Storage Virtualisatie Sjef van Ham Storage Support Meerheide 101 5521 DZ Eersel s.van.ham@storagesupport.nl 26-maart-2009
Storage Opslag Direct Attached Storage NAS Network Attached Storage DAS Direct Attached Storage NAS Network Attached Storage SAN Storage Area Network
Wat is een SAN ? Gecentraliseerde opslag op basis van een extern schijvencabinet voor verschillende typen servers « alles in één doos » gecentraliseerd beheer hoge schaalbaarheid door veel schijfruimte hogere hardware redundantie en databeveiligingsfuncties : RAID, spare disks 2) Speciaal storage netwerk Fibre-Channel grote afstand (optische kabels ) hoge bandbreedte (4Gb) iSCSI ethernet direct toepasbaar en flexibel lage TCO en infrastructuurkosten
Voordelen van gecentraliseerde opslag Flexibel en eenvoudig toewijzen van capaciteit Gemiddelde benutting in een DAS-omgeving is < 50% (SAN = 85%) Schijf en tape in hetzelfde opslag netwerk (50-75% minder tape drives) Minder LAN-verkeer Gecentraliseerd beheer Beheer met minder mensen (efficiency stijgt met factor 10) Opslag, switches en tape toevoegen zonder applicaties stop te zetten Zo veel redundantie als gewenst
iSCSI Storage iSCSI staat voor Internet SCSI (Small Computer System Interface). Dit is een op het Internet Protocol (IP) gebaseerde standaard voor IP netwerken. Deze technologie is een versmelting van twee technieken die zich in de industrie bewezen hebben. Het gaat om de technologie die ontwikkeld is voor communicatie (ip netwerken) en de technologie bedoeld voor dataopslag , SCSI opslag. Door deze te combineren werd het mogelijk zowel de data als de dataopslagcommando's over ip-netwerken te versturen. Een netwerk protocol voor storage netwerken Gebruikt de standaard ethernet tcp/ip netwerktechnologie eb hardware
Fibre Channel SAN Fibre Channels work on a single popular storage networking protocol that offer compatibility to all common storage and database servers. Fibre Channels use the FCP protocol ( Fibre Channel Protocol) which is simply a serial bus SCSI protocol. This FCP (Fiber Channel protocol) works really well in networks with a high bandwidth requirement. Fibre Channel is een protocol dat voornamelijk wordt gebruik om data binnen storage netwerken te transporteren. Vaak wordt Fibre Channel geassocieerd met alleen glasverbindingen, maar het protocol kan zowel over glas als koperverbindingen worden gebruikt.
iSCSI voordelen door Eenvoud Geen noodzaak meer om te investeren in FC technologie of FC kennis Ethernet is bekende technologie Ethernet heeft een lagere TCO en infrastructuur Geen noodzaak meer om te investeren in FC technolog of FC kosten Geen FC HBAs meer nodig, optionele iSCSI ethernet kaarten voor hogere prestaties
IP based versus FC based FC HBA’s Standaard Netwerk ip-kaarten 2-4-8 GB 1GBe-10GBe Fibre Kabels UTP kabels Fibre Switches Ethernet Switches 2-4-8 GB 1-10 GBe
De markt Fibre - iSCSI
Storage Requirements for the Traditional Data Center Performance Supports current workloads. No performance degradation Scalability Scale capacity and performance without disruption Availability Data stays online during a site failure, recover from disaster Manageability Simple to admister
Typical Storage Array Architecture Monolithic Array Not scalable Controller head becomes bottleneck Scales capacity only Single point of failure Forklift upgrades
WHICH CAN BE COMBINED IN OUR FRAMELESS ARCHITECTURE Scale when you need, as large as you need it, but view as a single SAN Performance Number of disks
Typical iSCSI SAN Pages RAID RAID 10, 50 or 5 Disks PS Series Array 1 Exchange SQL Archive Switched Gb Ethernet Exchange Archive SQL Storage Pool 1 Disks RAID PS Series Array 1 RAID 10, 50 or 5 Pages 2 N
Automatic and transparent load balancing across arrays Exchange SQL Archive Switched Gb Ethernet Exchange Archive SQL Storage Pool 1 PS Series Array 2 RAID 10, 50 or 5 Pages 2 2 3 RAID RAID 10, 50 or 5 4 4 Disks PS Series Array 1 N Automatic and transparent load balancing across arrays
The LeftHand Difference
