Evo­lu­tion of the tra­di­tional data cen­tre

Network Middle East - - SPECIAL REPORT / MARKET FOCUS -

IT bud­gets for data cen­tres are go­ing down and Gart­ner pre­dicts that by 2025, 80% of en­ter­prises will shut down their tra­dional data cen­tres. With fu­ture pro­jec­tions in­di­cat­ing a mas­sive shift away from tra­di­tional data cen­tres, Net­work Mid­dle East ex­am­ines what the fu­ture of data cen­tres looks like

The amount of data gen­er­ated by 2025 is set to rise to 175 zettabytes and as such data cen­tres will con­tinue to play an es­sen­tial role in the stor­age, com­pu­ta­tion and man­age­ment of informatio­n. Ma­jor tech com­pa­nies are con­tin­u­ing to in­vest in new data cen­tres by buy­ing land near power sources for fu­ture sites. Across a wide range of in­dus­tries — from health­care to fi­nance to man­u­fac­tur­ing — com­pa­nies rely on data cen­tres to sup­port the grow­ing cre­ation and con­sump­tion of data.

How­ever, IT bud­get spend on data cen­tres has re­duced as per an­a­lysts. With Gart­ner pre­dict­ing a mas­sive shift away from tra­di­tional data cen­tres, what does the fu­ture of data cen­tres look like?

SMBS have evolved from rent­ing phys­i­cal servers, to vir­tual ma­chines, to Saas ap­pli­ca­tions, to fi­nally func­tion as a Ser­vice or server­less com­put­ing plat­forms. The terms for rent­ing have been re­duced in scope and time, with con­tract cy­cles com­ing down from months to min­utes. This trend will con­tinue to gain pop­u­lar­ity as Saas ap­pli­ca­tions are more cost- ef­fec­tive and take less time to get to market.

Shailesh Davey, vice pres­i­dent, Man­ageengine and a co-founder of Zoho Corp ex­plains how data res­i­dency reg­u­la­tions have made it manda­tory to re­tain data in en­ter­prise data cen­tres (EDCS) for larger en­ter­prises to stay in com­pli­ance.

But even larger en­ter­prises are mov­ing to the public cloud or Saas ap­pli­ca­tions to de­liver a bet­ter cus­tomer ex­pe­ri­ence and lever­age big data an­a­lyt­ics, like analysing cus­tomers’ so­cial me­dia data. The data within EDCS is con­trolled within a pri­vate net­work, and con­nects to the data in the public cloud us­ing se­cure tun­nels. This hy­brid cloud setup en­ables en­ter­prises to ac­cess data stored both pub­licly and pri­vately, help­ing them stay com­pli­ant.

Ac­cord­ing to Fadi Kanafani, manag­ing direc­tor and gen­eral man­ager – Mid­dle East, Netapp, this shift in the data cen­tre in­dus­try was an­tic­i­pated few years ago. Hence, Netapp is ad­vis­ing its cus­tomers to stop build­ing data cen­tres and in­vest in build­ing a Data Fab­ric in­stead. “Although the tra­di­tional data cen­tre as we know it will see a down­ward spi­ral, it will not dis­ap­pear. It will go through an evo­lu­tion driven by busi­ness out­comes.

The fu­ture data cen­tre is es­sen­tially soft­ware de­fined, bor­der­less, dy­namic and makes up a crit­i­cal part of the Data Fab­ric cir­cle of data man­age­ment.

As or­gan­i­sa­tions start mov­ing work­loads back from public to hy­brid clouds, it is im­por­tant to look at the ar­chi­tec­ture it­self. Is it open enough to al­low seam­less data mo­bil­ity? Is data grav­ity be­ing ad­dressed? What are the cost im­pli­ca­tions? Are data ef­fi­cien­cies avail­able in the hy­brid cloud? What is the se­cu­rity pos­ture that needs to be in place? Has data clas­si­fi­ca­tion been done? Is data tier­ing be­ing re­viewed as part of the data man­age­ment life cy­cle? And lastly, does the de­sign lend it­self for multi- cloud ar­chi­tec­ture? Hav­ing said that, the suc­cess of a hy­brid cloud

strat­egy will di­rectly tie into the de­sign of the Data Fab­ric ar­chi­tec­ture and ser­vices that have the an­swers to all these ques­tions – and many more.

Kh­waja Sai­fud­din, se­nior sales direc­tor, Mid­dle East at Western Dig­i­tal points out that hy­brid cloud isn’t nec­es­sar­ily about sep­a­rat­ing work­loads be­tween pri­vate cloud vs. public cloud. “A hy­brid cloud works in a par­tic­u­lar ap­pli­ca­tion use case where you can take best ad­van­tage of each plat­form. So, for ex­am­ple, if you have an ap­pli­ca­tion that re­quires a petabyte or more of stor­age ca­pac­ity then it is not a good can­di­date for cloud host­ing due to the cost. Up­loads to the cloud are gen­er­ally ex­empt from band­width charges, which means you can up­load your data, process it, re­trieve the re­sult, then delete it to avoid the stor­age fees,” adds Sai­fud­din.

