Successfully reported this slideshow.
We use your LinkedIn profile and activity data to personalize ads and to show you more relevant ads. You can change your ad preferences anytime.

MapR and Cisco Make IT Better

883 views

Published on

You’re not the only one still loading your data into data warehouses and building marts or cubes out of it. But today’s data requires a much more accessible environment that delivers real-time results. Prepare for this transformation because your data platform and storage choices are about to undergo a re-platforming that happens once in 30 years.

With the MapR Converged Data Platform (CDP) and Cisco Unified Compute System (UCS), you can optimize today’s infrastructure and grow to take advantage of what’s next. Uncover the range of possibilities from re-platforming by intimately understanding your options for density, performance, functionality and more.

Published in: Data & Analytics
  • Be the first to comment

MapR and Cisco Make IT Better

  1. 1. Robert Novak, Big Data Partners Consulting SE, Cisco Bill Peterson, VP Partner Strategy, MapR January 2017 It's No Use Going Back to Yesterday's Storage Platform for Tomorrow's Applications
  2. 2. Who is Bill and why is he here? Today: Vice President of Partner Strategy MapR Current: Worldwide Partner Marketing North America Field Marketing Past: Analyst (IDC) PR Flack (PetersonPR, Page One PR) Product Marketing (NetApp and more) IT Manager (Harvard University) Blogger at mapr.com Tweeter at @thebillp
  3. 3. Who is Rob and why is he here? Today: Consulting Systems Engineer for Cisco’s Americas Partner Organization Focused on big data and analytics UNIX Sysadmin for ~20 years (retired) Full stack: servers, storage, network, coffee 149 to 149k person companies Sun, Nortel, 3PAR, eBay, Trulia, Disney, etc “Big Data” herder since 2003 Hadoop admin (certifiable) since 2009 Cisco UCS C-Series admin since 2011 (early adopter!) Charter Cisco Champion, VMware vExpert since 2013 Blogger at rsts11.com and Cisco Blogs Tweeter at @gallifreyan and @rsts11 and @rsts11travel
  4. 4. ‘I could tell you my adventures—beginning from this morning,’ said Alice a little timidly: ‘but it’s no use going back to yesterday, because I was a different person then.’ 1. Traditional app scaling 2. Moving into the Big Data Era™ 3. Models of Scale and Density 4. MapR • Next Gen Technologies • Converged Data Platform (CDP) • How is it being used 5. Cisco Unified Computing System (UCS) • Utility Computing • Policy-driven scalability • S-Series storage servers 6. Scaling and sustaining with CDP on UCS 7. Where do we go from here? What are we talking about today?
  5. 5. Traditional app scaling (“Are you from the PAST?”)
  6. 6. Monolithic applications Monolithic servers Large somewhat-modular storage arrays Growing pains involve forklifts At the turn of the century… By Peter Hamer - Ken Thompson (sitting) and Dennis Ritchie at PDP-11 Uploaded by Magnus Manske, CC BY-SA 2.0, https://commons.wikimedia.org/w/index.php?curid=24512134
  7. 7. Moving into the Big Data Era™ (“We can’t stop for gas, we’re late already!”)
  8. 8. Grid, Beowulf, Hadoop Divide and conquer, scale storage and compute together Hadoop started putting data where it needed to be Applications change, storage changes Beowulf page public domain via Wikimedia Commons. Aiyara cluster By Kaewkasi at English Wikipedia, CC BY-SA 3.0, https://commons.wikimedia.org/w/index.php?curid=52546611 . NASA cluster By NASA Ames Research Center/Tom Trower - http://www.nas.nasa.gov/News/Images/columbia_3.html., Public Domain, https://commons.wikimedia.org/w/index.php?curid=43593
  9. 9. Modern platform scaling Match the compute and the storage to the application, and not the other way around
  10. 10. Hadoop is great, if you can re-architect your apps, or start from scratch Access to the cluster transparent only to native apps Ain’t nobody got time for that Modern big data still requires virtual forklifts
  11. 11. What if… Your cluster provided industry-standard access methods like NFS Existing applications could plug right in with no recoding New applications could live side by side
  12. 12. With Cisco Unified Computing System and MapR Converged Data Platform… You can. Amazing but true.
