Is fog computing the future of databases?

Send­ing an e‑mail, watch­ing a video on YouTube, organ­is­ing a video con­fer­ence meet­ing, or play­ing an online game… Each of these activ­i­ties requires sig­nif­i­cant data flows to and from servers, locat­ed in data cen­tres. The Cloud is the pre­ferred remote pro­cess­ing and stor­age sys­tem for devel­op­ing all the Inter­net appli­ca­tions we use every day. But oth­er decen­tralised com­put­ing infra­struc­tures exist. Fog com­put­ing is grow­ing in pop­u­lar­i­ty. Accord­ing to spe­cial­ist con­sul­tant Future Mar­ket Insights, the glob­al fog com­put­ing mar­ket is expect­ed to reach $2.2 bil­lion by 2032, up from $196.6 mil­lion in 2022.

What is fog com­put­ing? “It is a high­ly decen­tralised cloud, with small, geo­graph­i­cal­ly dis­persed com­put­ing units, clos­er to the data sources, and there­fore clos­er to the users,” says Guil­laume Pierre, a pro­fes­sor of com­put­er sci­ence at the Uni­ver­si­ty of Rennes, who is cur­rent­ly work­ing on this tech­nol­o­gy. Indeed, users’ data trav­els back and forth to data cen­tres that are gen­er­al­ly very far away, which may be locat­ed in anoth­er coun­try or con­ti­nent, and which con­sume a lot of ener­gy. Fog com­put­ing makes it pos­si­ble to short­en the routes of these flows. Guil­laume Pierre is work­ing on the use of small machines, the size of a cred­it card, often used to teach com­put­ing, the Rasp­ber­ry Pi. 

This infra­struc­ture there­fore responds to cer­tain lim­i­ta­tions of the Cloud. How­ev­er, Fog com­put­ing will not replace the Cloud, warns the com­put­er sci­ence pro­fes­sor: “Fog com­put­ing is rather the exten­sion of the Cloud into new ter­ri­to­ries, new types of needs.” Its main inter­est: the speed and effi­cien­cy of data trans­mis­sion. Fog com­put­ing can there­fore be par­tic­u­lar­ly use­ful when the appli­ca­tion used requires the low­est pos­si­ble response time, such as aug­ment­ed real­i­ty or video games.

Fog com­put­ing is the exten­sion of the cloud into new ter­ri­to­ries, new types of needs.

Copié !

“When we move towards demand­ing usage sce­nar­ios, the response time can be sig­nif­i­cant enough that inter­ac­tiv­i­ty is com­pro­mised and the appli­ca­tion func­tions poor­ly or not at all,” says Guil­laume Pierre. For vir­tu­al or aug­ment­ed real­i­ty, for exam­ple, experts say that if the time between a move­ment and the dis­play update exceeds a delay of 20 mil­lisec­onds, the user may suf­fer from sea­sick­ness because the objects appear to be unsta­ble. “Fog com­put­ing can be a solu­tion to reduce the laten­cy between the user and the game,” says the professor.

The oth­er advan­tage of fog com­put­ing is the devel­op­ment of the Inter­net of Things. Con­nect­ed objects such as smart­phones, tablets, cars or smart TVs are now ubiq­ui­tous. This is also the case in many areas such as indus­try, agri­cul­ture, sci­en­tif­ic research, urban plan­ning and secu­ri­ty. Con­nect­ed objects, such as a tem­per­a­ture sen­sor, a cam­era or an ener­gy meter, pro­duce data at cer­tain points. “When we work with sci­en­tists who observe the flood­ing of a riv­er, for exam­ple, it can be inter­est­ing to process the data on the spot so as to be able to pro­gramme reac­tions such as chang­ing the fre­quen­cy of mea­sure­ments, apply­ing a par­tic­u­lar type of treat­ment, etc.”, explains Guilaume Pierre.

Is fog com­put­ing a solu­tion to the envi­ron­men­tal chal­lenges posed by data cen­tres? Accord­ing to Guil­laume Pierre, the answer is unclear. Indeed, these cen­tres con­sume a lot of ener­gy, but they are also well opti­mised. Fog tech­nol­o­gy is still being devel­oped. “If we do things bad­ly, it is pos­si­ble to con­sume more. On the oth­er hand, we may have access to more renew­able ener­gy to pow­er the small machines. We are think­ing about ways to pow­er tools with renew­able sources, per­haps solar pan­els, which would reduce the eco­log­i­cal impact con­sid­er­ably,” says the specialist. 

Our mas­sive use of dig­i­tal tech­nol­o­gy could also lead to a stor­age cri­sis. Accord­ing to researchers at Aston Uni­ver­si­ty in Eng­land, the cloud will reach sat­u­ra­tion point, with a 300% increase in the amount of data in the world in the next three years (study pub­lished in Decem­ber 2022). “Organ­is­ing stor­age sys­tems based on small­er units may be part of the solu­tion, espe­cial­ly if the data is already dis­persed at the out­set, such as with the Inter­net of Things,” says Guil­laume Pierre.

Beyond the com­par­i­son with the Cloud, research into Fog com­put­ing has already led to con­crete advances. Guil­laume Pierre coor­di­nat­ed the Euro­pean Fog­Gu­ru doc­tor­al train­ing project. Eight doc­tor­al stu­dents were able to work with the city of Valen­cia, in Spain, on water con­sump­tion, a major issue in this semi-desert area. The city has been deploy­ing smart meters, sim­i­lar to Enedis’ Linky device, for the past 15 years: the team of researchers devel­oped the appli­ca­tion that process­es the data in order to inter­vene more quick­ly in the event of a water leak affect­ing con­sumers. Pre­vi­ous­ly, the response time was between three and six days. With Fog com­put­ing, data is trans­mit­ted more fre­quent­ly and effi­cient­ly, and the response time is reduced to only a few hours, thus avoid­ing the waste of this pre­cious resource.

Fog com­put­ing is also already being deployed in indus­try. Tele­phone oper­a­tors are also very inter­est­ed in devel­op­ing this tech­nol­o­gy, as are muse­ums and tourist offices, which see it as a way of offer­ing smooth and fast ani­ma­tions and sim­u­la­tions. The uses are var­ied: sci­en­tif­ic data pro­cess­ing, video games, indus­tri­al process­es, restau­rants that want to opti­mise their fre­quen­ta­tion, med­ical analy­ses, vis­its to a city or an exhi­bi­tion in aug­ment­ed real­i­ty, etc.

So what is miss­ing for the large-scale devel­op­ment of this tech­nol­o­gy? “What we don’t have today are gen­er­al sys­tems, where any­one can deploy their appli­ca­tion in the Fog, and where sev­er­al uses can coex­ist. This mas­si­fi­ca­tion will take a lit­tle time, because we lack appli­ca­tion deploy­ment and plat­form man­age­ment tech­nolo­gies, which we are cur­rent­ly work­ing on. With­in ten years or so, we will have this tech­nol­o­gy deployed and avail­able, per­haps not for indi­vid­u­als, but cer­tain­ly for com­pa­nies,” says Guil­laume Pierre.

Sirine Azouaoui