HP P4000 SAN Solutions technical presentation Daniel Stamm Senior Technology Consultant Storage daniel.stamm@hp.com 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
Storage Clustering, iscsi and Multipathing 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
Storage consolidation HP StorageWorks Portfolio Consolidation and Performance HP Network Storage System Simple Unified Storage iscsi SAN and Optimized NAS with integrated snapshots, backup, replication, and simple management. Windows, Linux, VMware, more. MSA200 0 Low cost consolidation 4GB FC, 1GbE iscsi, and 3Gb SAS SAS & SATA together Controller-based snapshot/clone Windows, Linux, HP-UX, VMware + more WEB, Exchange, SQL LeftHand SAN Scalable, Clustered IP SAN iscsi, 10GbE optional Scalable Performance SATA & SAS HDD Cost effective high availability VMware, Citrix, Hyper- V, Windows, Linux EVA Outstanding TCO FC & iscsi host ports FC & FATA HDD Storage consolidation + disaster recovery Simplification through virtualization HP-UX, Windows, Linux, VMware + more Mid Market MS, Oracle & SAP applications Business Continuity and Availability XP20000 XP XP24000 Always-on availability Data center consolidation + disaster recovery Large-scale Oracle/SAP applications HP-UX, Windows, + 20 more, including NonStop & mainframe
HP LeftHand SAN Appliance internal redundancy A HP LeftHand SAN Appliance provide internal redundancy for certain components. Redundant fans, Redundant power supplies, Redundant NICs Hardware RAID for disks Hardware RAID for disks can be selected during initial setup of a node and cannot be changed once the node is added to a cluster Available Hardware RAID levels RAID 5 8 Disk model: 7+1 12 Disk model: 2 sets 5+1 RAID 6 8 Disk model: 6+2 12 Disk model: 2 sets 4+2 RAID 10 8 Disk model: 4 drives striped and mirrored to 4 drives 12 Disk model: 6 drives striped and mirrored to 6 drives 4
SAN/iQ storage virtualization engine SAN Platforms Hardware RAID = R1/10, R5/50, R6 Manager Mgmt Services Manager Storage Server / IO Server RAID (10, 5/50, 6) @ 256KB Chunk Storage Server / IO Server RAID (10, 5/50, 6) @ 256KB Chunk Target Services (iscsi) Storage I/O Services (Storage Server & I/O Server) Distributed Cluster Management (Manager) Target Services (VIP, Target Failover, iscsi, MPIO) Virtual Volumes (Volumes, Disks, LUNs) Data Pages @ 256KB (Virtualized Disk Blocks) Manager Storage Server / IO Server RAID (10, 5/50, 6) @ 256KB Chunk iscsi 5
HP LeftHand SAN components HP LeftHand Central Management Console Clients Client Network Exchange SQL File Application Servers iscsi Initiator IP Storage Network (1 or 10GBit) Storage Node iscsi Target Volume Exchange SQL data File Storage Cluster HP LeftHand SAN Storage Cluster Management Group 6
SAN/iQ Storage Clustering iscsi Cluster for reliability, performance and ease of management Customer problems Large up front investment Inherent performance bottlenecks Storage controller upgrades Silos of storage Troubleshooting performance issues CPU NICs Storage SASNode Disks Cache Hardware RAID Redundant Cooling Redundant Power SAN/iQ Storage Software HP LeftHand storage clustering 7 Pay as you grow, scale out architecture An all inclusive feature set -> SW! Scale performance non-disruptively Manage all systems via single pane of glass Integrated performance monitor Centralized Management Console Cluster I SATA Cluster II
Network RAID and Multi-Site HA 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
HP LeftHand Network RAID Data availability beyond hardware RAID Customer problems System or box failures Human errors Power failures Air conditioning failures Availability/utilization trade off P4000 Cluster A D A D A B A B C B C B HP LeftHand Network RAID C D C D Data availability in addition to hardware RAID Data availability implemented at the network layer Adjust data availability per volume The goal for SAN availability is "no nines," or 100% availability. (Gartner,2007) Human error and firmware bugs are the weakest links, even in properly deployed SANs. (Gartner, 2007) 9
