Published on February 27, 2014
@randybias Pets vs. Cattle:! The Elastic Cloud Story ! DevOps Chicago Meetup! February 26, 2014 CCA - NoDerivs 3.0 Unported License - Usage OK, no modiﬁcations, full attribution*! * All unlicensed or borrowed works retain their original licenses
A Tale of Two Clouds 2
Enterprise Computing Approach GUI Driven! Ticket-Based! Hand-Crafted! Reserved ! Scale-up! Smart Hardware! Proprietary! Traditional Dev! … 3
Cloud Computing Approach API Driven! Self-Service! Automated! On-demand! Scale-out! Smart Apps! Open Source! Agile DevOps! … 4
Elastic Cloud Shifts Uptime Responsibility Enterprise Model Cloud Model 99.9%! Applications! (8h46m down) 99.999% Applications! (5m down) 99.999%! Infrastructure! ($$$$) 99% Infrastructure! ($$) 5
Elastic Cloud Origins Elastic Public Cloud Designed for Server Consolidation ! IT Admins manage Infrastructure! Ticket-based manual provisioning! Improves virtualization value + On-premise Deployment! Enterprise Virtualization! Private Cloud = Standardization, Automation,! Chargeback, Self-Service! + Virtual Infrastructure = ≠ Elastic ! Private Cloud Designed for Agility! Cloud Admins manage Services! Self-service automated provisioning! Delivers cloud value on-premise 6 Elastic & Virtualization 2.0 Clouds are very different.! ! Different workloads.! ! Different ! architectures.! ! Different ! skills.! ! Different economics.
What Companies Care About? Line of Business Enablement New App Initiatives (Mobile, SaaS, etc.) Data Center Modernization Business ! Agility! IaaS / PaaS ! ! Agile Methodologies Continuous Integration Agile Development! Continuous Testing & Delivery ACCELERATING! TIME TO VALUE! Public / Private / Hybrid Cloud Computing! ! ! Big Data / Analytics ! ! Public APIs Operational Discipline! DevOps Continuous Deployment 7 Data Center & App Automation
Elastic Cloud is a Mindset Change bowzer.company.com! (scale-up) web001.company.com! (scale-out) (Virtual) Servers *are* cattle Attribution: Bill Baker, Distinguished Engineer, Microsoft 8
Pets vs. Cattle Takes Off Cloudscaling Microsoft CERN Scale-out, not UP in Cloud IBM Red Hat Rackspace Scalr 9
(Some) Elastic Cloud Patterns ! ! ! What follows are *some* Elastic Cloud Patterns! There are many more, but these are mine! Input, ideas, & other thoughts welcome via twitter / email 10
Big Failure Domains ! Make Big Craters 11
Big Failure Domains ! Make Big Craters An An tiPa tte r tiP at te r n n 12
Smaller Failure Domains Would you rather have the whole cloud down ! or just a small bit of it for a short time? vs 13
Loose Coupling Synchronous, blocking calls mean cascading failures. Async, non-block calls mean failure in isolation. 14
Open Source Software Excessive software taxation is the past. You can ! always fork. Black boxes create lock-in. 15
Uptime in Software Self-management Hardware fails.! Software fails.! People fail. Only software can measure itself & respond to failure in near real-time. Applications designed for 99.999% uptime can run anywhere 16
Scale Out vs Scale up A B A B A B C ... Horizontal Scaling ➔ Make more boxes ➔ Vertical Scaling Make boxes bigger (usually an HA pair) 17 N
Circuit Breaker Pattern When failing service detected, stop calling that API and serve fallback responses Fallback mechanisms (e.g. cached data) ensure uninterrupted service while giving service time to recover 18
Buy from ODMs ODMs operate their businesses on 3-10% margins. AMZN, GOOG, and Facebook buy direct without a middleman. Only a few enterprise vendors are pivoting to compete. 19
Less Enterprise “Value” in x86 Servers Generic servers rule. Full stop. Nothing is better because nothing else is *generic*. “... a data center full of vanity free servers ... more efﬁcient ... less expensive to build and run ... “ - OCP 20
Fully Routed (L3) Networking The largest cloud operators all run layer-3 routed, networks with no VLANs. Cloud-ready apps don’t need or want VLANs. Enterprise apps can be supported on elastic clouds using Software-deﬁned Networking (SDN) 21
Software-deﬁned Networking (SDN) • • • • x86 server is the new Linecard! network switch is the new ASIC! VXLAN (or NVGRE) is the new Chassis! SDN Controller is the new SUP Engine 22 “Network Virtualization”
Flat Networking + SDNs Flat + SDN co-exist & thrive together Internet VM VM Availability Zone VM VM VPC Gateway VM Virtual L2 Network 1 2 VM Standard Security Group VM VM Physical Node 23 VPC Security Group VM Virtual Private Cloud Networking
