SUC­CESS STORY

Au­tomat­ing to cre­ate a Trans­par­ent Cul­ture

The HR Digest - - Content Features -

In­side Face­book Inc.’s op­er­a­tions: Au­tomat­ing to cre­ate a trans­par­ent cul­ture

When a com­pany as large as Face­book sketches a blue­print for the fu­ture of op­er­a­tions, it mainly thinks about how its own op­er­a­tions will change. Rarely does the vi­sion of au­tomat­ing ex­clude the work that hu­mans do to­day.

Run­ning the op­er­a­tions of Face­book Inc. is not an easy feat. First of all, there’s the sheer scale of its global net­works and the supreme need it puts on re­li­able ser­vice and sat­is­fy­ing user ex­pe­ri­ences. Then there’s the need to cre­ate new flex­i­bil­ity and ca­pac­ity for the busi­ness to pur­sue its broader am­bi­tions. This in­cludes Face­book’s most re­cent ini­tia­tives, the Con­nec­tiv­ity Lab, AI, deep learn­ing and vir­tual re­al­ity as a nex­tGen com­put­ing plat­form.

To com­bat this, Face­book is us­ing au­toma­tion be­cause the team has a ton of com­plex­ity to man­age. They have an in­fra­struc­ture that sup­ports hun­dreds and thou­sands of com­put­ers all around the world. Face­book is ac­tu­ally serv­ing 2 bil­lion peo­ple in the map, and mil­lions of peo­ple on the other apps. Their engi­neers are con­stantly im­pro­vis­ing on the fea­tures, mak­ing things more op­ti­mized, and pro­vid­ing newer and bet­ter ser­vices. So, in re­al­ity, Face­book has a lot on its plate, and a lot that can be done via au­toma­tion, in or­der to keep up with the scale and the pace of its prod­uct de­vel­op­ment.

Re­cently, Face­book built a sys­tem called FBAR, for Face­book Auto Re­me­di­a­tion, to do the ba­sic hard­ware re­me­di­a­tion tasks. Be­fore, if a server had a hard drive fail­ure or some kind of hard­ware er­ror, an alarm would go off and a Face­book worker would have to log in, or walk to the com­puter, and try to fix it. The

worker would have to fix the soft­ware and re­boot the ma­chine, be­fore it starts work­ing al­right. To­day, all of the soft­ware re­me­di­a­tion and de­bug­ging is au­to­mated. No hu­man is in­volved in the process. The au­to­mated sys­tem can de­tect the er­ror, be it a disk drive, or a CPU, or a net­work­ing card fail­ure. The sys­tem will rec­og­nize the er­ror – and sim­ply fix it.

In any case, these are truly idi­otic things – things that are out­ra­geously unim­por­tant to au­to­mate. But, it al­lows Face­book engi­neers to work on higher-level things.

Face­book has a long list of me­nial tasks it’s look­ing to au­to­mate in the months to come. Take for ex­am­ple, clus­ters, .i.e. a bunch of servers do­ing a cer­tain type of func­tion in the in­fra­struc­ture. They re­quire a lot of con­fig­u­ra­tion – in­stalling soft­ware and a lot more, to make sure the right things are con­nected to one an­other. Back in 2009, this process was done man­u­ally. Face­book had to lit­er­ally write things down on the white­board to as­sign clus­ter job to a cer­tain group of engi­neers. The en­tire process was time con­sum­ing, and more im­por­tantly, er­ror-prone. You don’t need to worry about that with au­toma­tion, since work con­sis­tently gets done the same way, every sin­gle time.

In sim­ple words, au­toma­tion al­lows Face­book engi­neers to fo­cus on sim­ple, but time-re­ward­ing tasks. It al­lows them to think about the fu­ture projects, rather than what they’ve al­ready built. An­other rea­son why au­toma­tion is im­por­tant is that it al­lows teams to do things for the fu­ture. Imag­ine, a ma­jor­ity of tech com­pa­nies in the world in­vest most of their time bring­ing the best in tech. If em­ploy­ees end up do­ing hum­drum tasks for a long time, they aren’t re­ally learn­ing any­thing. This leads to job burnout and dis­sat­is­fac­tion. It is nec­es­sary to keep au­tomat­ing your sys­tems so that tasks don’t be­come

me­nial, bor­ing, or repet­i­tive for the em­ploy­ees.

Com­pa­nies must set em­ploy­ees in a place where one group, which needs to keep on im­prov­ing, and one needs to make ev­ery­thing cost-ef­fec­tive, are both aligned. Let’s take a look at Face­book’s prod­uct teams – it con­sists of the mid­dle­ware as well as back­end en­gi­neer­ing teams, the op­er­a­tions teams, the se­cu­rity team and the IT team. Back in 2009, it was con­sid­er­ably nor­mal to have so many teams. But then, Face­book re­al­ized it was caus­ing in­ef­fi­cien­cies in its op­er­a­tions. As a mat­ter of fact, it was slowing peo­ple down. The teams couldn’t make the best de­ci­sions. Some of the de­ci­sions made were short-term cost­based, while some were long-term. Au­toma­tion didn’t ex­ist in a lot of op­er­a­tions. Face­book, then, had to re­think its en­tire op­er­a­tional struc­ture, the kind of peo­ple it wanted to hire, and break every pos­si­ble wall. It is dif­fi­cult to man­age sep­a­rate teams. There are daily in­ter­rup­tions coming in con­sis­tently that dis­tract you from your long-term goal. You truly need to en­sure you have the right team, with the nec­es­sary ex­per­tise and the right skills you re­quire. In any case, there are one too many ad­van­tages of au­tomat­ing. To be­gin with, it makes work much more trans­par­ent. No one likes it when there is a team work­ing in a se­cured lo­ca­tion do­ing some­thing cool that might re­place your prod­uct and your team, some­day. It ends all sorts of un­wanted wor­ries. The core busi­ness team can go back to fo­cus­ing on more im­por­tant is­sues.

Newspapers in English

Newspapers from USA

© PressReader. All rights reserved.