3PAR Inform OS

HP Confidential | October 2010,30,DOUBLE-LINE TITLEFOR long titles,Click to edit Master text styles,Second level,Third level,October 2010,3PAR Inform OS,Course Objectives,At the end of this presentation the student should be able to :,Understand InFormOs Software Architecture,Understand InFormOS RAID concepts,Understand concepts of InServ Chunklets,Understand InFormOS Logical Disks (LDs),Understand InFormOS Virtual Volumes,PD (Physical Disk),Chunklet,Mesh-Active Cluster,Wide Striping,Virtual Volume,Set Size,Step Size,Virtual LUN,CPG (Common Provisioning Group),Region,Host,VV Set, Host Set, (Autonomic Group),Key Concepts and Terms*,*,There will be a quiz.,3PAR Combines the Best Attributes of Modular and Monolithic Storage Arrays,point-to-point or switched connection,point-to-point or switched connection,Host Connectivity,Data Cache,Disk Connectivity,Traditional Monolithic Architecture,Traditional Modular Architecture,3PAR InSpire,Architecture,A finely, massively, and automatically load balanced cluster,Legend,InSpire Architectural Differentiators,AutonomicManagement,3PAR,Gen-3 ASIC,Fine Grained Virtualization,Simplified Storage and Server Provisioning,3PAR Rapid Provisioning,Seconds to provision storage with no pre-planning,Self-configuring and auto load-balanced storage volumes,Labor-intensive and careful planning eliminated,3PAR InServ Virtual Volume,A virtual disk. Assembled by policy. Backed by every physical disk of the specified service level.,Virtual,Volume,Host Sees Virtual Volume as a LUN,Every disk array is “Virtualized” to some degree,Physical disks are re-mapped into LUNs that are exported to hosts,What makes 3PAR different?,LUN,Chunklets,Physical drives broken into fine-grained uniform 256 MB chunklets,First Things FirstHow to Build a VV (Virtual Volume),3PAR SAs LOVE to talk about CHUNKLETS,Chunklets are the very first step in building 3PAR storage,Chunklet size is fixed at 256MB for all drive sizes,Chunklets are like miniature physical disks, the lowest common denominator,Physical Disk Chunklets (256 MB),C,= 256 MB Data Chunklet,SC,= 256 MB Spare Chunklet,C,C,C,C,C,C,Physical Disk,SC,SC,SC,Each InServ Physical disk is initialized with data chunklets and spare chunklets,3PAR InServ Data Layout,256,L,D,LD,LD,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,256,L,D,Drive Magazines have,4 of the same drives,FC 15, SATA, SSD etc. Drive magazines can be mixed inside the same drive chassis.,Nodes are added in pairs for cache redundancy.,An InServ with 4 or more nodes also supports,“Cache Persistence”,Which enables maintenance windows and upgrades without performance penalties.,Drive Chassis are,point-to-point connected,to controllers nodes in the T Series InServs to provide “cage level” availability.,The ability to withstand the loss of an,Entire,drive enclosure without losing access to your data.,Each Physical Drive,is cut up into “Chunklets”,each 256 MB in size.,Each VV created is automatically widely striped across chunklets on all disk spindles of the same type (Fiber channel, SATA, etc.,Creating a massively parallel system,RAID sets ( 3+1 R5 for example will stripe the 4 members of the RAID set into separate chassis. Massively striping data and enabling high availability.,RAID Sets are bound together to form logical disks,Logical Disks are bound to, serviced by and load balanced across all Nodes,All LDs are bound together transparently to form a single volume. And presented to the host across all ports and all nodes. Enabling a TRUE active/active configuration on a PER LUN basis. Not Active/passive on a per LUN basis as other architectures.,3PAR Wide Striping: Completely and Finely Load Balanced,8GB LUN spread across 64 disks (RAID 10),8GB LUN spread across 4 disks (RAID 10),Channel Directors,Cache Boards,Disk Directors,Market Leading,Monolithic Array,InServ T400Storage Server,Chunklets are Automatically Grouped into Global Pools by Drive Rotational SpeeD,SATA (Nearline) disks are usually the biggest pool (depends on configuration),Fibre Channel disk chunklet pool,Solid State disk chunklet pool is smaller and reserved for high-value I/O,Why are Chunklets so Important?,Same drive spindle can service many different LUNS, and many different RAID types at the same time,Enables array to be managed by policy, not by administrative planning,Drives up utilization rate of the disks (no leftover space),Improved High Availability and Sparing,Enables easy mobility between physical disks, RAID types, service levels,Several Layers of Virtualization to go from PD to VLUN,VLUN,Logical Disk,Logical Disk,Regions,Common Provisioning Groups,Common Provisioning Groups (CPG),CPGs are FUNDAMENTAL to administering 3PAR.,CPGs are the combination of a RAID type and a drive type which equals SERVICE LEVEL and AVAILABILITY LEVEL.,CPGs have many functions:,They are the policies by which free chunklets are assembled into logical disks.,They are a container for existing volumes and used for reporting,They are the basis for service levels and our optimization products.,High Availability + High Utilization Rate,3PAR Availability