DATA MIGRATION BEST PRACTICES NetApp Glbal Services January 2006
Intrductin Fr IT managers, data migratin has becme ne f the mst rutine and challenging facts f life. Wrklad balancing, technlgy refresh, server and strage cnslidatin, data center relcatin, data classificatin, and mergers/acquisitins are all drivers t mve data frm ne strage device t anther n a regular basis. Yu d think that any peratin perfrmed rutinely wuld becme easy. But nt s. Migratin mving data frm ne device t anther (especially as the capacity f strage devices cntinues t grw) and then redirecting all I/O t the new device is an inherently disruptive prcess. With the increase in the percentage f missin-critical data and the prprtinate increase in data availability demands, dwntime with its huge impact n a cmpany s financial bttm line becmes unacceptable. In additin, business, technical and peratinal requirements impse challenging restrictins n the migratin prcess itself. Resurce demands staff, CPU cycles, and bandwidth and risks applicatin dwntime, perfrmance impact t prductin envirnments, technical incmpatibilities, and data crruptin/lss make migratin ne f IT s biggest challenges. Since the majrity f strage systems purchased by custmers is used t stre existing rather than new data, getting these new systems prductin-ready requires that data be cpied/mved frm the ld system t be replaced t the new system being deplyed. As the services arm f ne f the industry s leading strage cmpanies, NetApp Prfessinal Services has therefre develped a cre cmpetency in data migratin, finetuning its migratin methdlgy ver time, with each migratin services engagement delivered. Whether the migratin is a same vendr system upgrade (fr example, NetApp FAS940 t NetApp FAS3050) r a vendr takeut (fr example, EMC Celerra t NetApp FAS3020), whether the migratin is perfrmed by internal IT r an external services prvider, the migratin methdlgy is the same. Migratin Methdlgy Althugh the term data migratin is ften used t refer t several different peratins, this paper will fcus n data migratin as the prcess f mving data frm ne strage device (surce) t anther (destinatin), as in a technlgy refresh r an archiving ( HSM ) peratin. All migratins fllw the same basic methdlgy illustrated in Figure 1. 2
Business Impact Analysis Discvery Mapping & Design Creatin f Migratin Plan Prvisining Pre-Migratin Test Cutver Final Signff Migratin & Cutver Migratin Validatin Figure 1) Migratin methdlgy. Althugh the ultimate gal f a data migratin is t mve data, a lt f upfrnt planning needs t happen prir t the mve in rder t ensure a successful migratin. In fact, planning is the number-ne success factr fr any migratin prject, independent f the cmplexity. Nt nly des upfrnt planning help shrten the duratin f the migratin prcess, but als it reduces business impact and risk fr example, applicatin dwntime, perfrmance degradatin, technical incmpatibilities, and data crruptin/lss. The migratin plan the end result f the planning defines what data is mved, where it is mved, hw it is mved, when it is mved, and apprximately hw lng the mve will take. There are a few key envirnmental factrs that influence the migratin methdlgy selected hw the data is mved. The first factr is the number f tiers within the strage hierarchy that are included in the destinatin. Single-tier migratin is a migratin in which all surce data is migrated nt ne r mre devices within a single tier that is, primary strage. Within a single-tier migratin, three ther factrs are imprtant in determining the methdlgy: Strage vendr: Data layut: Hmgeneus: ne vendr Hetergeneus: multiple vendrs One-t-ne mapping: surce and destinatin have the same strage layut 3
Relayut: surce and destinatin (ne r bth f which can be ne r mre devices) have different strage layuts Access prtcl: NAS (file) SAN (blck) Multitier migratin is a migratin in which data is migrated t different tiers f the hierarchy based n a best fit between data/applicatin requirements and strage tier attributes. There are tw types f multitier migratin: One-time migratin: data placement recmmendatins, based n the results f an analysis f the data and the strage tiers in the hierarchy (primary filer, ATA filer, and NearStre ), are prvided Cntinuus and plicy-driven migratin: unlike all ther data migratins, this is a cntinuus prcess (requiring the installatin f an archival applicatin) that uses a set f user-defined