Cloudera Enterprise 5 Reference Architecture

Transcription

Cloudera Enterprise 5Reference ArchitectureA PSSC Labs Reference Architecture GuideDecember 201620432 North Sea Circle, Lake Forest, CA 92630 * www.pssclabs.com * tel: 949-380-7288 * fax: 949-380-9788

IntroductionPSSC Labs continues to bring innovative compute server and cluster platforms to market. Focusing onspecific applications where performance and reliability are critical, highlights PSSC Labs strengths. TheApace Hadoop data framework requires substantial compute and storage capabilities coupled tightlytogether. Cloudera Enterprise 5 adds layers of unification, security and management.PSSC Labs is the first manufacturer to design a server platform specifically for Hadoop. Introducing theWorld’s highest density, lowest power consuming, Enterprise Ready Big Data server platform designedspecifically for Hadoop workloads. The CloudOOP 12000 offers the absolute highest possible computeand storage density combined with high performance Data IO throughput. PSSC Labs already deliversthe CloudOOP 12000 for small POC to large production clusters spanning hundreds of nodes.Key Features: Cloudera Enterprise 5 CertifiedReduce Data Center Footprint By 50%Reduce Power Consumption By 40%Nearly 50% Greater Data IO RatesPatent Pending Tool Free MaintenanceTechnical Specifications: Up to 12 x 3.5” SATAIII or 14 x SSDs in 1U Rack Spaceo 96 TB using twelve x 8 TB Hard Drives Supports UP to 2 x Intel Xeon E5 Series Processors Supports Up to 256 GB ECC Enterprise Memory 10 GigE, 40 GigE and Infiniband Network Support Red Hat , CentOS, Ubuntu, MS Windows Compatible“Certification of the CloudOOP 12000 on Cloudera Enterprise 5 ensurescustomers have the most innovative data management technologyavailable.”Tim StevensCloudera , Vice President20432 North Sea Circle, Lake Forest, CA 92630 * www.pssclabs.com * tel: 949-380-7288 * fax: 949-380-9788

CloudOOP 12000 Sample ConfigurationsEvery organization or use case requires different computing needs. The CloudOOP 12000 offers thegreatest flexibility possible. Below are three different proposed architectures for different workloads:high density storage, high computational requirements and a balanced configuration.CloudOOP 12000 High Density 72 TB Total Storage16 Xeon E5 Processor Cores128 GB ECC Memory2 x 10GigE Network Ports2 x GigE Network PortsRemote IPMI ManagementCentOS Linux OSPower Draw Estimate215 Watt Idle / 275 Watt Full LoadCloudOOP 12000 High Compute 24 TB Total Storage28 Xeon E5 Processor Cores256 GB ECC Memory2 x 10GigE Network Ports2 x GigE Network PortsRemote IPMI ManagementCentOS Linux OSPower Draw Estimate265 Watt Idle / 345 Watt Full LoadCloudOOP 12000 Balanced 48 TB Total Storage20 Xeon E5 Processor Cores256 GB ECC Memory2 x 10GigE Network Ports2 x GigE Network PortsRemote IPMI ManagementCentOS Linux OSPower Draw Estimate230 Watt Idle / 315 Watt Full LoadA Sample of Organizations Currently Using PSSC Labs for Cloudera Enterprise 520432 North Sea Circle, Lake Forest, CA 92630 * www.pssclabs.com * tel: 949-380-7288 * fax: 949-380-9788

CloudRAX : Cloudera Enterprise 5 Validated Turn Key Cluster SolutionsPSSC Labs offers a complete, turn-key cluster that is ready to run on delivery. With over 2000 HPCClusters delivered to date, PSSC Labs understands everything that is necessary for a successfuldeployment. All necessary hardware including servers, network equipment, power and infrastructureare included. PSSC Labs Cluster Engineers preconfigure network, storage, operating system and BIOSsettings to the end user’s specifications. Cloudera Enterprise 5 is installed at PSSC Labs factory. Thefinal step is the running of sample data sets to ensure proper functionality and performance. EachCloudRax Turn Key Cluster is custom configured for specific needs and budget.Below is an overview of each different server platform PSSC Labs offers for Cloudera Enterprise 5 turnkey deployments. Depending on the complexity of the environment, some software resources can beinstalled on different server platforms.CLOUDOOP 12000 DATA NODETech SpecsKey FeaturesSoftware Resourceo 1U High Density Form Factoro 2 x Intel Xeon E5 Processorso 12 x SATAIII or SAS Hard Drivesor 14 x SSDso 12 TB to 72 TB Storage Capacityo 32 GB to 256 GB ECC Memoryo 2 x GigE Network Adapterso Optional 10 GigE, 40 GigE,Infiniband Supporto Dedicated IPMI / iKVMo Enterprise Platformo Redundant Power Supplyo Improved Data IO Throughputo 40% Reduction in PowerConsumptiono 2 x the Density of StandardServero Flexible Configuration Optionso 3 Year Warranty Included(24 x 7 x 365 NBD Available)o TaskTracker Daemono DataNode Daemono Zookeepero Hbase Region Servero Impala Daemono Solr ServersSURESTORE U2000 MANAGEMENT NODE (NAME NODE & SECONDARY NAME NODE)Tech SpecsKey FeaturesSoftware Resourceso 2 x Intel Xeon E5 Processorso 1 TB to 24 TB SATA III, SAS, SSDHard Driveso 32 GB to 256 GB ECC Memoryo 2 x GigE Network Adapterso Optional 10GigE, 40GigE,Infiniband Supporto Dedicated IPMI / iKVMo Enterprise Platformo Redundant Power Supplyo Redundant Storageo Raid Levels 0,1,5,6,10,50o Flexible Configuration Optionso 3 Year Warranty Included(24 x 7 x 365 NBD Available)o Cloudera Managero Hbase Mastero JobTrackero Standby JobTrackero NameNodeo Standby NameNodeo JournalNodeso Zookeepero OozieTech SpecsKey Featureso 2 x Intel Xeon E5 Processorso 1 TB to 12 TB SATAIII, SAS, SSDHard Driveso 32 GB to 256 GB ECC Memoryo 2 x GigE Network Adapterso Optional 10GigE, 40GigE,Infiniband Supporto Dedicated IPMI / iKVMo Enterprise Platformo Redundant Power Supplyo Redundant Storageo Raid Levels 0,1,5,6,10,50o Flexible Configuration Optionso 3 Year Warranty Included(24 x 7 x 365 NBD Available)Resourceso Hive command line cliento Hadoop command line cliento Impala command line cliento Flume Agentso Hue ServerSURESTORE U1000 EDGE NODE20432 North Sea Circle, Lake Forest, CA 92630 * www.pssclabs.com * tel: 949-380-7288 * fax: 949-380-9788

