Scale-out Storage Solution And Challenges PDF Free Download

2y ago

55 Views

1 Downloads

1.17 MB

27 Pages

Report/dmca

Download PDF

Transcription

Table of Content Overview of Scale-out StorageScale-out NAS Solution Architecture Storage IO Workload Distribution Development ChallengesScale-out NAS Testing Test Strategy Test Challenges Open Source ToolsScale-out NAS – Next2013 Storage Developer Conference. iGATE. All Rights Reserved.2

Scale-out Storage OverviewDistributed scalable storage Distributed scalable IO computing Infrastructure versus system Interoperable Fault tolerant Infrastructure manageability 2013 Storage Developer Conference. iGATE. All Rights Reserved.3

Scale-out NAS Approaches Scale-out NAS with no metadata Uses consistent hashing (e.g. MD5 algorithm) for distributionand look-up of files in the cluster Scale-out NAS with some metadata Uses consistent hashing and maintains some metadata fordistribution and look-up of files in the cluster Scale-out NAS with metadata Uses pre-defined parameters (e.g. file count, capacity of thestorage node) for distribution of files Metadata is used for file look-up2013 Storage Developer Conference. iGATE. All Rights Reserved.4

Scale-out NAS ApproachesScale-out NAS with nometadata Highlights Fast look-up No single point offailure as nometadata server Good scalability Lowlights No load balancing File movement incase of rename,node add RebalancingmandatoryScale-out NAS with somemetadata Highlights Good load balancing Only metadatamovement and no filemovement in case ofrename, node add Metadata can bedistributed onmultiple servers toprevent single pointof failure Good scalabilityScale-out NAS withmetadata Highlights Look-up time couldbe proportional tonumber of files Good load balancing No file movement Lowlights Large metadata Metadataconsiderations forscalability Lowlights Average look-up time2013 Storage Developer Conference. iGATE. All Rights Reserved.5

Scale-out NAS Components –Storage ControllerMeta DataServer-YClient-1File Meta DataNFSStorageController-1Meta DataServer-1File DataStorage Node-1 Client-NCIFSStorageController-XStorage NodeZStorage Controller provides: Global Namespace- Provides a single unified global file system view to clients Metadata Management- Distribution and lookup of file metadata using consistenthashing Data Management - Distribution and lookup of file data using weight-based (on filecount, node capacity) algorithm Client Request Management – Accepts and responds to client requests Automatic failover by deploying redundant controllers2013 Storage Developer Conference. iGATE. All Rights Reserved.7

Scale-out NAS Components –Metadata ServerMeta DataServer-YFile MappingMeta Data Replicationand RebalanceMeta DataServer-1Storage Node-1Storage NodeZ Metadata server provides: Repository for directory hierarchy File to Storage Node mapping Metadata servers are clustered for: Scalability - distribution of file metadata uniformly enables metadata workloaddistribution and capacity scaling Availability - replication of metadata Elasticity - metadata servers can be added or removed in cluster; metadata isautomatically rebalanced2013 Storage Developer Conference. iGATE. All Rights Reserved.8

Scale-out NAS Components –Storage NodeStorageController-1File DataStorageController-NFile DataStorage Node-1Storage NodeZData Replicationand Rebalance Storage Node provides: Storage: File based data storage (Native file system) Scalability: Nodes can be added/ removed non-disruptively Load Balancing: Data Rebalancing when nodes are added/ removed Availability: Policy based data replication IO Handling: Services IO requests2013 Storage Developer Conference. iGATE. All Rights Reserved.9

Storage IO Workload eNode-NStorageNode-1 StorageNodeStorageNode-NSingle large file IOParallel read/write of filechunks to multiple StorageNodes Multiple files IOParallel read/write toStorage Node withmultiple Storage NodeManagers2013 Storage Developer Conference. iGATE. All Rights Reserved. Multiple clients IOParallel read/write bymultiple StorageControllers10