  13. 13. © 2016 MapR Technologies 13© 2016 MapR Technologies 13© 2016 MapR Technologies Driving Business Transformation with Data How the MapR Converged Data Platform Drives Innovation, Growth and Efficiency
  14. 14. © 2016 MapR Technologies 14© 2016 MapR Technologies 14 We will disrupt ourselves through Data JPMorgan Chase
  15. 15. © 2016 MapR Technologies 15© 2016 MapR Technologies 15 Industry Leaders Are Investing in Disruptive Technology Now Legacy technology investmentNext-Gen technology investment Source: IDC, Gartner; Analysis & Estimates: MapR Next-gen consists of cloud, big data, software and hardware related expenses (80,000) (40,000) - 40,000 80,000 120,000 2013 2014 2015 2016 2017 2018 2019 2020 $ (millions) Investment in Next-Gen vs. Legacy Technologies for Data Total $ growth of IT market Innovating and Reducing Costs at the Same Time 90% of data is on next-gen technology in just four years
  16. 16. © 2016 MapR Technologies 16© 2016 MapR Technologies 16 MapR is Transforming Business with Data WHAT WE DO Bring together analytics and operations into next-generation Converged Applications for the business WHY IT MATTERS Empowers companies to grow revenue through innovation and cutting costs HOW WE DO IT Patented technology architecture with the world’s only complete Converged Data Platform Leading companies around the world are transforming their business with the industry’s only Converged Data Platform
  17. 17. © 2016 MapR Technologies 17© 2016 MapR Technologies 17 Enabling Transformation Through Converged Applications OPERATIONAL APPLICATIONS Immediate ANALYTICAL APPLICATIONS Historical Complete access to real-time and historical data in one platform Converged Applications
  18. 18. © 2016 MapR Technologies 18© 2016 MapR Technologies 18 Powered by the World’s Only Converged Data Platform Breakthrough Reliability Operate globally at enterprise grade for mission critical apps Breakthrough Value Radically cut costs of big data IT infrastructureBreakthrough Innovation Enable continuous innovation with proprietary technology and open source access A platform engineered to support next-generation applications
  19. 19. © 2016 MapR Technologies 19© 2016 MapR Technologies 19 Optimized for Speed • Optimized compute engines • High speed ingest & I/O • Fast recovery and snapshots • High scale elasticity MapR Platform Lowers Friction For Next Gen App Developers Innovative architecture delivers uncompromising scale, speed and availability Optimized for Availability • Global reliability • Inter-cloud, Containerized • Data protection & recovery • Strong security model Optimized for Scale • High scale data store • Document Database • Persistent Streaming • Schema Free SQL engine • Community innovation
  20. 20. © 2016 MapR Technologies 20© 2016 MapR Technologies 20 MapR Converged Data Platform Product Line
  21. 21. © 2016 MapR Technologies 21© 2016 MapR Technologies 21 Flexible processing where change is the norm Distributed processing across clusters, data centers, public & private cloud environments Supports global apps that can scale arbitrarily A Single Platform: On-Premises, In the Cloud, or Hybrid
  22. 22. © 2016 MapR Technologies 22© 2016 MapR Technologies 22 We Make It Easy to Get Started 1 Understand capabilities of big data platform Experimentation 2 Develop first use cases and put into production Implementation 3 Expand to multiple use cases across key lines of business Expansion 4 Integrate and expand data driven apps and analysis to all lines of business and more business functions Optimization Take the MapR Big Data Maturity Model
  23. 23. © 2016 MapR Technologies 23© 2016 MapR Technologies 23© 2016 MapR Technologies
