EMC S200 manual

Whi te Paper Abstra ct This white pape r details ho w the EMC Isilo n OneFS archi tecture provides high ava ilability and dat a pr otecti on n eeded t o m eet the chall enges organizations fa ce as they deal with the deluge o f digital conte nt and unstructured data and the growing importance of data protection. November 2013 HIGH AVAI LABILITY AND[...]

2 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Copyright © 2013 E MC Co r por at ion . A ll R igh ts R eser ved. EM C be lie ve s the info r mat ion in t his pu bl ica t ion is a cc ur at e as of its publicat ion date. The infor mation is subject to c hange without not ice. The info r mat ion in th is pub l ica tio [...]

3 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Table o f Cont ents Intro ductio n ................................................................................................... 5 High av ailabi lity and reliab ility at the peta byte sca le ..................................... 5 Area l densit y and dri ve reb ui[...]

4 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Replic atio n ................................................................................................... 23 Syn c IQ ................. ............................................. ......................................... 23 SyncIQ linear re store .............[...]

5 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Introducti on Today , organ iz ation s of all siz es acro ss th e full spectru m of the bu sin ess aren a are fa cin g a s im ila r p ro ble m: A n e xp los io n in t he sh ee r q uan tity o f f ile - based data they are gen erat ing and, b y vi rtue, ar e fo rced to man[...]

6 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS la w. Howev er, th e reli abil ity an d perfo rman ce of disk drives a re n ot in creasi ng a t the same pa ce, and thi s is compo unded by the growing amount of time it tak es to rebuild dri ves. Large cap aci ty disks, su ch as t he cur rent t hree and fou r te rabyte [...]

7 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Furt her al ong th e cont inuu m li e the da ta reco very ap proa che s in ord er of d ecrea sin g tim eliness. Th ese sol utions typically include a form of point - in - time snapshot s for fa st recov ery, follo wed by synchrono us and async hronous replication . Final[...]

8 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 2: Isilon S cale - out NAS Arc hitectur e OneFS a r chitectur al o verview OneFS col lap ses the t rad iti onal el ement s o f the st orage st ack — data protection , volu me man ager, fi l e system , etc . — into a single, unified so ftware layer (see Figure [...]

9 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS re liab le , ult ra lo w - latency back - end clu ster com muni cati on. These RBM messag es — every thing from cluster heartbe at pings to distributed lock ing control - are t hen proce ssed b y a nod e’s L ocal Block Manag er vi a th e Devi ce Wo rker Th rea d (DWT[...]

10 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS clien t access p rotocol s - across mult iple hea ds. The GMP is built on s everal distribute d al gorith ms and strict ly a dheres t o B rewer’ s Th eorem, w hi ch stat es th at it i s i mpossi ble for a distributed c omputer system to simultaneo usly gu arantee a ll[...]

11 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS All files, inodes and other me tadata structures (B - trees, etc) wit hin OneFS are eit her mi rrored up to ei ght times or p arity protected, with the data spre ad across the variou s disk cyl in der groups of mul tiple nod es. Pari ty prot ection uses an N+M schem e w[...]

12 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 4 : OneFS H ybrid Par ity Protection S chemes (N+M :x) As men tion ed earl i er, OneF S als o provi des a vari ety of m irro ring opti ons rang ing from 2x to 8x, allowing fr om two to eight mirro rs of the specified co ntent. Metadata, f or exampl e, i s mirro r[...]

13 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS With Automatic Pr ovisioning, ev ery set of equivalent node har dware is automatica lly divided into no de pools comprising up to forty nodes and six drives per node. Thes e node poo ls are protected by default at N +2:1, and multiple p ools c an then be co mb ine d int[...]

14 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS OneFS f ault t olerance File s ystem j our nal Every Is ilon node is equipped with a dual - battery backed 512MB NVRAM card, which guard s that nod e's fi le syst em j ournal . Each jou rnal is used by On eFS as stab le stora ge, and guards write tra nsactions agai[...]

15 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS MediaScan MediaScan’s role within OneFS is to check disk sectors and deploy the above DSR mechan ism in order to force di sk dri ves to fi x an y sect or ECC err ors th ey ma y encoun ter. Impl ement ed as one of t he ph ases of th e On eFS job engi ne, Med iaS can i [...]

16 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS and t he crea tion of i ncreasi ngl y larg er sin gle vo lu mes and file sy stems, ty pical recov ery ti mes fo r mul ti - terabyt e dri ve fai lu res are b ecomi ng m ulti ple day s or e ven weeks. During this MTTDL peri od, stora ge system s ar e vuln erabl e to ad di[...]

17 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Some of thes e met hod s are b iased towa rds c ost effi ci ency bu t hav e a h igh er risk associ ated wit h th em, and others repre sent a hig her co st but also o ffer an i ncreased level of protect ion. Tw o ways t o measu re cost versus risk fr om a da ta protectio[...]

18 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 6: Isilon D ata Protection technolog y alignm ent wit h prot ection continuum Connecti on l oad b alancing and failover SmartConnect As mentioned previously, at the leading edge of the data pro tection continuum lies high av ai lab il it y. Th is no t only inc lu[...]