Technical ChallengesChallengeSUN-RPC based daemons are single threaded (in Linux)IssueHandling simultaneous requestsSolution Multiple SUN RPC server processes Dynamically spawn/kill processes depending upon workloadChallengeIn TCP/IP, sockets once released are not available immediately, toavoid denial of serviceIssueNo socket available for Client-Server communication after athresholdSolution Connection pool of sockets TCP/IP tuning (tcp fin timeout, ip local port range,tcp tw reuse, tcp tw recycle, tcp max syn backlog,netdev max backlog, somaxconn) SUN RPC tuning (tcp fin timeout, rpc timeout)2013 Storage Developer Conference. iGATE. All Rights Reserved.12

Technical Challenges cont.ChallengeDirectory tree is replicated on metadata serversIssueSynchronizing simultaneous directory operations across metadataserversSolution Use of master metadata server for synchronization. All directory operations except readdir() are routed through mastermetadata serverChallengeFiles are distributed across storage nodes leading to directory treereplication on each nodeIssueSynchronization and performance issues associated with create,delete, rename operation of directories on each storage node dueto replication of directory tree on multiple storage nodesSolution Maintaining the directory tree at the Metadata Server File is uniquely identified by prefixing directory inode to file name2013 Storage Developer Conference. iGATE. All Rights Reserved.13

Test Strategy - Traditional v/sScale-out NASTraditional NAS Test Considerations Designed to handle low volume of dataPerformance bottleneck when multiple users access simultaneouslyCapacity scalingPerformance degradation as data traffic growsRisk of data unavailability in case of NAS head failureScale-out NAS Test Considerations Ability to scale and store huge dataLoad balancing across nodes – Performance improvementCapacity scalingDynamic scaling by on-the-fly addition of nodesScale-out NAS Testing challenges Huge data creation for scalability testing Performance testing to consider node scale-out and scale-down conditions Initiation of multiple simultaneous operations from multiple clients2013 Storage Developer Conference. iGATE. All Rights Reserved.14

Scale-out NAS Test genceScalabilityInteroperabilityStress2013 Storage Developer Conference. iGATE. All Rights Reserved. Orchestration Software upgradesand maintenance User accountmanagement GUI complexity Ease ofdeployment

Scale-out NAS Test nceScalabilityInteroperabilityStress2013 Storage Developer Conference. iGATE. All Rights Reserved. Authentication(RPC/ AD/ LDAP /NIS) Network Security Data security Antivirus support

Scale-out NAS Test StrategySecurityManagementMulti lityStress2013 Storage Developer Conference. iGATE. All Rights Reserved. NFS, CIFS, FTP,HTTP protocols BackwardCompatibility Infiniband/FCinteroperability Support of SSH,SCP for backendoperations

Scale-out NAS Test igenceScalabilityInteroperabilityStress2013 Storage Developer Conference. iGATE. All Rights Reserved. Determine storagecontrollerthreshold foroptimumperformance Determine capacitythreshold perstorage controllerfor linearperformance File systemcapacity

Scale-out NAS Test IntelligenceScalabilityInteroperabilityStress2013 Storage Developer Conference. iGATE. All Rights Reserved. Support to thirdparty applications Backupapplicationsusing NDMPsupport Cloud APIsupport fordifferent vendors

Scale-out NAS Test genceScalabilityInteroperabilityStress2013 Storage Developer Conference. iGATE. All Rights Reserved. Failover/ Failback Replication Redundant storageaccess path NIC teaming – NICfailover / failback

Scale-out Storage – Next Cloud integration (OpenStack etc.) STaaS OpenStack Swift IntegrationBig Data Hadoop Platform for BI AppsSolid State Device Scale-out NAS SSD array PCIe SSD cardsSoftware Defined Storage (SDS) SLAs & QoS Orchestration & Storage HypervisorHealthcare or other Domain/Industry Online data, history, diagnosis, medication, billing, etc.2013 Storage Developer Conference. iGATE. All Rights Reserved.26

Scale-out NAS Components – Storage Node Storage Node provides: Storage: File based data storage (Native file system) Scalability: Nodes can be added/ removed non-disruptively Load Balancing: Data Rebalancing when nodes are added/ removed Availability: Policy based data replication IO