  24. 24. © 2016 MapR Technologies 24© 2016 MapR Technologies 24 A Crisis of Complexity Expensive to stitch together Fragile not agile “Connected” and “Federated” not converged Limited in scale, no global Many security models, points of failure Hadoop & Spark cluster Cassandra for event or content logging Classic data warehouse Message middleware Application serverDocument JSON DB Search server vs. the Complete Data Platform Engineered as single platform Powers legacy and next-gen apps Enables continuous innovation Supports all big data technologies Multiple deployment environments BUSINESS MODEL: SUPPORT FREE SOFTWARE BUSINESS MODEL: ENTERPRISE SOFTWARE LICENSES
  25. 25. © 2016 MapR Technologies 25© 2016 MapR Technologies 25 Converge, Transform and Grow with MapR Innovate for Growth Realize Cost Savings
  26. 26. © 2016 MapR Technologies 26© 2016 MapR Technologies 26 Our Customers Are Leading the Way Financial Services Telco & Media Ad tech Government RetailOver 80 use cases including payment efficiency and accuracy of claims processing. $2M/month reduction in payment errors and fraud. Provides 95% of Fortune 500 CPG and retailers with data and analytics. Achieved $2.5M/year annual savings from mainframe & DW offload. Ported credit scoring use case to MapR resulting in 20X cost savings over DB2. Biometric identification system for more than 1.25 billion people in India. $1.3B yearly savings thru fraud reduction. Developed a new self-service analytics platform to give their customers better market insights to help them operationalize their decisions. Protects $1 trillion in charge volume from fraud every year. Amex offers program has saved card members over $180M.INNOVATION COST REDUCTION
  27. 27. © 2016 MapR Technologies 27© 2016 MapR Technologies
  28. 28. © 2016 MapR Technologies 28 MapR Performance Advantage: Summary YARN Tasks Benchmark: Terasort Measurement: Total time (s) Distributed Filesystem Benchmark: DFSIO Measurement: Read/Write MB/s NoSQL Database Ops Benchmark: YCSB 50/50 Measurement: Ops/s/client Test configuration: 10 nodes, 16x2 cores (2.6 GHz), 11x1TB disk (7200 RPM, 1 SP), 128G RAM, jumbo frames, 1x10GE Double the speed for YARN Up to 2X faster distributed I/O 0 10000 Other Vendor MapR 5.0 Operations per second 3-4X more NoSQL throughput + True Read/Write Distributed Filesystem 0 500 1000 1500 2000 Other Vendor MapR 5.0 YARN Jobs Complete in Half the Time 0 500 1000 1500 Other Vendor MapR 5.0 Write MB/sec write read
  29. 29. © 2016 MapR Technologies 29 MapR Leads TPC-HS Benchmark for 1TB, 10TB Sizes • Cisco + MapR were the first to complete the Big Data benchmark from the Transaction Processing Council in 2015 • MapR 5.x recorded the highest performance for the 1TB and 10TB tests • “… stresses both hardware and software including Hadoop run-time, Hadoop Filesystem API compatible systems and MapReduce layers”
  30. 30. © 2016 MapR Technologies 30 File Creation Benchmark MapR Scales to Billions of Files, Competition Fails at 1.25M 0 2000 4000 6000 8000 10000 12000 14000 16000 18000 0 2000 4000 6000 Filecreates/s Files (M) Other distribution fails Test Configuration: 100 byte file creation, 10 nodes, each w/2 x 4 cores, 24G RAM, 12x1TB (7200 RPM), 1x10GE 0 100 200 300 400 0 0.5 1 1.5 Filecreates/s Files (M) Other distributionMapR Other Advantage Rate (creates/s) 14-16K 335-360 40x Scale (files) 6B 1.3M 4615x platform maintains a consistent high rate of file creation
  31. 31. © 2016 MapR Technologies 31 World Record Performance PREVIOUS RECORD: 1.6 TB with 2200 nodes 1.65 TBIN 1 MINUTE 298 NODES NEW MINUTESORT WORLD RECORD With a Fraction of the Hardware Hardware Specs: Cisco UCS C240M3S 2x Intel ® Xeon ® E5-2665 @ 2.4GHz (16 cores) CPU; 128 GB RAM; SATA 24x1 TB Storage; Cisco VIC 10 GbE NIC 2200 nodes - competitor 1/7th of hardware required 298 nodes -
  32. 32. Cisco UCS for Scalable Platforms
  33. 33. Hardware Still Matters A quick glance at infrastructure for Big Data a.k.a. Why is Cisco here?