HP LeftHand Multi-site SAN Real-time protection from site failure Customer problems Traditional arrays only protect against component failures HW RAID Redundant components No protection against System failure Room maintenance Data center outage HP LeftHand Multi-Site SAN Keep data online during: Facility disruption Natural disaster Site maintenance SAN/iQ ClusterSAN/iQ Multi-site SAN A D A D A B A B C B C B C D C D Volumes Remain Online 10
SAN/iQ network RAID levels Network RAID Level-0 Network RAID Level-2 Network RAID Level-3 Network RAID Level-4 Aka: 0-way replication in the UI 1-copy (Stripe) of the data for the volume Aka: 2-way replication in the UI 2-copies (Mirror) of the data for the volume Aka: 3-way replication in the UI 3-copies (triple-mirror) of the data for the volume Aka: 4-way replication in the UI 4-copies (quad-mirror) of the data for the volume SAN/iQ Cluster SAN/iQ Cluster SAN/iQ Cluster SAN/iQ Cluster A A D A C D A B C D B A B A B D A B C D C B C A B C A B C D D C D B C D A B C D 11
5 features with every LeftHand SAN 12
HP LeftHand P4000 thin provisioning Reduce costs, increase capacity utilization Customer problems Full Provisioned Unused capacity in volumes Setting growth increments and thresholds Performance impact Reserving capacity for snapshots Guessing snapshot sizes Allocated 200 GB Available 0 GB Used 050 100 GB Wasted 200 150 100 GB 50 GB written 50 GB written 200 GB Volume Waste HP LeftHand Thin Provisioning 13 Allocate on write, no performance impact No reservation of capacity for volumes No thresholds to set, complete automation No guessing at snapshot size No up front reservations for snapshots Allocated 50GB 100GB Available 200GB 150GB 100GB Used 50GB 100GB Wasted 0GB Thin Provisioned 50 GB written 50 GB written 200 GB Volume Available
HP LeftHand P4000 snapshots Simplify backup and recovery Customer problems Backups take too long Recovery from tape is difficult Traditional snapshots waste capacity Traditional snapshots are hard to manage SAN/iQ Cluster 1:00 HP LeftHand snapshots Instantaneous snapshots Mountable snapshots simplify recovery Thin provisioning snapshots No reservation required Microsoft VSS integration included 1:00 2:00 3:00 14
HP LeftHand P4000 Remote Copy Implement cost effective disaster recovery Customer problems Central backup at remote sites Managing bandwidth Failover/failback difficult Inefficient capacity utilization at recovery site SAN 1 SAN 2 Recovery Server HP LeftHand Remote Copy Scheduled or manual Thin provisioned, no reserve space required Incremental failback Bandwidth management Failover / Failback Wizard Integrated in to VMware Site Recovery Manager 1:00 2:00 3:00 Baseline Copy Incremental Copy Incremental Copy 15
iscsi and Multipathing 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
iscsi SAN networking best practices Set up a separate network for your SAN Subnet or VLAN Multiple subnets are supported for multi-site environments Switches Any 1GbE switch is supported NIC bonding ALB (recommended) 802.3 (slightly better on writes, but requires switch) LeftHand Networking Best Practices Guide 17
iscsi terminology iscsi login: The process of establishing an iscsi session between an initiator and a target iscsi session Establishes the initial path for IP communications and non-persistent meta data about the connection An initiator opens one session per volume Cluster Virtual IP (VIP) Each cluster has a unique virtual IP address iscsi initiator connects to the Virtual IP of the LeftHand cluster (iscsi login) During iscsi session login, iscsi target redirects initiator to a physical node IP based to establish the actual iscsi session VIP is hosted by a given node, but will failover to another in case of a node problem For multi-subnet clusters (see later chapter) a cluster has one VIP per subnet Node IP: The physical IP Address of a cluster node Multi-path I/O (MPIO): The ability to use multiple network paths to the same target Device Specific Module (DSM): A vendor-specific module for multi-path I/O 18
virtual IP: standard session load balancing Server 1 writes to D (located at node 4) D iscsi initiator connects to VIP of cluster A LeftHand redirects initiator to physical node (based on least #connections per node) B C D HP LeftHand Storage Cluster Cluster VIP Vendors offering VIPLB. Dell EQL Compellent Isilon Pivot3 Intransa NetApp 19