19 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 7: Se amle ss Clien t F ailove r with S martC onnec t Snapshots SnapshotI Q Next along the high availability and data protection co ntinuum ar e sna pshots. The RTO of a snaps hot can be very small and the RPO is also highly flexible w ith the use of rich poli ci[...]

20 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS available sna pshots is via a /. snapshot hidden director y under each file syste m di rectory. I silo n SnapshotI Q can a lso creat e unl imit ed snapsh ots on a clust er. This pr ovides a substantial bene fit over the major ity of other snapshot implementat ions bec a[...]

21 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 9: Snaps hot Ch ange Tra ckin g Snapshot s cheduling Snap shot sch edu les ar e conf igu red at a dai ly, w eekl y, mont hly or yearl y i nterval , wi th sin gle or mu lt ipl e job freq uency per schedu le, d own to a per - minute granularity. And automatic delet[...]

22 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Snapsho tDelet e, it is only run if the group is in a p ristine state, i.e., no drives /nodes are dow n. Th e mo st ef fic ie nt m et ho d fo r d ele tin g mu lt ip le sn ap sho ts sim ult ane o us ly is t o process older through newer , and SnapshotIQ will automatica l[...]

23 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 10: F ile Clo nes S hadow stores also provide the basis for Isilo n Smart Dedup e TM softwa re . Smart Dedup e maxi mi zes th e sto rage e ffici ency of a cl uster b y dec reasi ng th e amoun t of physical stor age required to house an organization’s da ta . Ef[...]

24 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 11: D isast er Re covery with Sync IQ I silo n SyncIQ TM soft ware del ivers hi gh - p erforman ce, asynch ron ous repl i cation of unstructur ed data to address a broad range of recover y point objectives (RPO ) and recovery time objectives (RTO ). This enables [...]

25 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS SyncIQ l inear r es tor e Leve ragi ng OneF S Snapsho tIQ infras tructure , the Li near R est ore fun ction ali ty of Syn c IQ is ab le to det ect and restore (commit) consiste nt, point in time, block - level changes between cluster replication se ts, with a minimal im[...]

26 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS SyncIQ f ailover and f ailback In the event that a pri mary clu ster b ecome s un avail abl e, Sync IQ provi des th e ab ili ty to fai lover t o a mi rro red, DR clu ster. Du ri ng su ch a sc enari o, th e admi nistrat or ma kes the decision to r edirect client I/O to t[...]

27 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Contin uous re plication mode Complementar y to the manu al an d schedul ed repli cation p olici es, SyncI Q also off ers a continuo us mode , o r rep lica te on change , option . When the “Wh eneve r th e sour ce is modified” po licy co nfiguration o ption is se le[...]

28 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS inde f in ite ly . Additiona lly, SmartLock s atisfies the regulator y compliance demands o f stringent data r etention policies, including SEC 17a - 4. Data E ncryption at Rest I silo n a lso pr ov ide s a solu t ion for the secu rity of data at rest. This involve s de[...]

29 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS based ta pe and VTL infrastructure. This is enabled by the EMC Isil on A100 Backup Accel erato r (BA) nod e, whi ch fea tures a quad - port 4GB /s Fi bre Ch annel card , qua d - core p roces sors, and 8 GB of RAM. A s ing le Is ilon A100 Backup Accelera tor can concurre[...]

30 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 15: Back up Using S napshotIQ Parallel s tr eams Isilon’s distr ibuted archite cture all ows backup s to be sp read across mu lti ple netw ork streams from th e clu ster, whi ch can signi fican tly improv e perfor mance. Thi s is achieved by dividing the roo t [...]

31 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS • Restore - to - arbi trar y sys tems • Seamless integration with acc ess control lists (ACLs ), alternate data streams and resou rce f orks • Sel ectiv e Fil e Recov ery • Replicate the n backup While some back up software vendors may support backing up OneFS o[...]

32 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Figure 16: R ecomme nded Two - w ay ND MP w ith B ackup Acc eler ator Remote NDMP m o del In the remot e NDM P sc enari o, th ere is n o Back up A ccel erator p resen t. I n thi s cas e, the DMA uses NDMP over the L AN to instruct the cluster to start backing up data to[...]

33 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Incremental b acku ps Isil on OneF S acc ommod ates t he ran ge of ful l, incr ement al an d tok en - based backups. In standard D R nomenclature, Level 0 indicates a full backup, and levels 1 - 9 ar e in crementa ls. A ny l evel speci fi ed as 1 - 9 will back up all th[...]

34 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Summary Organizations of all sizes aro und the globe are dealing with a de luge of digital content and unstructur ed data that is driving mas sive incr eases i n st orage need s. As these enter prise datasets continue to expand to unprece dented sizes, data p rotection [...]

35 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS Isilon a cronyms g lossa ry[...]

36 Hig h Availability and Data Protecti on w i th E MC Isilon Scale - o ut N AS About E MC EMC C orpo rati on is a globa l l eader i n enab ling busi nesses and s ervi ce provi der s to tran sform th ei r operat ion s and d eli ver IT as a ser vice. Fund ament al t o thi s tran sformat ion i s cloud computing. Thr ough innovative products and servi[...]