  34. 34. Why does hardware still matter? 34 • Cisco customers’ big data pools tend to grow 2-3x/year (or more… lots more) • Customer IT staff doesn’t grow as fast • The Cisco Unified Computing System (UCS) provides scalable, repeatable, predictable, and manageable deployments across dozens to thousands of servers for any application deployment • Pallet to production in hours, not days or weeks • Deep engineering integration between Cisco and MapR with tested and proven configurations More on this later…
  35. 35. Ten years from now, tech industry historians will remember at least two things about 2009: the economic mess and the Cisco UCS announcement. If nothing else, Cisco just made the industry much more exciting than it was last Friday. Jon Oltsik, CNet, March 16, 2009 Why Cisco UCS for any hardware deployment? Ten second edition • Scalability • Manageability • Performance
  36. 36. Ten years from now, tech industry historians will remember at least two things about 2009: the economic mess and the Cisco UCS announcement. If nothing else, Cisco just made the industry much more exciting than it was last Friday. Jon Oltsik, CNet, March 16, 2009 Why Cisco UCS for any hardware deployment?  Single point of management and access control for thousands of servers  Centralized host/network/storage/lights-out firmware management built in  High performance networking at the core  Flexible network configuration with vNICs for security and scalability  Open XML API for automation and third party integration  Fully functional remote console (including virtual media) at no extra cost
  37. 37. Why do these especially matter for highly scalable platforms?  Single point of management and access control for thousands of servers  Centralized host/network/storage/lights-out firmware management built in  High performance networking at the core  Flexible network configuration with vNICs for security and scalability  Open XML API for automation and third party integration  Fully functional remote console (including virtual media) at no extra cost  Big Data grows faster than most platforms. Ever added 100 Oracle servers?  Big Data environments tend to grow 2x-3x (or more) within two years. IT staff do not.  More data moving around means heavier pressure on the network to perform  New software models may require different networking and storage  Larger companies have existing management infrastructures to work with  Some vendors nickel and dime for management features and licenses.
  38. 38. Cisco UCS Reference Architectures • Integrated Infrastructures for Big Data (a.k.a. CPAv4) updated May 2016 • Proven, predictable configs to start from and grow with • Scale to petabytes of data with high performance and low TCO • Seven infrastructure designs for different compute/capacity/ performance requirements • CPAv4 Blog link
  39. 39. Cisco UCS Integrated Infrastructure for Big Data 4th Generation of Reference Architectures and Bundles UCS-SL-CPA4-S UCS-SL-CPA4-H UCS-SL-CPA4-P1 UCS-SL-CPA4-P2 UCS-SL-CPA4-P3 UCS-SL-CPA4-C1 UCS-SL-CPA4-C2 Extreme Capacity Coming Soon Network: 2x 6248 Servers: 8 X UCS-BD- C220M4-S1 Server Type: C220 M4 SFF CPU: 2x 2620v4 Memory: 128GB DDR4 Drives: 8 x 1.2TB 10K SAS HDD VIC: VIC 1227 RAID: 12Gps SAS, 2GB UCSD: no Cores: 128 Memory: 1024 Raw Storage: 76.8 I/O Bandwidth: 7.5 Gbytes/sec Network: 2x 6332 Servers: 8 X UCS-BD- C220M4-H1 Server Type: C220 M4 SFF CPU: 2x 2680v4 Memory: 256GB DDR4 Drives: 8 x 960GB SSD VIC: VIC 1387 RAID: 12Gps SAS, 2GB UCSD: no Cores: 224 Memory: 2048 Raw Storage: 60 I/O Bandwidth: 20 Gbytes/sec Network: 2x 6296 Servers: 16 X UCS-BD- C240M4-P1 Server Type: C240 M4 SFF CPU: 2x 2680v4 Memory: 256GB DDR4 OS: 2 x 240GB SSD Drives: 24 x 1.2TB 10K SAS HDD VIC: VIC 1227 RAID: 12Gps SAS, 2GB UCSD: yes Cores: 448 Memory: 4096 Raw Storage: 460.8 I/O Bandwidth: 45 Gbytes/sec Network: 2x 6296 Servers: 16 X UCS-BD- C240M4-P2 Server Type: C240 M4 SFF CPU: 2x 2680v4 Memory: 256GB DDR4 OS: 2 x 240GB SSD Drives: 24 x 1.8TB 10K SAS HDD VIC: VIC 1227 RAID: 12Gps SAS, 2GB UCSD: yes Cores: 448 Memory: 4096 Raw Storage: 691.2 I/O Bandwidth: 48.75 Gbytes/sec Network: 2x 6332 Servers: 16 X UCS-BD- C240M4-P3 Server Type: C240 M4 SFF CPU: 2x 2680v4 Memory: 256GB DDR4 OS: 2 x 240GB SSD Drives: 24 x 1.8TB 10K SAS HDD VIC: VIC 1387 RAID: 12Gps SAS, 2GB UCSD: yes Cores: 448 Memory: 4096 Raw Storage: 691.2 I/O Bandwidth: 48.75 Gbytes/sec Network: 2x 6296 Servers: 16 X UCS-BD- C240M4-C1 Server Type: C240 M4 LFF CPU: 2x 2620v4 Memory: 128GB DDR4 OS: 2 x 240GB SSD Drives: 12 x 6TB 7.2K SAS HDD VIC: VIC 1227 RAID: 12Gps SAS, 2GB UCSD: yes Cores: 256 Memory: 2048 Raw Storage: 1152 I/O Bandwidth: 26.25 Gbytes/sec Network: 2x 6296 Servers: 16 X UCS-BD- C240M4-C2 Server Type: C240 M4 LFF CPU: 2x 2620v4 Memory: 256GB DDR4 OS: 2 x 240GB SSD Drives: 12 x 8TB 7.2K SAS HDD VIC: VIC 1227 RAID: 12Gps SAS, 2GB UCSD: yes Cores: 256 Memory: 4096 Raw Storage: 1536 I/O Bandwidth: 26.25 Gbytes/sec Network: 2x 6332 Servers: 9 X UCS-BD-C3220- HC1 Server Type: C3260 (2 x servers) CPU: 2 x 2680v4 Memory: 256GB DDR4 OS: 2 x 240GB SSD Drives: 424 6TB 7.2K SAS HDD VIC: VIC 1387 RAID: 12Gps SAS, 2GB UCSD: yes Cores: 504 Memory: 4608 Raw Storage: 2544 I/O Bandwidth: 57.97 Gbytes/sec
  40. 40. High performance fabric for distributed storage Automation for rapid scalability Capacity / performance versatility for all apps API controls for developers reduction in provisioning time83% reduction in ongoing management costs62% reduction in power and cooling costs49% Reduction in cabling78% UCS: Ideal for Active Data UCS Customer Results
  41. 41. Cisco UCS S3260: Modular Platform Massive Capacity • 600TB data storage capacity in 4U • Up to 90TB SSD Flash Lower Capex34% Multi-Node Power • Single or Dual Server Options • New cache acceleration capabilities with NVMe and Fusion IoMemory I/O flexibility • 40Gb Cisco Virtual Interface Card (VIC) Technology • 256 virtual adapters per node plus 16Gb native Fabre Channel options Total Automation • Scale to Petabytes in minutes with UCS Manager • Cisco SystemLink Technology with flexible storage profiles Lower Ongoing Management80% Less Cabling70% Less Space60% Lower Power59% Significant Benefits Compared to Conventional 2RU Servers