HP LeftHand multi-pathing options Support Load balancing/ path selection Inbox iscsi initiators Virtual IP load balancing Standard mode of all modern in-box iscsi initators Static redirected to one node at iscsi session logon Decision based on # connection per node not changing until next logon or path failure Failover To next node with least # connections Failback/ re-balance Not before next session logon HP LeftHand MPIO DSM for Windows Dynamic load balancing Currently the Windows-only DSM can be downloaded free of charge from www.hp.com Dynamic, iscsi sessions opened to all nodes in a cluster Path for I/O selected based on data location (avoiding internode communication) I/O done to node that holds copy Automatic failback 20
HP LeftHand MPIO DSM: block load balancing Server 1 writes to A (located at node 1) A D Server 1 writes to D (located at node 4) iscsi initiator connects to VIP of cluster A LeftHand MPIO DSM does open an iscsi session to each cluster node. B C Cluster VIP D HP LeftHand Storage Cluster 21
Virtual SAN Appliance (VSA) 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
Virtual SAN Appliance for VMware ESX high availability for remote/branch offices Full featured virtual SAN Centralized Management Console LeftHand within an ESX virtual machine Virtualizes ESX server s internal and direct attached disk resources VSA VSA VSA Up to 10TB per license Only SAN appliance on VMware SAN/Storage HCL LeftHand cluster within ESX A C A B B C Highly Available storage across multiple ESX systems Shared storage for VMs iscsi Speicher 23 Download laptop demo version and 30 days VSA for ESX trial at www.hp.com/go/tryvsa
V S A V S A V S A V S A HP LeftHand P4000 VSA Two main use cases Internal shared storage for BladeSystem Cost effective, easy and scalable shared storage for HP BladeSystem Turn internal and direct attached storage into a full featured P4000 SAN Grow within AND beyond the BladeSystem P4000 Cluster Logical Representation Physical Representation Shared storage for branch offices and remote sites Replicate between remote or to central site Central backup Central management No storage hardware, infrastructure or staff required at remote sites Branch office VSA VSA Branch office VSA Central Data Center VSA Branch office VSA VSA 24
VSA support Supports any internal or direct attached storage Internal disks, SB40c, MDS600, other direct attached storage 10 TB per VSA instance Cluster of multiple VSA instances can support much more All LeftHand features and functions Network RAID, Thin provisioning, Snaps, Smart Clone, Remote Copy, Managed from same management console Can replicate remotely to physical HP LeftHand SAN Performance About 90% on IOPS compared to physical SAN About 70% on throughput compared to physical SAN Virtual SAN Appliance for VMware ESX 25
P4000 management 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
Starting the live demonstration 27
LeftHand SAN components in the CMC Available storage nodes Nodes which are powered on but not in a Management Group Management Groups - Are groups of storage nodes within which one or more modules are designated as Managers. Everything having to do with LeftHand storage is contained within the management group Clusters - Clusters are sub-groupings of nodes within a management group. However a cluster can be as small as a single Storage Module Volumes - Volumes are data storage areas created on clusters. Snapshots - Snapshots are read-only copies of volumes created at specific points in time. 28
Virtual Connection Manager Helps manage the complexity of SAN connection proliferation Streamlines mapping servers to volumes Embedded storage virtualization allows for a single step mapping process Simplifies navigation of storage environments Start with servers, find volumes Start with volumes, find servers Servers added to the GUI One step server-volume mapping 29
SmartClone User Interfaces 30
HP LeftHand - Advanced API/CLI Enables integration and automation of storage tasks Simplifies management and reduces costs by enabling integration and automation of management tasks Provision virtual machines from a single interface Integrate and automate backup Manage and report on capacity Collect data, monitor, analyze trends Integrate with Virtual Desktop Manager, Virtual Machine Manager, XenCenter Full coverage of LeftHand feature set available Based on SMI-S management standard Included for Partners and Customers at no additional cost 31