CloudRAX Turn-Key Cluster Sample ConfigurationsCloudOOP Rax 150 150 TB Total Storage1 Name Node3 Data Nodes48 Xeon E5 Processor Cores196 GB ECC Memory10GigE Network BackplaneRemote IPMI ManagementCentOS Linux OSCloudera Installation ServiceCloudera Validation ServiceCloudOOP Rax 500 500 TB Total Storage2 Name Nodes1 Edge Node10 Data Nodes160 Xeon E5 Processor Cores1280 GB ECC Memory10GigE Network BackplaneRemote IPMI ManagementCentOS Linux OSCloudera Installation ServiceCloudera Validation Service Rack & Roll ServiceCloudOOP Rax 1500 1500 TB Total Storage2 Name Nodes1 Edge Node30 Data Nodes480 Xeon E5 Processor Cores3840 GB ECC Memory10GigE Network BackplaneRemote IPMI ManagementCentOS Linux OSCloudera Installation ServiceCloudera Validation Service Rack & Roll Service“We believe strongly in our ability to deliver thehighest performance, highest reliability serverplatforms to Cloudera Enterprise end users. Ourexperience delivering clusters ranging from severalhundred TBs to several dozen PBs ensures asuccessful Cloudera Enterprise deployment.”Larry LesserPSSC Labs, CTO20432 North Sea Circle, Lake Forest, CA 92630 * www.pssclabs.com * tel: 949-380-7288 * fax: 949-380-9788

Total Cost of Ownership ComparisonPSSC Labs goal is to offer solutions with the absolute lowest total cost of ownership. The below chart compares different server manufacturer’ssolution for a 1 Petabyte (raw) Hadoop environment. PSSC Labs CloudRax 1000 requires 50% less rack space and consumes 40% less powerRequired Data CenterFootprintPower ConsumptionEstimate*Required PowerCircuitsPre-installation andValidation ofCloudera Enterprise 5at FactoryOnsite PhysicalInstallationCluster ManagementTrainingDedicated RemoteMonitoringCapabilitiesHardware WarrantyPSSC Labs CloudRAX 1000for 1PB Total Storage SpaceSingle x 42U RackDell Configuration for1PB Total StorageTwo x 42U RackHP Configuration for 1PBTotal StorageTwo x 42U RackLenovo Configuration for1PB Total StorageTwo x 42U Rack4300 Watts Total @ Idle5500 Watts Total @ LoadSingle x 30 Amp / 208V /Single PhaseYes. Cloudera Enterprise 5preinstalled and tested.5800 Watts Total @ Idle8500 Watts Total @ LoadTwo x 30 Amp / 208V /Single PhaseNo. Additional servicesand fees required.6000 Watts Total @ Idle8800 Watts Total @ LoadTwo x 30 Amp / 208V / SinglePhaseNo. Additional services andfees required.5700 Watts Total @ Idle8700 Watts Total @ LoadTwo x 30 Amp / 208V /Single PhaseNo. Additional servicesand fees required.Yes. Cluster arrives preracked, cabled and labeled.Yes.No. Additional servicesand fees required.No. Additional servicesand fees required.Yes.No. Additional services andfees required.No. Additional services andfees required.Yes.No. Additional servicesand fees required.No. Additional servicesand fees required.Yes.3 Year NBD ServiceAvailable.3 Year NBD Service Available.3 Year NBD ServiceAvailable.Yes. IPMI 2.0 NetworkStandard3 Year NBD ServiceAvailable.*Dell, HP and Lenovo power estimates based on manufacturers website power draw estimates.20432 North Sea Circle, Lake Forest, CA 92630 * www.pssclabs.com * tel: 949-380-7288 * fax: 949-380-9788

Cloudera Enterprise 5 Reference Architecture A PSSC Labs Reference Architecture Guide December 2016 . 20432 North Sea Circle, Lake Forest, CA 92630 * www.pssclabs.com * tel: 949-380-7288 * fax: 949-380-9788 Introduction PSSC Labs continues to bring innovative co