  42. 42. Versatility for All Data-Intensive Applications Object Storage • Email Storage • Medical imaging • Video Storage 600 Terabytes Raw Storage 1G or 4G RAID Cache Capacity Play Data Protection • Consolidated backup target • Multi-site replication • Disaster Recovery 160 GB Aggregate VIC I/O 8 and 16GB Fiber Channel PCIe I/O Flexibility Big Data Analytics • Recommendation engines • Fraud detection • Network security 1.6TB NVMe 2 x Fusion ioMemory3 PX 90 TB SSD Flash Cache Optimized Content Distribution • Video Surveillance • Facial Recognition • Content Delivery Dual Server Nodes 72 CPU Cores Compute Optimized
  43. 43. • Updated by Cisco and MapR engineers in October 2016 • 250+ page guide to design and deployment, pallet to production • Based on UCS C-Series (C220, C240, S3260) servers and MapR Converged Data Platform • Download for free at cisco.com/go/bigdata_design • MapR + UCS S3260 CVD (PDF) • MapR Streams CVD (HTML/PDF) Cisco Validated Designs (CVD) for MapR
  44. 44. • CVD Updated November 1, 2016, to optimize for S-Series • CVD for SAP HANA on MapR Converged Data Platform released June 2016 • Flexible server and storage offering means you can plug it into whatever you need to do, CVD or reference architecture or not What’s coming for Cisco and MapR?
  45. 45. Call To Action MapR & Cisco Make IT Better Webinar & Roadshow Series – It's No Use Going Back to Yesterday's Storage Platform for Tomorrow's Applications – Cisco & MapR for SAP HANA – Cisco & MapR for Software-Defined AND Web-Scale Storage Read the CVD’s (or at least the parts that interest you) Follow: @gallifreyan – @mapr – @Cisco – @thebillp Attend Cisco Live, Big Data Everywhere, Strata Hadoop World
  46. 46. © 2016 MapR Technologies 46© 2016 MapR Technologies
  47. 47. Cisco UCS S3260 System Overview Drives 4 Rows of Hot-Swappable HDD 4TB/6TB/8TB/10TB with up to 2 Rows of 400GB/800GB/1.6TB/3.2TB SSD Total Top Load: 56 drives FAN 8 Hot-Pluggable Fans Server Node Up to (2) Based on Intel V4 CPUs, LSI 12G RAID, Up to 512GB DDR4 RAM (1024GB Post-FCS), and NVMe Optional Second Node Server Node or Drive Expansion or PCIe Expansion Up to (4) 120GB/480GB/1.6TB SSDs HW RAID, Hot-Plug, OS/Boot System I/O Controller (SIOC) Up to (2) Cisco VIC 1300 on Chip Power Supply 4 Hot-Pluggable PSUs *Shown with Single Server Node and IO Expander
  48. 48. What’s new (from C3260) M4 Server Node Dual-Socket Intel XEON E5-2600 V4 Processors Up to 512GB of DDR4 memory Single 800G or 1.6TB 2.5" NVMe SSD 12G SAS RAID with 4GB Cache I/O Expander Module Dual 8x PCIe half-height half-width slots Support for 3rd party add-in Modules Works only with M4 server nodes Connectivity 16G FC8/16G FCDual- 10GbE Quad- 1GbE Flash Memory 64000GB32000GB10000GB
  49. 49. UCS S3260 Management • UCSM 3.1.2 Integration  Fully Managed by 2nd and 3rd Generation Fabric Interconnects  Connects via FI Server Ports to 3260 SIOC Ports  Each 3260 Physical Box is a Chassis  Chassis-Wide and Per Server Node Management  Inventory, Compute and Storage Configuration, FW Mgmt, Pools, Policies, Profiles, Templates, vNICs/vHBAs, and Much More  Storage Profiles – Disk Group (RAID) and LUN Configuration

×