VITASENIOR-MT: A distributed and scalable cloud-based telehealth solution

VITASENIOR-MT: A dis ibu ed and scalable cloud-based eleheal h solu ion Diogo Mendes VITA.IPT Lab Poly echnic Ins i u e o Toma Toma , Po ugal [email p o ec ed] Dá io Jo ge VITA.IPT Lab Poly echnic Ins i u e o Toma Toma , Po ugal [email p o ec ed] Rena o Panda VITA.IPT Lab Poly echnic Ins i u e o Toma Toma , Po ugal [email p o ec ed] Rica do An ónio VITA.IPT Lab Poly echnic Ins i u e o Toma Toma , Po ugal [email p o ec ed] Ped o Dias VITA.IPT Lab Poly echnic Ins i u e o Toma Toma , Po ugal [email p o ec ed] Luis Oli ei a VITA.IPT Lab Poly echnic Ins i u e o Toma Toma , Po ugal [email p o ec ed] Gab iel Pi es VITA.IPT Lab Poly echnic Ins i u e o Toma Toma , Po ugal [email p o ec ed] Abs ac — VITASENIOR-MT is a eleheal h pla o m ha allows o emo ely moni o biome ic and en i onmen al da a in a domes ic en i onmen , designed speci ically o he elde ly popula ion. This pape p oposes a highly scalable and e icien a chi ec u e o anspo , p ocess, s o e and isualize he da a collec ed by de ices o an In e ne o Things (IoT) scena io. The cloud in as uc u e ollows a mic ose ices a chi ec u e o p o ide compu a ional scalabili y, be e aul isola ion, easy in eg a ion and au oma ic deploymen . This solu ion is complemen ed wi h a p e-p ocessing and alida ion o he collec ed da a a he edge o he In e ne by using he Fog Compu ing concep , allowing a be e compu ing dis ibu ion. The p esen ed app oach p o ides pe sonal da a secu i y and a simpli ied way o collec and p esen he da a o he di e en ac o s, allowing a dynamic and in ui i e managemen o pa ien s and equipmen o ca egi e s. The p esen ed load es s p o ed ha his solu ion is mo e e icien han a monoli hic app oach, p omo ing be e access and con ol in he da a lowing om he e ogeneous equipmen . Keywo ds—Teleheal h, Cloud Compu ing, Fog Compu ing, IoT, Mic ose ices. I. INTRODUCTION In he Eu opean Union (EU), demog aphic ends poin o a ise in aging popula ion, wi h an es ima ed g ow h o 74% in he popula ion o e 65 yea s o age un il 2060, and a decline o 42 million in he wo king-age popula ion [1]. This ac , oge he wi h he high dese i ica ion o he coun yside caused by he u al exodus, leads o an inc eased numbe o isola ed elde ly people. Coupled wi h his eali y is he high numbe o age- ela ed diseases, such as ype 2 diabe es and ca dio- ce eb o ascula diseases, o immobili y condi ions due o ac u es caused by alls. As an example, in 2015 he e we e almos 49 million people diagnosed wi h ca dio ascula diseases in he EU [2]. Con inuous moni o ing and ollow-up o he elde ly can help o mi iga e such si ua ions. Teleheal h has become an impo an esea ch and de elopmen opic in he a ea o Cloud sys ems, p omo ing he use o mobile de ices in heal hca e se ices o mo e ulne able popula ions [3]. To add ess he abo emen ioned issues, we p opose a eleheal h pla o m, VITASENIOR-MT, ha allows o moni o he pa ien s’ clinical pa ame e s as well as he en i onmen whe e hey li e [4]. The elde ly in e ac wi h he sys em h ough a ele ision, a de ice ha is al eady amilia o hem, he eby a oiding he in oduc ion o new echnology which could lead o cons ain s and ejec ion o he sys em. A cen al de ice, ha we call VITABOX, suppo s he in e ac ion be ween he use and he ele ision. I is also esponsible o collec da a om he medical de ices and en i onmen al da a om he wi eless senso ne wo k, as well as o send all da a o he Cloud in as uc u e (see de ails in [4]). Cloud compu ing is a well-de ined and consolida ed concep o ge hos ing o ou eleheal h sys em wi h dis ibu ed p ocessing powe wi hou he need o know he ha dwa e physical loca ion. To be e o ganize he business logic and in o ma ion s o age, i is necessa y o model all da a in a homogeneous way, abs ac ing all de ices dissimila i y and impose access es ic ions and pe sonal da a p o ec ion wi h di e en access laye s. The en i e p ocess s a s in he VITABOX wi h he p e- p ocessing o he senso s’ collec ed da a, ollowed by ansmission o he cloud, whe e i is s o ed and p ocessed o la e no i y he pa ien and hei ca egi e s o possible issues. Da a in o ma ion is also p esen ed o he use s, namely, in o mal ca egi e s, doc o s o he pa ien s hemsel es. In his pape we desc ibe he implemen a ion o VITASENIOR-MT eleheal h solu ion, which agg ega es da a om dis inc equipmen , based on di e en p o ocols. In addi ion, i also p o ides a pla o m o emo ely moni o he en i onmen and biome ic s a e o he use s in a plug-and-play way, wi hou equi ing special echnological knowledge. Wi h ha aim, we eso o concep s such as og compu ing and mic ose ices o p o ide a secu e, scalable and lexible sys em whe e each Fig. 1. Gene al a chi ec u e o he VITASENIOR-MT p ojec . module can be scaled as equi ed o accommoda e mo e use s, equipmen o analy ical p ocessing needs, independen ly. II. STATE OF THE ART Cloud Compu ing p omo es he p o ision o new business models o he p o ide s, such as SaaS (So wa e as a Se ice) and PaaS (Pla o m as a Se ice) [3]. Unlike he use o he classic IaaS (In as uc u e as a Se ice) model, in which he p o ide is only esponsible o p o ide he physical p ocessing powe , s o age and ne wo k esou ces, lea ing o clien he con igu a ion and main enance esponsibili y. In SaaS i is possible o he clien o use an applica ion in a Cloud en i onmen wi hou equi ing any de elopmen (e.g., e-mail and calenda ools). In PaaS, he p o ide se s all he ecosys em equi ed o a gi en echnology (ope a ing sys em, amewo k and deploymen ools) and he clien is equi ed only o de elop o e he p o ided s uc u e [3]. The PaaS model gi e us he oppo uni y o se up he ini ial sys em equi emen s, wi h he possibili y o scaling hem as needed in he u u e. As a esul , he use is billed only on he esou ces needed, in a model known as pay- as-you-go (PAYG). