7 Must-Know Disaster Recovery Strategies Jamie Gruener Senior Analyst, Enterprise Infrastructure The Yankee Group What You Will Learn In This Session . . 1. The Main Goal: Bullet-proofing Disaster Recovery Strategies 2. Remote Replication vs. Mirroring Strategies 3. Disaster Recovery and Virtualization: A New Angle 4. Integrating Mirroring and Replication into Disaster Recovery Strategies 5. Metro SANs: Options for Remote Mirroring And Replication 6. Your Network Options: Questions to Guide You The Goal: Bullet-Proofing Disaster Recovery Strategies • You’ve heard the statistics – ~ $5 billion in computing infrastructure lost in 9-11 tragedy – Close to 100 businesses declaring disasters in NYC • You’re here for a reason – 9-11 was a call to action for all of us – Disaster Recovery and Business Continuity Strategies must be done – this is only one leg of a multi-leg strategy • You’re entering a new era – It is now about Business Continuance and Risk Management Defining Terms: Remote Mirroring and Data Replication Remote Mirroring generates a mirrored image of Checklist • data on two or more disks • Data Replication scans data periodically for changes and copies new data to the other disk or file system on another system • Factors to consider – – – – Value of data (and lost data) being backed up Costs for network bandwidth and software Your existing infrastructure Product features (OS, File System, Disk or Application) Measuring Data Protection: Point-in-Time to Synchronous Tip More There will be a need for multiple tools to protect data Data Protection Synchronous Semi-Synchronous Lost Transactions Line of Tolerance Asynchronous Point In Time Less Network Bandwidth Consumed More Questions to Consider When Looking At Remote Mirroring and Replication – What are you protecting? (applications, transactions, files, disks) – What level of protection do you need? (We’ll come back to this!) – What are your network requirements? – What is your expected budget for this project? – Is the ROI greater than acquisition costs? – What will best fit your larger business continuity strategy? Best Practice The Luxury Sedan: Disk-to-Disk Remote Copy • Key advantages – Operates at the disk level – Can be (not always) less complex to set up and administer than host-based approaches – Can offer the benefit of capturing all application changes. . . • Key Disadvantages: Costly – Lacks transaction knowledge or what the data represents – Can be wasteful of network bandwidth if not properly set up The Rising Alternative: Block-and-File Replication • Fundamentals – Makes comparisons and only copies changes (at disk or file level) • Key Advantages – Can be less expensive – Can be flexible to replicate all enterprise data regardless of disk system – Copies only the most important files/data – Many-to-one replication architectures available – Limits amount of data transferred, reducing network load and cost • Key Disadvantages Warning Block and File Replication Details You need to Consider File WAN Replication Block WAN Replication • Less expensive, host-based (or array-based) file and block replication • Specific to storage vendor, OS or file system Database Replication • Typically done by replicating or mirroring log files • A number of variations: – Continuous Mirroring: updates DBMS as changes(adds, updates, deletes) occur – Change Data Capture: captures DBMS changes and stores them until a predetermined replication time – Full Copy Refresh: replicates entire DBMS copy to target systems (done to resynchronize DBMS after outage) • Trigger-based native DBMS is not usually appropriate for DR because of high system and Snapshots: The Hotshots of Backup • Establishes a separate identifiable storage entity and run operations against it – Primary purpose: backup, testing, conversion and batch process • Is dependent on OS, host and array support… • Advantage: it takes up less network bandwidth than mirroring • Disadvantage: resynchronization of data is an art – If you do not resynchronize, you must build snapshot mirror totally from scratch • New Kid on the Block: Virtualization and Disaster Recovery Virtualization software offers a new alternative – – – – Data replication (over both IP and Fibre Channel) Snapshot High Availability Fail-over A cost-effective approach to disaster recovery • Key Challenges – Recreating the virtualization system can be difficult – Specifying file level information for replication can be difficult Warning – Still a relatively new technology, so test well – Research virtualization players thoroughly • Ask hard questions about number of customers doing this Prioritizing Data: Integrating Mirroring or Replication to Your Strategy Tool Hot Site More Remote Disk Mirroring Data Replication Importance Of Data Snapshot Campus Disk Mirroring Electronic Vaulting Tape Onsite Less Tape-Backup Offsite More Less OK, Now What? It’s The Network, Stupid • Long-distance Remote Mirroring/Replication requires significant network integration – Mission: Connecting two or more islands of storage – Could