SAN/iQ Storage Clustering SAS Storage cluster Aggregates all components for performance Data is load balanced across all nodes Predictable scalability Grow on your terms Non-disruptive scalability No forklift upgrades Scale everything Throttle bandwidth Create tiers of storage Each optimized for type of data Online volume migration Simple centralized management Provisioning Monitoring Security SATA Centralized Management Console
SAN/iQ Network RAID Beyond component redundancy Protects data from array failure Synchronous replication Configure on a per-volume basis Configure on-the-fly, no down time High availability Multiple disks, controllers, or arrays Zero disruption of data access Ensures “high availability” for data SAN/iQ Cluster A D A D A B A B C B C B C D C D
SAN/iQ Multi-site SAN Protect Storage By: Rack Room Floor Building Keep Data Online During: Facility disruption Natural disaster Site maintenance SAN/iQ Cluster SAN/iQ Multi-site SAN A D B A C B C D Volumes Remain Online
SAN/iQ Thin Provisioning Eliminate Initial Over-Purchasing No Wasted and/or over-allocated capacity Purchase what you need today Delay Future Storage Purchases Grow capacity as warranted Simple to Manage Enable/Disable with radio button on per-volume basis Allocate only as data is written to the volume No Reserve necessary SAN/iQ Cluster Full Provisioned Thin Provisioned Allocated 200GB Available 0GB Used 100GB Wasted 100GB Allocated 200GB Available 0GB Used 50GB Wasted 150GB Allocated 200GB Available 0GB Used 0GB Wasted 200GB Allocated 0GB Available 200GB Used 0GB Wasted 0GB Allocated 100GB Available 100GB Used 100GB Wasted 0GB Allocated 50GB Available 150GB Used 50GB Wasted 0GB 50 GB written 50 GB written 50 GB written 50 GB written 200 GB Volume 200 GB Volume
SAN/iQ Snapshots Simplified Backup and Recovery SAN/iQ Cluster Simplified Backup and Recovery Point-in-time volume copies Scheduled, Ad Hoc or Scripted Functional Volume rollback Recover granular data Backup Integration Microsoft VSS VMware VCB Space Efficient Thin provisioned always Delta changes only No snap reservation 1:00 1:00 2:00 3:00
SAN/iQ Remote Copy SAN 1 SAN 2 Asynchronous Replication Recovery Server Asynchronous Replication Per volume basis Scheduled or manual Thin provisioned No reserve space required Simple to Manage Bandwidth management Failover / Failback Wizard SAN 1 Vol_1 (Primary) SAN 2 Vol_1 (Primary) Vol_1 (Remote) 1:00 1:00 Baseline Copy 2:00 2:00 Incremental Copy 3:00 3:00 Incremental Copy
Network Storage Server (NSS) Intelligent Storage Virtualization True heterogeneous intelligent storage virtualization Application-aware data protection WAN optimized disaster recovery services protect operations and save on WAN costs Increased storage efficiency through Thin Provisioning of resources Your Storage, Your Infrastructure, Your Solution
TOTALLY Open Storage Virtualization Platform Exchange Oracle SQL VMware Notes DB2 SAP Windows Linux Solaris HP-UX AIX Netware iSCSI FC Infini- Band Any application Any storage protocol Any disk array Any operating system Data Virtual Physical Existing SAN volumes via Storage Service Enabler Virtualized SAN volumes Thin provisioned SAN volumes
Control storage management cost and complexity Consolidate heterogeneous storage management through virtualization Minimize storage costs with thin provisioning Centralized management of client systems simplifies operations
The Solution Starts with Consolidation File Servers Critical Applications Application Servers Backup Severs DAS DAS DAS DAS DAS DAS DAS SAN SAN Enhanced storage utilization Consolidated storage management Opens the door for a host of solutions and innovations
How storage virtualization works Quickly create virtual volumes from across or within any physical disk resources - Simple wizard interface - Physical LUN and virtual devices up to 16TB Expand volumes as needed Securely assign volumes to servers through NSS interface Virtual disks Physical disks
Storage abstraction layer Move the storage intelligence and security into the SAN fabric NSS creates a storage abstraction layer Storage services no longer dependent on vendor-exclusive disk functionality Consolidates services and operations Hosts Fabric Storage abstraction layer Disk