plicies t determine, in real time, when data needs t be mved where Figure 2 illustrates the basic migratin envirnments defined by these key factrs. cmplexity single tier multi-tier ne-time NAS plicy-based & cntinuus SAN ne-time NAS plicy-based & cntinuus SAN hmgeneus hetergeneus cmplexity Figure 2) Migratin envirnments. All migratins fall int ne f the abve migratin envirnments. Business Impact Analysis The bjective f the business impact analysis is t identify the business and peratinal requirements that impact the migratin prcess. Varius stakehlders within the cmpany need t be cnsulted t ensure that their requirements are factred int the migratin planning. IT staff defines available netwrk bandwidth and CPU cycle lad, allwable dwntime, and migratin schedule 4
Database and system administratrs define applicatin and database requirements Business wners define the imprtance f, and requirements fr, specific applicatins and types f data Security and cmpliance grups define cmpliance requirements Discvery Details abut the migratin hardware and sftware envirnment are cllected during the discvery step. The migratin envirnment whether a hetergeneus SAN migratin r a hmgeneus NAS migratin determines the specific data that has t be cllected. Fr example, in a NAS migratin envirnment, current I/O wrklad (IOPS) n the servers, number f files/flders/vlumes invlved in the migratin, share permissins (NFS, CIFS, r bth), netwrk cnfiguratin, DNS and AD dmain infrmatin, and s n are imprtant. A SAN migratin envirnment has different discvery requirements. Accurate and cmplete discvery is critical t the success f the migratin, s althugh discvery can be manual (via cmmand line) r autmated (via a discvery tl), using an autmated discvery prcess is the preferable ptin. Nt nly des autmated discvery reduce discvery time, but als it minimizes the chance f errr. Unfrtunately, n ne discvery tl prvides all the required data, s ften multiple tls must be used. Mapping and Design During discvery, the migratin surce what data is mved is identified. During mapping and design, the secnd half f the equatin, the migratin destinatin where is the data mved is identified. There are tw basic mapping layuts: ne-t-ne, where the surce and destinatin layut are the same, and relayut, where the surce and destinatin layuts are different. Althugh a ne-t-ne mapping enables a much simpler migratin, migratin is ften seen as an ptimal pprtunity t cnslidate and/r ptimize perfrmance and/r capacity utilizatin, s relayut is a very cmmn scenari. A cmbinatin f the migratin gals and best practices drives the layut f the new (destinatin) strage envirnment. Creatin f Migratin Plan The migratin plan, which is the end deliverable f the planning phase, functins as the blueprint fr the migratin implementatin, specifying custmer expectatins, defining prject deliverables, and identifying migratin methdlgies t be used. There are fur majr inputs int the migratin plan: Business and peratinal requirements, which prvide the cnstraints Data t be migrated, with all assciated attributes Available migratin tls Strage and applicatin best practices Creating a wrkable migratin plan is ften quite challenging. Different types f data may require different migratin tls and strategies, and business and peratinal requirements the dwntime windw may require creative ways f mving the data. The migratin plan cmpnents shuld include, but are nt limited t: Migratin strategy and key activities Dependencies Required equipment and migratin tls Custmer expectatins (custmer business, technical and peratinal requirements) Test plan 5
Verificatin prcedures Risks and cntingency plans Change cntrl prcedures Prject schedule Pstimplementatin activities/respnsibilities Migratin cmpletin criteria Althugh the migratin plan is the end deliverable f the planning phase, it is really a living, rather than static, dcument. Variables in the envirnment can change, r executin can lead t unexpected results, impacting the migratin plan as dcumented. Prvisining During prvisining, the destinatin strage envirnment is prepared fr the data mve. LUNs, vlumes, directries, and s n are allcated, security attributes are set, and shares/exprts are created. Prvisining fr a ne-t-ne mapping is simple; fr a relayut, it is a mre cmplex task. Hwever, by using infrmatin generated frm the mapping and design step, it is pssible t autmate many f the prvisining