WIXLIB

Disaster Recovery SolutionAchieved by EXPRESSCLUSTERJuly, 2019NEC Corporation,Cloud Platform Division,(EXPRESSCLUSTER)1 NEC Corporation 2019

Index1. Clustering system and disaster recovery2. Disaster recovery with remote clustering3. Configuration of remote clustering4. Features of remote clustering5. Case studies

1. Clustering system and disaster recovery

1. Clustering system and disaster recovery Even highly available clustering system has risk of system disruptionin case of widespread disaster Clustering system ensures redundancy for wide system components such ashardware, software, and network. There is still risk of system disruption in case of widespread disaster such asfire, power outage, earthquake, floods etc.IT infrastructure disruption due to firedisaster, electric outage, earthquake, etc.Network 1NIC 1NIC 2Network 2NIC 1ActiveHBA 1HBA 2Controller 1Disk 15NIC 2Stand-byHBA 1HBA 2Controller 2Disk 2 NEC Corporation 2019

2. Disaster recovery with remote clustering

2. Disaster recovery with remote clusteringEarthquake, fire disaster, electric outage ensures business continuity from any disasterFireApplication FailoverEarthquakeWAN/VPNVirusMain siteElectric outageData Mirroring7 NEC Corporation 2019Backup site

3. Configuration of remote clustering

3. Configuration of remote clustering① 1 1 RemoteClustering Configuration Clustering configured across remote sites Only one communication network is enough between serversAsynchronous mirroring eliminates performance degradation due to network delayClustering can be configured among different network segmentsCompresses mirroring data, and saves bandwidth used by transferred data Standby server takes over the workload in case of failure / disaster Automatic application level failover Synchronous / Asynchronous data mirroringWAN/VPNIP addressHost nameIP addressHost nameApplicationApplicationData Mirroring9 NEC Corporation 2019

② 2 2 HybridClustering Configuration3. Configuration of remote clustering Ensures redundancy even within backup site① Local failover within primary site in case of system failure② Failover to backup site in case of disaster③ Ensures further availability by maintaining clustering configuration at backup siteWAN/VPNRedundancy inbackup siteSelects automaticallyor manuallyIP addressHost name①ApplicationIP addressHost nameIP addressHost name③ApplicationApplication②Mirroring of shared disk10 NEC Corporation 2019IP addressHost nameApplication

3. Configuration of remote clustering③ 2 1 HybridClustering Configuration Scale up existing cluster to remote cluster configuration with low cost① Prepare for system failure with shared disk type of clustering at primary site② Automatic application level failover in case of disaster③ Leverage local disk at back up site and achieve low cost disaster recoveryWAN/VPNAutomatic / Manualfailover①IP addressHost nameApplication②IP addressHost nameApplicationIP addressHost nameApplicationShared disknot requiredMirroring of shared disk to local disk11 NEC Corporation 2019③

3. Configuration of remote clustering④ 2 1 Hybrid ClusteringConfiguration (with virtualization) Cost reduction by virtualization of backup site Supporting leading hypervisors such as VMware, Hyper-V, KVM, XenServer, etc Enables flexible configuration such as physical/virtual mixed configuration, virtualmachine only configuration, etcWAN/VPNIP addressHost nameApplicationIP addressHost nameApplicationIP addressHost nameApplicationMirroring to virtual disk12 NEC Corporation 2019Virtual diskHypervisor

4. Features of remote clustering

4. Features of remote clustering① Synchronous mirroringAlways ensures to have the latest data on both serversWriteRead①EXPRESSCLUSTER receives write data from theapplication.①Read is done from active disk only thus will notbe affected by network performance at all②Data is written to primary server, andtransferred to standby server at the same time③Writing is completed when data is written to theboth serversApp①③Writing to active serverWriting to stand-by serverAppReading from active server①②NICNICNICMirroringPrimary14 NEC Corporation 2019NICMirroringStand-byPrimaryStand-by

4. Features of remote clustering② Asynchronous mirroringPrevents application performance degradation due to narrow network* In some products, function corresponding to this function is called semi-synchronous.In case disk mirroring is done through narrow band network with largecommunication delay such as remote clustering, asynchronous mode mirroringcan prevent degradation of the write performance.Write flow① EXPRESSCLUSTER receives writetransaction from application② At the same time as writing to disk, thedata will also be written to queueApp①③②③ Writing transaction completes whenwriting to the disk and queue is done④NICMirroringActive NEC Corporation 2019QueueNIC④ The queued data will be written to the diskof standby server in background15Writing to active serverWriting to stand-by serverWriting to queueStand-by

4. Features of remote clustering③ Schedule mirroringSchedule mirroring* enables data synchronization at quiescent point !* Sometimes this function is called as “Asynchronous Mirroring”In case synchronization must be done ensuring quiescent point in database etc,schedule mirroring is helps you to automate the synchronization process1. During normal operation Update of data on standby server can beprevented by disabling the data mirroring2. During data synchronization Secure quiescent point, and mirror theupdated data Command to take quiescent point of Oracle, SQLServer,and MySQL is supportedAppAppNICNICNIC16Updated blocks NEC Corporation pendedStand-byPrimaryStand-by