Level Concepts:,HA Cage,HA Magazine,HA Port,These are policies by which the system lays out RAID sets in order to protect against hardware failures.,The default configuration will always be the “safest” but can be overridden by an administrator.,Default HA level “cage” unmatched in industry.,3PAR RAID Levels,Inform OS can assemble RAID sets (RAIDLETS) into RAID 00, RAID 10, RAID 50, RAID 60,Data to Parity Ratios (3PAR Set Size),Raid 10 mirror or 3-way mirror,Raid 50 from 2+1 to 8+1 (HA-magazine),Raid 60 at 6+2 or 14+2 (protect against double disk failure with same capacity tradeoff of RAID 50,3PAR RAID 10 Concepts,RAID 10 is “mirrored” data,Data is written as “paired” chunklets,Each “chunklet” on the RAID set is on a different physical disk,Setsize = 2,Default size (RAID 1),Usable space,= 256 MB,C,C,3PAR RAID 5 Concepts,(1 of 2),RAID 5 uses parity to reconstruct data,RAID 5 uses a setsize equal to number of drive cages (shelves) by default to guarantee cage level HA.,Setsize = 4 (3+1),Default size (RAID 5),Usable,space = 768 MB (3*256),C,C,C,p,Setsize = 6 (5+1),Usable space = 1280,MB,(5*256),C,C,C,c,c,p,C,C,p,C,C,p,C,C,3PAR RAID 5 Concepts,(2 of 2),C,C,C,c,c,p,C,C,C,Setsize = 9 (8+1),What is the usable space for this setsize?,Usable space = 2048 MB (8*256),3PAR RAID Multi Parity (MP) Concepts,RAID MP uses parity (double parity can deliver data in a double disk failure) to reconstruct data and performed in the ASIC XOR engine,RAID MP only supports two setsizes ( 8 and 16),The default set size of 8 has the same data to parity ratio as the default RAID-5 set size of 4 - 3:1 in both cases.,23,23,Usage Meaning of “-ha cage” For RAID MP,The system defines -ha cage as will tolerate the failure of ONE cage.” This means:,Up to two chunklets are allowed to share the same cage.,The default R6 “-ha cage” set size 8 requires 4 cages per node-pair, just like the default R5 set size 4.,If 8 cages are available, the layout will use one chunklet per cage.,The same rules apply to “-ha mag” up to two chunklets are allowed per mag but the system will place only one chunklet per mag if possible.,24,24,Virtual Volume Step Size,“Step Size” is The number of contiguous bytes that the system accesses before moving to the next chunklet.,Step size varies based on RAID type.,R10 default is 256K,R50 default is 128K,R60 default varies by Set Size.,Step size is how we avoid hotspots on the back end disks!,NOTE:,Administrators can override defaults, but not recommended,Unified Processor and/or Memory,Control Processor & Memory,3PAR ASIC & Memory,disk,Heavy,throughput,workload,applied,Heavy,transaction,workload,applied,I/O Processing: Traditional Storage,I/O Processing: 3PAR Controller Node,hosts,hosts,small IOPs,wait for,large IOPs,to be processed,control information and data are pathed and,processed,separately,Heavy,throughput,workload,sustained,Heavy,transaction,workload,sustained,Disk interface,=,control information (metadata),=,data,Host interface,Host interface,disk,Disk interface,Mixed Workload: Ready for any application,InServ HA: Built for the demands of multi-tenancy,Traditional Arrays,3PAR InServ,- RAID 1 Group -,- RAID 1 Group -,- R1 Raidlet -,- R1 Raidlet -,Shelf-dependentRAID,No accessto data,Shelf (chassis),Shelf (chassis),Shelf-independentRAID,Data access preserved,Drive,Shelf,(chassis),y,Traditional CacheMirroring,Poorperformance (write-thru mode),PersistentCacheMirroring,Consistent performance(no write-thrumode) *,Controller Node,DiskDrive,Few-to-one rebuild,Hotspots* & prolongedrebuild exposure,Many-to-many rebuild,Non-disruptiverebuilds in the time,sparedrive,spare chunklets,Component,3PAR Host Definition,Host Definition (Stand alone or into Host Set),Host Persona - Different operating systems require slightly different behaviors both from the port level and the SCSI layer,Follow Implementation Guides Per Host OS,VLUNS are what you get when you match a Virtual Volume (VV) to a host definition (or host set),Clustered environments add additional complexity to storage provisioning,VMware clusters create new storage management challenges,1 cluster of 5 hosts and 10 volumes, requires 50 provisioning actions on most traditional arrays!,At 10 min/action: More than 1 day!,Error-prone,VMware clusters are dynamic resources subject to growth and frequent change,3PAR Autonomic Groups eliminates repetitive error-prone tasks,3PAR Autonomic Groups,Simplifies and automates volume provisioning in a single command,Exports a group of volumes to a host or cluster of hosts,Automatically preserve same LUN ID for each volume across hosts,When a new host is added into the host group:,All volumes are autonomically exported to the new host,When a new volume is added into the volume group:,New volume is autonomically exported to all hosts in the host group,Quiz!,Explain the concept behind “setsize” for RAID.,What does an InServ Virtual Volume look like to a host?,What is the size of a 3PAR “region” and why is this important?,Where is data stored when a disk drive is being rebuilt?,Can you specify a “setsize” of 9 if you have 4 drive cages?,Name 3 functions that CPGs provide.,