tasks. Premigratin Test Befre any data is mved, it is imprtant that sme prtin f the migratin plan with the scpe depending n the specific situatin and the agreement with the custmer be tested and validated. Results f the migratin test determine whether mdificatin f the migratin plan fr example, timeline, migratin tls used, amunt f data migrated per sessin, and s n is required. Fr example, if testing shws that allwable dwntime wuld prbably be exceeded, the migratin methdlgy needs t be revisited. Migratin and Cutver This is the pint within the migratin prcess when the surce data can be mved t the destinatin devices. There are many ptins fr mving data, each with its wn advantages and disadvantages: Migratin utilities can run n the hst, strage device, r netwrk appliance/switch Data can be mved in tw basic ways: Out-f-band (r ut f the data path), by creating a baseline cpy, mving the baseline cpy via media transfer r electrnically, and then applying incremental changes, red lgs, and s n Inband (r in the data path), by creating a mirrr in real time and then breaking the mirrr The ptimum migratin methdlgies vary significantly, based n many factrs, including: Surce and destinatin strage systems Netwrk strage tplgy (NAS r SAN) Equipment lcatins Specific applicatins and data usage Custmer business, technical, and peratinal requirements Available/supprted migratin utilities Many vendrs ffer prducts that play in the migratin space, s there are many chices. 6
Basically, there s n ne right way; it s all abut making trade-ffs based n knwledge f the envirnment, business and peratinal requirements, and migratin experience, which prvides a prven set f best practices. In fact, the entire planning phase fcuses n develping the best-in-class migratin plan fr the specific migratin prject at hand. By migratin and cutver, all the upfrnt planning, analysis, and trade-ffs have been made; the migratin plan serves as the implementatin guide. Once the data has been mved, all clients must be redirected t the destinatin devices. Migratin Validatin Befre the migratin can be cnsidered a success, ne critical step remains: t validate the pstmigratin envirnment and cnfirm that all expectatins have been met prir t cmmitting. At a minimum, netwrk access, file permissins, directry structure, and database/applicatins need t be validated, which is ften dne via nnprductin testing. Internal r External? The disadvantage f ding a task ver and ver is that it can get pretty bring; the advantage is that yu fine-tune yur prcedures, learn t recgnize the pitfalls and develp wrkarunds, and becme prficient in ther wrds, yu becme an expert. Althugh data migratin has becme a rutine peratin within IT, when migratins are cmplex, new technlgy is being intrduced int the envirnment (fr example, NetApp strage r FlexVl technlgy), r migratin skills are in shrt supply, bringing in an external services prvider makes a lt f business sense. NetApp Prfessinal Services cmplements yur internal capabilities, wrking with all the key stakehlders t understand gals and requirements and prviding Migratin expertise and field-prven methdlgy and tls Incrpratin f best practices Risk mitigatin Prfessinal prject management thrughut the migratin prcess lifecycle, including: Creating a statement f wrk (SOW) and btaining custmer apprval Identifying the varius custmer stakehlders Establishing a migratin team, cmpsed f bth custmer and service prvider representatives Creating a migratin plan and btaining custmer apprval Prviding migratin prgress reprts Cmmunicating issues and their impact and recmmending alternatives as required Obtaining custmer prject acceptance based n predefined cmpletin criteria Data migratin is and will cntinue t be a way f life fr IT as wrklad balancing, technlgy refresh, server and strage cnslidatin, data center relcatin, data classificatin, and mergers/acquisitins drive the need t mve data frm ne strage device t anther. Netwrk Appliance, Inc. Cmpany Cnfidential and Prprietary. 2005 Netwrk Appliance, Inc. All rights reserved. Specificatins subject t change withut ntice. NetApp, the Netwrk Appliance lg, and NearStre are registered trademarks and Netwrk Appliance and FlexVl are trademarks f Netwrk Appliance, Inc. in the U.S. and ther cuntries. All ther brands r prducts are trademarks r registered trademarks f their respective hlders and shuld be treated as such.. DS-XXXX Rev. XX/XX www.netapp.cm 7