RAIS instead of HA Pairs/Clusters Redundant arrays of inexpensive services (RAIS)! Load balanced with no state sharing! Active … active … active … active … ! On failure, connections are lost, but failures are rare! Rolling upgrades are easier, because each server is an island! Think: scale-out + fault isolation (sharding)! Ridiculously simple & scalable! Hardware failures are infrequent & impact subset of trafﬁc! (N-F)/N, where N = total, F = failed! 10 RAIS servers - 1 failure == 90% capacity! Most things retry anyway! Cascade failures are unlikely and failure domains are small 24
Service Array (RAIS) Example Public IP Blocks Backbone Routers OSPF Route Announcements RAIS (NAT, LB, VPN) Cloud Access Switches API Return Trafﬁc (default or source NAT) Cloud Control Plane AZ (Spine) Switches 25
Lots of Inexpensive 1RU Switches Simple spine-and-leaf ﬂat routed network Rack 1 Rack 2 Multiple Racks2 Rack Rack 1 Rack 3 1RU: 6K-30K VMs / AZ Multiple Racks2 Rack Rack 1 Multiple Racks2 Rack Rack 1 Modular: 40K-200K VMs / AZ 26
Direct-attached Storage (DAS) Cloud-ready apps manage their own data replication. DAS is the smallest failure domain possible with reasonable storage I/O. SAN == massive failure domain. SSDs will be the great equalizer. 27
Elastic Block Device Services EBS/EBD is a crutch Bigger failure domains (AWS outage anyone?), complex, sets high expectations Sometimes you need a crutch. When you do, overbuild the network, and make sure you have a smart scheduler. AWS EBS Outage! http://aws.amazon.com/message/65648/ 28
More Servers == More Storage I/O >1M writes/second, tripleredundancy w/ Cassandra on AWS Linear scale-out == linear costs for performance 29
Hypervisors are a Commodity Cloud end-users want OS of choice, not HVs. Level up! Managing iron is for mainframe operators.! … hypervisors are bare metal APIs Hypervisor of the future is open source, easily modiﬁable, & extensible. 30
The Hypervisor of the Future May Be NO Hypervisor Bare Metal Cloud LXC ironic 31
Quiz Time 32
Quiz Time Pets Cattle LACP? 33
Quiz Time Pets LACP Cattle ➔ 34
Quiz Time Pets Cattle LACP Managing a Server at a Time? 35
Quiz Time Pets Cattle LACP Managing a Server at a Time ➔ 36
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling? 37
Quiz Time Pets Cattle LACP Managing Server at a Time ➔ 38 Auto-scaling
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure? 39
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling ➔ 40 Design-for-Failure
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals? 41
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals ➔ 42
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals HA pairs for redundancy? 43
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals HA pairs for redundancy ➔ 44
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals HA pairs for redundancy Shared Nothing Architecture? 45
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals HA pairs for redundancy ➔ 46 Shared Nothing Architecture
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals HA pairs for redundancy Shared Nothing Architecture Persistent Block Storage? 47
Quiz Time Pets Cattle LACP Managing Server at a Time Auto-scaling Design-for-Failure 100% Uptime Goals HA pairs for redundancy Shared Nothing Architecture Persistent Block Storage ➔ 48
Q&A Randy Bias! Founder & CEO, Cloudscaling! Director, OpenStack Foundation! @randybias 49
Cloudscaling is the leading provider of on-premise elastic cloud infrastructure solutions to support dynamic applications.
Nhớ lại những ngày cũ, trước khi các dịch vụ như của AWS hay GCE trở nên phổ biến như bây giờ, phương cách quản trị ...
The talk was called Pets vs. Cattle: The Elastic Cloud. ... Stories of Linux; ... Computing Systems Management Are Your Servers Pets or Cattle?
... James Weatherell updated his online resume on LinkedIn with his recent experience working with new cloud ... Cloud Pets Vs. Cattle ...
... most cloud providers are built around the concepts of shared ... The Rest of the Story. ... I hope you enjoyed reading about the pets vs. cattle ...
Virtual CCAP: NFV Pets vs. Cattle ... Pets vs. Cattle: The Elastic Cloud Story. 27,176 Views. markhubernyc. How to Keep Cattle in and the Predators Out. 0 ...
View 174 Elastic Cloud posts, presentations, experts, and more. Get the professional knowledge you need on LinkedIn. LinkedIn Home What is LinkedIn?
Here's a key distinction that I've been trying to make when talking about cloud vs ... pets vs cattle" is ... elastic qualities of cloud ...