A presentation at online webinar by erik riedel
OPEN UNCOMPLICATED SIMULTANEOUS SCALING FOR USER AND APPLICATION GROWTH Erik Riedel, PhD Senior VP, Engineering | ITRenew @RiedelAtWork
Hyperscale for All: Powering the Circular Data Center ITRenew delivers maximum financial & sustainability returns from open technology CIRCULAR CLOUD DECOMMISSIONING SESAME BY ITRENEW EDGE SOLUTIONS Strategic Advisory Services and Data Security Rack-Scale Solutions and Components Infrastructure planning Data center decommissioning services Rack-scale solutions for data centers Edge solutions & building blocks TCO & Sustainability Modeling Teraware data sanitization platform Open systems, HCI, AI/ML Server components Breakthrough TCO Laptop and PC memory Lifetime value maximization Value Recovery ($1B+ TCO to date) End-to-end logistics solutions
We have the proof that Open for All is not just a utopian vision, it is real and ready today: preconfigured, plug-in-ready rack-scale solutions, built on open architecture. No assembly. No guesswork. Soz No army of engineers required. Now anyone can power their data center like a hyperscaler while dropping their TCO and upping their sustainability game.
open hardware open software
Massive budgets and talent pools The big cloud providers have all the advantages Most advanced technology Greatest flexibility to scale Lowest costs
If everyone had the advantages, they could… Deliver more new services Respond faster to customer needs Grow the user base Increase margins & accelerate growth
Open Is Necessary, But Not Sufficient Per Se design, engineering and product expertise, rigorous commitment to component quality [hardware] design, engineering and product expertise, rigorous commitment to component quality from capacitors to consensus algorithms to cognitive load wide dynamic range of skills must be applied & appreciated ORCHESTRATING AN OPEN HARDWARE + OPEN SOFTWARE STACK [software] requires baseline partnership across hardware & software devops – collaboration from designers & developers to operations automation, analytics, and machine learning for all the things can’t be efficient at hyperscale without the machines making many, most decisions
The Benefits of Open Hardware MORE FLEXIBILITY HIGH DENSITY COMPUTING OPTIMIZED POWER OPTIMIZED COOLING STREAMLINED MAINTENANCE Multi-vendor, standards-based hardware for modular solutions to fit your needs More server, storage, and network capacity, in less space saves costs Rack-level power vs. individual server power. More efficient. Less cost. Fewer points of failure Rack-level cooling to operate more efficiently. Even more with free-air cooling, if the data centers support it Flexible, easy-access design enables faster troubleshooting, updates, and upgrades
The Benefits of Open Software MORE FLEXIBILITY HIGH DENSITY COMPUTING OPTIMIZED POWER OPTIMIZED VALIDATION STREAMLINED MAINTENANCE Multi-vendor, standards-based software for modular solutions to fit your needs More automation, with API-driven scalability, allows more software per silicon in2 Stack-level power vs. individual packaged software. More efficient. Less cost. Fewer points of failure Stack-level continuous integration, continuous deployment (CI/CD) to validate more efficiently. Fewer points of failure in the field Flexible, API-based, devops-considered design enables faster troubleshooting, updates, and upgrades
THE POWER OF HYPERSCALE FOR ALL Optimized for your workload, from deskside to data center. No assembly. No guesswork. Just plug them in. PROVEN HYPERSCALE TECH BUILT ON OPEN ARCHITECUTURE CONSISTENT PRODUCT DEPENDABLE SUPPLY BUSINESSCHANGING TCO
Optimized For Workloads. Ready To Deploy NO GUESSWORK. NO ASSEMBLY. JUST PLUG THEM IN. SESAME FOR OPEN SYSTEMS SESAME FOR AI/ML SESAME FOR CONVERGED Massive scale, cross-rack switching interconnects up to 20 racks; 750+ nodes in a single cluster, network domain High bandwidth, low latency platform to optimize learning Simplicity of design with standardized converged nodes Powerful application processing on large and small accelerated compute nodes Compute, storage and network capability scalable in lockstep as needed 100G and bandwidthoptimized connectivity for top throughput Fully open configuration and manageability tools, qualified and optimized for Linux Pre-validated with Kubernetes software and infrastructure stacks Linux-ready with scalable hardware management
Sesame for Open Systems • Massive scale; cross-rack switching interconnects up to 20 racks; 900+ nodes in a single cluster, network domain • Pre-validated with Kubernetes software and infrastructure stacks • Linux-ready with scalable node-level management CAPACITY • 6 to 36 nodes per rack – mix of node types • • • PERFORMANCE • BASE config: 150+ cpu cores & >1 TB memory single-socket x86 compute • SCALE config: 700+ cpu cores & >8 TB memory 2-socket x86 compute • 5 TB to 250 TB of high IOPS flash storage/ rack storage nodes w/ NVMe flash • 2.5 kW to 18.9 kW per rack • 3, 6, or 9 infrastructure nodes/rack • Full 25 GbE connectivity within the rack WORKLOAD FLEXIBILITY • Pre-designed/integrated racks fit most common Kubernetes/VM orchestration deployment architectures • Servers, storage, and networking hardware is pre-qualified and pre-tested • Designed to fit the space & power constraints of most modern data centers external TOR switches (2x) ingress ingress ingress internal TOR switches (2x) compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute power zone BB compute compute compute compute compute compute compute compute compute storage storage storage storage storage storage storage storage storage mgmt mgmt mgmt infra infra infra power zone AA
S e r ve rs external TOR switches (2x) ingress ingress ingress internal TOR switches (2x) compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute compute power zone BB compute compute compute compute compute compute compute compute compute compute compute storage storage storage compute storage storage storage compute storage storage storage infra mgmt mgmt mgmt power supply + switch infra infra infra up to 5 nodes Fast-Start power zone AA up to 45 nodes for Open Systems single or 2- socket nodes, 25 GbE connectivity flash-based storage nodes; millions of IOPS and terabytes of capacity
C a s e S t u d y – P ro j e c t G o a l s a n d G e n e s i s Many service provider & enterprise SaaS companies OCP adoption could be accelerated by offering pre-designed and pre-qualified solutions for key computing use cases. are looking for solutions to roll onto the floor, plug in, and quickly run workloads. We are working with infrastructure software stacks and software partners to pre-design and pre-qualify solutions with OCP equipment.