④ Difference copy,used-area-only copy4. Features of remote clusteringEasy/rapid recovery by differential copy and used-area-only copy!After restoration of server damaged in disaster/failure, disk can beresynchronized automatically and brings back to normal clustering conditionDifferential CopyIf disk is not damaged① Data can be mirrored in background whileapplication is under operation at backup site.② Only blocks which was updated after site failurewill be mirrored①NICUsed-area-only copy① Data can be mirrored in background whileapplication is under operation at backup site.② Only used area in the disk will be mirrored, butnot the empty area①AppNICIf disk is renewedNICNICEmptyEmpty② Difference copyMain site17Updated blocks NEC Corporation 2019② Used-area-only copyBackup siteMain siteAppUsedBackup site

4. Features of remote clustering⑤Compression ofmirroring dataEfficient data transfer by compressing the data to be mirrored!Compress beforesending it to networkExtract beforewriting it to diskAlso acts as a antipeeping measure sinceit is compressedAverage 50% reduction in data size ascompared to the previous version(Results differ depending on file type)* This feature is only valid in asynchronous mirroring mode.Helps configuring remote cluster even with narrownetwork!18 NEC Corporation 2019

4. Features of remote clustering⑥ Virtual IP addressCluster configuration even among different segment is possibleby taking over IP addressSame IP address can be available after failover even between differentnetwork segment by leveraging RIPConnection with same IP .0.0.100/24192.168.0.10/24192.168.0.20/24During failover, EXPRESSCLUSTERupdates the routing information19 NEC Corporation 2019192.168.1.30/24

4. Features of remote clustering⑦ Dynamic DNS integrationTaking over the host name when RIP or VPN is not availableAs DNS can be updateddynamically at the timing offailover, user can access to theserver without changing anyconfiguration at client side.1. IP address will changeat the time of failoverDynamic DNSServer3. Acquires currentIP address from DNSby unique host name2. EXPRESSCLUSTER X notifieschange of IP address to thedynamic DNS serverMakes remote clustering more easier, more convenient!20 NEC Corporation 2019

⑧ Automatic / manualfailover option4. Features of remote clusteringSupports customers’ demand to failover automatically in case of failure,and failover manually in case of disaster① Automatically failover if failure can be recovered within site.(Manual operation is also available by configuration)② In case of switching to another site during disaster etc, manual failover can beperformed. (Automatic operation is also available by configuration)WAN/VPNSelects automaticallyor manually①IP address/Host nameBusiness21IP addressHost nameApp NEC Corporation 2019②Redundancy inbackup site①IP address/Host nameAppIP addressHost nameBusiness

4. Features of remote clustering⑨ Manual failbackEasy operation to failback to main siteStep-1)Recovers main site. Cluster settings can be easily recoveredby push-button from GUI EXPRESSCLUSTER WebManagerStep-2)Resynchronization of data to main site by differential copyStep-3)Failback can be done by a few clicks from GUI③Easy failback from GUIIP addressHost nameIP addressHost nameBusinessBusiness①Example: Main site recovery Renewal of failed hardware Setup of OS / software Setup of EXPRESSCLUSTERNIC② Re-synchronization of the dataMain site22 NEC Corporation 2019NICBackup site

5. Case studies

5. Case studiesCase -1 ManufacturingEXPRESSCLUSTER has successfully deployed forvarious business continuity requirementsMajor Deployments of EXPRESSCLUSTER DR tanceSystemConfiguration1,500km DatabaseOne-to-one data mirroring50km DatabaseOne-to-one data mirroring260km DatabaseOne-to-one data mirroringFinance50km DatabaseHybrid clusteringFinance80km DatabaseHybrid clusteringInfrastructure390km DatabaseHybrid clusteringManufacturing100m DatabaseManufacturing5km DatabaseOne-to-one data mirroring500m DatabaseConstruction390km DatabaseHybrid clusteringFinance100km Log Collection1 to 1 mirroringPublic Sector350km Database1 to 1 mirroringDatabaseBuilding-A(Main Site)DatabaseOne-to-one data mirroringRetail50km DatabaseHybrid clusteringRetail15km Custom App1 to 1 mirroringManufacturing120km Database1 to 1 mirroringManufacturing120km DatabaseHybrid clustering NEC Corporation 2019(Backup Site)Three-to-one data mirroringPublic Sector24Building-B Adopted as business continuitysolution against fire disaster

5. Case studiesCase -2 Manufacturing- Remote cluster of factory between 5km distance (Campus Cluster)- Consolidated 3 standby servers to 1 physical server on VMwareMain siteFile serverFailoverBackup siteFailoverFile wareFailoverCustom Application25 NEC Corporation 2019Custom Application

5. Case studiesCase -3 Finance- EXPRESSCLUSTER was adopted to take over both “data and business application”- Dynamic cost reduction achieved compared to traditional storage-based replicationDistance: 50km, NW: 100Mbps, Data update: 40GB /dayMain siteBackup siteManualFailoverDownAppKey Decision Factor Failover can be performedautomatically within same site incase of server failure In case of disaster, failover can beperformed manually User can judge whether to failoveror not in case of disasterAutomatic FailoverAutomatic Failover Failover can be also automatic The process of switch over of thesystem can be pre-defined Prevents manual operation errorand realizes immediate recoveryAsynchronous data mirroringShared storage26 NEC Corporation 2019Shared storage Easy failback operation afterfailover testing