HA with Managers 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
What is a Manager? A Manager is a process that runs on one or more storage nodes. It controls the following tasks for the entire Management Group Data replication Keeps track of system status Coordinates reconfigurations as nodes are brought up and taken offline Re-synchronize data when a module fails and recovers Provide majority/quorum decision in case of node failure/split A manager is associated with a management group NOT a cluster. There can be less managers than nodes in a management group Typically not more than 5 managers even for large group of nodes Manager Identification 33
Quorum of Managers You must maintain a QUORUM of managers - they control data flow between modules in the Management Group Typical configurations Number of managers Number for a quorum Fault tolerance 1 1 None 2 2 None 3 2 Medium 4 3 Medium 5 3 High A QUORUM by definition is half plus one so the best number of managers to maintain are odd numbers (aka majority) If QUORUM is lost You lose access to your data Until quorum is regained You do not lose any data XM XM M Cluster: 34 Question: What about a 2 site setup? Can I lose one complete site? Answer: Yes, see next slide!
F ailover Manager Failover Manager For multi-site SANs or two-node environments Designed to break the tie if a site fails and provide automated fail over An instance of HP LeftHand SAN software optimized to run the manager only Runs in a ESX Virtual machine (like VSA) Recommended configuration Run at a third site if not possible run at primary site with most important applications M M F Site C Site A 2 managers + 1 failover manager = Quorum Cluster will stay online M M Site B Cluster: 35
V irtual Manager Virtual Manager An additional manager used only to regain quorum Must be added to the Management Group prior to the loss of quorum. Remains unassigned until it is needed to regain quorum Started manually on one node by user User starts Virtual Manager on one of the nodes Typical situations A management group in a single location with two storage nodes A management group across two sites AND no failover manager set up or Failover Manager lost OR any other situation where quorum is lost M V M Site A 2 managers + 1 virtual manager = Quorum Cluster will start again M V M Site B Cluster: 36
P4000 product offering 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice
HP LeftHand SAN solution portfolio - overview Physical SAN appliances For central main sites Multi-Site SAN solutions All features included No extra licenses Virtual SAN appliances For small & remote sites Virtual SAN Appliance for VMware ESX Virtualization SAN solutions Expansion Nodes SAS/SATA Starter SAN solutions 38
The HP LeftHand SAN appliance 39
HP LeftHand SAN Solutions Comparison P4300 SATA Starter SAN P4300 SAS Starter SAN P4500 Virtualization SAN P4500 Multi Site SAN Cluster Nodes & Disk Drives Starts with 2 nodes &16 drives Starts with 2 nodes &16 drives Starts with 2 nodes & 24 drives Starts with 4 nodes & 48 drives Starting Capacity 12.0 TB 7.2K SATA 4.8 TB 15K SAS 10.8 TB 15K SAS 21.6 TB 15K SAS Dual Active/Active Controllers (two per site) Every SAN/iQ feature 3 Yrs 9x5 Software Support 3 Yrs Hardware Maintenance On-Site, Next Business Day Customer Self Installable Windows Solution Pack Remote Office Solution Pack optional optional (10 included) (10 included) Expansion nodes optional optional optional optional 40 Site 1 Site 2
Limitation / Best Practices 10 Nodes per Storage Cluster is best practice 80 Disks with P4300 / 120 Disks with P4500 Up to 32 possible (technical limit) Up to 1000 Objects (Volumes, Snapshot, etc.) per Management Group 2ms latency for Multisite Cluster / 50 MB/s per node 4000 iscsi Initiator assigned to LUNs per Management Group 50 Snapshots per LUN recommended 41
Performance P4000 SAN Appliance 42
Thank you! 2008 Hewlett-Packard Development Company, L.P. The information contained herein is subject to change without notice