C a s e S t u d y – F ro m S e r ve rs to S e r ve r l e s s i n 1 0 2 3 M i n u te s Provisioning Server Linux DevOps Automation Our Choices Ubuntu 19.04 docker, Ansible Rancher Deployed Services DRP (PXE server) DHCP (static IPs) Digital Rebar docker registry Rancher
S e r ve rs SCALE compute compute compute compute infra power supply + switch DRP (PXE server) DHCP (static IPs) compute CRIMSON 2x 12c 256GB 10G compute CRIMSON 2x 12c 256GB 10G compute CRIMSON 2x 12c 256GB 10G compute CRIMSON 2x 12c 256GB 10G 96 cores 1 TB memory
1 st S te p – S u c c e s s ! O S e s D e p l o ye d
2 nd S te p – D e p l o y R a n c h e r
4 th S te p – S u c c e s s ! Workloads Configured
We D i d T h i s W i t h T h re e D i ffe re nt S ta c ks RANCHER KSPHERE TALOS 20 minutes to full cluster readiness 60 minutes to full cluster readiness 31.5 minutes to full cluster readiness 1,650 pods 1,800 pods 3,000 pods 500 pods* 500 pods* 500 pods*
c o n t a i n e r s b e g a n t o d i e , k u b e l e t c r a s h e d a t 5 5 0 p o d s ** note that all these tests were done after overriding the default 110 maximum pods per node, as set by Kubernetes
Serverless
INTRODUCING AN ALL-NEW OPEN RACK SCALE SERVER CHOICE PROVEN HYPERSCALE TECHNOLOGY PURPOSE BUILT READY TO DEPLOY DEPENDABLY AVAILABLE CONSISTENT PRODUCT BY BUSINESSCHANGING TCO MEET YOUR GROWING DATA AND INFRASTRUCTURE DEMANDS WITH THE SAME OPEN HARDWARE AS THE WORLD’S LEADING HYPERSCALERS. 35
C a l l To A c t i o n CHECK US OUT ON OUR WEBSITE: bit.ly/webinar-uncomplicated www.sesame.com Watch Our Videos: Sesame By ITRenew bit.ly/sesame-youtube QUESTIONS OR COMMENTS, REACH US: @RiedelAtWork
QUESTIONS
DATA CENTER IMPACT REPORT: THE FINANCIAL & SUSTAINABILITY CASE FOR CIRCULARITY Download at itrenew.com bit.ly/circular-report
BACKGROUND
Design and procure hardware that enables circularity HYPERSCALE DATA CENTERS 2 … 3 ITRENEW 4 5 INTERNAL RE-USE COST AVOIDANCE 1 LIFETIME VALUE MULTIPLIER RACK-SCALE SOLUTIONS EDGE SOLUTIONS EDGE SECONDARY MARKETS RE-CERTIFIED COMPONENTS CSP Industrial Retail Telco Gaming IoT Energy Enterprise RESPONSIBLE RECYCLING VALUE RECOVERY
6.7 million cars’ annual emissions
Simultaneous Scaling For User And Application Growth
The global pandemic has changed our considerations for a lot of things – including data center infrastructure. Overnight, market demands and business priorities have shifted, which means businesses can no longer afford a multi-year journey to cloud-native.
IT teams are now having to rapidly scale capacity up from thousands to millions of users. Due to the demand for more services, they must also scale datacenter infrastructure out to support hundreds of apps and workloads.
This talk will share insights on what it takes to scale users, infrastructure, and workloads overnight. Learn why industry leaders worldwide are taking an open software and open hardware approach to this multi-dimensional scaling dilemma. Drawing from case studies of implementations of ITRenew’s Sesame systems in data centers across the globe, this webinar will show how open hardware platforms deliver on the promise of elastic infrastructure; explain how open technology takes care of which apps run on which nodes, using frameworks like Kubernetes; and demonstrate how anyone can implement at scale, in record time. We also include a short case study on how to go from servers to serverless in 20 minutes.
It has never been more urgent to serve more customers, add more services and grow business revenues – accelerating cloud-native and elastic infrastructure initiatives using Kubernetes and Open Hardware can be done today.
The following resources were mentioned during the presentation or are useful additional information.
for free. You
can too.