By host­ing the data lo­cally, you can mi­grate your com­pute re­sources from one host­ing provider to an­other, with­out wor­ry­ing about mi­grat­ing the data. The re­sult is an ex­tremely ag­ile mul­ti­cloud en­vi­ron­ment, where you have ac­cess to the widest ar­ray of cloud ser­vices pos­si­ble, while cut­ting costs and pro­tect­ing your­self from ven­dor lockin all at the same time.

Davey fur­ther high­lights that to sup­port a suc­cess­ful hy­brid cloud im­ple­men­ta­tion, some im­por­tant con­sid­er­a­tions are avail­abil­ity of low la­tency con­nec­tiv­ity to mul­ti­ple public cloud net­works, iden­tity and ac­cess man­age­ment, same pro­ce­dures fol­lowed by the Devops team across both the public and pri­vate cloud en­vi­ron­ments and de­vel­op­ment of ex­per­tise to run vir­tual routers.

Scope for AI in data cen­tres

AI is now be­ing re­alised in real-world use cases in­clud­ing man­u­fac­tur­ing, se­cu­rity, farm­ing, au­to­mo­tive, health­care, ed­u­ca­tion and busi­nesses. The abil­ity of de­vices and ob­jects in­ter­act­ing through In­ter­net of Things to pour even more data al­low for greater rea­son­ing and re­ac­tion.

AI is play­ing an in­creas­ing role through ma­chine learn­ing, adds Sai­fud­din. “Those in au­to­mo­tive, EDA, and re­search em­ploy real-time data an­a­lyt­ics and ma­chine learn­ing work­loads ev­ery day. At the core of these emerg­ing work­loads are in­creased adop­tion of in-mem­ory data­bases and clus­tered data­bases that ben­e­fit more im­me­di­ately from low la­tency stor­age.”

Cloud ap­pli­ca­tions and net­work, stor­age, and HVAC equip­ment gen­er­ate a lot of teleme­try data, and need to be mon­i­tored and an­a­lysed for proper func­tion­ing of the data cen­tre. Wher­ever data is gen­er­ated in huge quan­ti­ties, there is a good op­por­tu­nity to utilise AI/ ML tech­nolo­gies to de­tect anom­alies and pro­vide pre­dic­tive in­sights and what-if analysis. Alert gen­er­a­tion, er­ror re­cov­ery pro­ce­dures, and pre­ven­tive ac­tions can be au­to­mated based on past data. AI/ ML tech­nolo­gies can also work in tan­dem with the on- call tech­ni­cians to re­cover from un­fore­seen sit­u­a­tions.

AI ini­tia­tives tend to be a very high re­source- con­sum­ing prac­tice due to high com­pu­ta­tional power re­quired and the ren­der­ing of the data needed for ma­chine and deep learn­ing. Netapp has teamed up with Nvidia, the leader in GPU com­put­ing, to help ac­cel­er­ate busi­ness out­comes through a val­i­dated de­sign in a con­verged ar­chi­tec­ture.

“This ap­proach can help data sci­en­tist be­come pro­duc­tive in a mat­ter of days in­stead of weeks and months – while, ad­min­is­tra­tively the so­lu­tion be­comes all con­tainer­ised to en­sure sim­plic­ity and agility. Although, these ar­chi­tec­tures are mainly on-premise, the plat­form is open to in­te­grate with the cloud to aug­ment re­source re­quire­ments when­ever needed, adds Kanafani.

Gart­ner pre­dicts that by 2025, 80% of en­ter­prises will shut down their tra­dional data cen­tres.

For larger en­ter­prises to stay in com­pli­ance, data res­i­dency reg­u­la­tions have made it manda­tory to re­tain data in en­ter­prise data cen­tres). But even larger en­ter­prises are mov­ing to the public cloud or Saas ap­pli­ca­tions, says Davey of Man­ageengine.

Hy­brid cloud isn’t nec­es­sar­ily about sep­a­rat­ing work­loads be­tween pri­vate cloud and public cloud. It works in a par­tic­u­lar ap­pli­ca­tion use case where you can take best ad­van­tage of each plat­form, says Sai­fud­din of Western Dig­i­tal.

The fu­ture data cen­tre is es­sen­tially soft­ware de­fined, bor­der­less, dy­namic and makes up a crit­i­cal part of the Data Fab­ric cir­cle of data man­age­ment, says Kanafani of Netapp.

Newspapers in English

Newspapers from UAE

© PressReader. All rights reserved.