Thin Provisioning: Better Storage Utilization Application server sees virtual volume size Create a virtual volume to match size of data provisioning request Allocate a smaller amount of actual physical disk space Additional physical space allocated automatically as needed Reduces wasteful over provisioning of storage Saves disk space and power consumption “800 GB” Virtual 800 GB Physical 200 GB Data Thin provisioned SAN volume
Minimize downtime Highly available systems ensure continuous operation Cross-array mirroring for ultimate data protection Seamless, zero-downtime data migration
NSS high availability deployment model Shared Storage Failover NSS appliances deployed in failover pairs Each NSS server supports its own set of client servers during normal operations If one devices fails, the second device takes over servicing clients SAN
Data mirroring, zero downtime migration Any-to-any storage array synchronous data mirroring Ultimate data protection across arrays Seamless data migration across storage devices. Create mirror Switch primary/mirror status without any break in application processing Remove mirror (old array) Writes acknowledged from both arrays to ensure integrity Primary Mirror Mirror Primary
Ensure comprehensive data protection Application-aware snapshots provide rapid data recovery WAN optimized data replication for disaster recovery and failback Host-free, accelerated backup eliminates all backup windows
Application aware snapshots MS Exchange Oracle VMware DB2 Informix SAP MS SQL Sybase Notes/Domino Progress Pervasive SQL Up to 256 snapshots per volume Instant recovery to any captured point in time Fully mountable as read/write volumes Application agents and consistency groups for 100% transactional integrity Each volume has its own snapshot schedule 8:00 9:00 10:00 11:00 1:00 5:00 1:30 2:00 2:30
WAN Optimized Disaster Recovery Easily configured any-to-any remote replication Continuous or periodic modes … … DR Site Full replicas WAN optimized out of the box Includes WAN data encryption, compression and QoS Snapshots Copy to tape
WAN Optimized Data Replication 512 byte = empty disk sector Applications and file systems write far more data to disk than actual changes NSS scans the changes at a 512 byte level and replicates only the changed data - Avoids constant re- sending of duplicate data = data written by application or file system = sector with actual data change 512 byte Disk sectors A typical 8k write of data. Many systems send all 8k or more. CDP sends only the truly changed sectors.
VMware Advanced Features Because continuous uptime is so critical, VMware has created several advanced features designed to help system uptime - VMware High Availability - VMotion - VMware Distributed Resource Scheduler
VMware High Availability/Vmotion/Scheduler VMware High Availability automatically restarts virtual machines affected by physical server failure VMware ESX Server X VMware ESX Server VMware ESX Server Shared storage pool
The Power of the Virtual SAN The FalconStor Network Storage Server Virtual Appliance transforms internal disk storage into an iSCSI SAN - No additional storage network investment needed Enables VMware advanced features Cost effective, highly available infrastructure for the SMB and ROBO
FalconStor NSS Virtual Appliance Architecture Virtual Machines ESX SW iSCSI Adapter VMware ESX Server Virtual Machines VMFS Volume VMFS Volume IP Network VMware ESX Server NSS VA Physical Ethernet Adapter FalconStor NSS VA V-LUN Virtual Switch Virtual LUN for VMDK files Server Internal RAID
High Availability via NSS VA Cross-Mirror Virtual Infrastructure HA supported without SAN storage VMware ESX Server To client Standby NSS VA NSS VA ESX SW iSCSI Adapter V-LUN Same size virtual LUN Internal Storage VMware ESX Server Cross-over Gigabit connection NSS VA V-LUN NSS-VA Cross-Mirror Internal Storage Active Access Point Passive Access Point