be SANs, hot sites, remote disk or tape • Myriad of network transport choices boil down to two fundamentals (from the POP out) – Fibre – IP • Where to Start: Evaluate Network Requirements – Ask storage vendors for requirements – Map that to service provider bandwidth services • Specific Network Challenges for Remote Mirroring and Replication Enterprise network performance is many times slower than storage performance • Things You Could Consider – What tools do the remote mirroring/replication vendor provide for performance on a Metro SAN? • Network throttling – adjusting data amounts sent over wire • Compression – compressing data to take up less network bandwidth • Time-stamping – marking data at time saved or accessed – Latency isn’t your friend • Measurable time it takes for an I/O transaction to reach destination • Distance is a factor – especially when extending data native limit • Storage traffic requires high bandwidth, low latency connections Tip Metro SAN Puzzle Pieces Headquarters Key Features Remote Site Remote Tape Library Remote Disk Storage Primary Disk Storage ATM, T3/E3, OC3+ or Fibre Optic (DWDM) Service Provider Connectivity Services Mirroring/Replication Software Edge of Network Equipment (Director, Router or DWDM) Metro SAN Network Transport Options Key Features Network Protocol Performance Length of Distance ESCON Full performance (200-M bits/sec. unidirectional) 8 km for full performance, 50% performance @ 20 km FICON Bi-directional channel protocol, runs over at 1.063 G bit/sec. 100 km distance limitation Dark Fibre Dedicated Fibre (depends on transport) 10 km without long-wave transceivers iSCSI, iFCP, FCIP Still in proof-of-concept Depends on applications, stage, but promise of 1 TB service, connection points – per hour over IP FCIP - primary for Metro SANs DWDM - Gigabit (Optical) Ethernet Requires high-speed IP network to connect Depends on applications, service, connection points and IP Network Options • Private Router Backbones – Leased, dedicated lines – Optimized for performance (racing the sun) • xSP VPNs – Customer purchases edge routers and ISP provides shared backbone – Leverages Multiprotocol Layer Switching (MPLS) for better Quality of Service • Internet . . . – Not a disaster recovery tactic I would trust Checklist • Tips To Consider When Speaking With Your Service Provider Ask your xSP to provide you what their expertise is in storage services – Consider how you could leverage your existing connectivity services to establish a cost-effective service contract for your backup services – Ask for a latency/network performance SLA – Your primary xSP should have partners to assist with storage services if they don’t • Consider leveraging storage vendor relationships to find service providers that can do integration • Do or have done a network analysis to determine requirements for your backup services • • • • • • • Metro SAN Challenges: Things You Need to Ask about IP Checklist Latency and Bandwidth Security Complexity Quality of Service (QoS) Dropped Packets Manageability NOTE: IP storage switch vendors are trying to solve these things – so gauge them based on this Does Outsourcing Make Sense? Best Practice • Determine if outsourcing disaster recovery services for Data Mirroring and Replication services – Do I have the skills, personnel and infrastructure? • Outsourcing provides a number of advantages – Enables enterprise to focus on core competence – Speeds IT’s ability to maintain and return to business operations – Leverages expertise and more plentiful specialized resources – Offers mirrored solutions for immediate recovery – Provides problem resolution expertise that speeds delivery of services related to backup, security and performance monitoring Ask for Help From a Professional . . . • This is a complex proposition • Storage vendors will assist with best backup technologies • Service Providers will offer options for services • Integrators with storage and networking practices are best candidates for additional implementation – they are still rare in the world Evaluate ROI for Disaster Recovery • Come up with a ROI calculation for your data protection strategy • Guidelines to remember Best Practice – Calculated over three-year term – The higher the ROI, the more favorable the project – TCO vs. ROI – ROI wins out • Key factors to be included in ROI analysis – – – – Costs: downtime, personnel, assets Reliability: data replication, mirroring, backups System performance: how is systems and software utilized Calculate savings: improved performance, revenue, processes Conclusions • Don’t forget the fundamental goal: Disaster Recovery • Consider your options for mirroring/replication and make a metric to measure them by – ROI analysis is one way of measuring options • Be cautious of new technologies – New IP storage networking technologies and software offer new options – but you need to be savvy in deployment • There’s not one solution – strategize on ways to prioritize your data protection Questions? Jamie Gruener jgruener@yankeegroup.com