
Ansible
1000 Node Stack Baseline
OpTmized OS
Container
OrchestraTon
Compute
Networking
Provisioning
Samsung Kubernetes Engagement
• We are not forking • We are pushing back our work either into kubernetes or into our github repo – Samsung-‐AG.
• 2016 – more work on networking and storage, with 1000 node clusters as the baseline
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 34
Why are we involved in K8Scale?
• We want very large clusters with cross applicaTon resource sharing – Also looking forward to Ubernetes
• We believe we can make a posiTve contribuTon to make this happen faster and beler.
• We believe we need deep technical involvement to build/deploy/operate at scale.
• We learn by doing.
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 35
Contact Info
[email protected] bobwise on kubernetes slack hlps://github.com/Samsung-‐AG Sealle, WA
(Please reach out if interested in K8scale as well)
Copyright © 2015 Samsung SDS Co., Ltd. All rights reserved 36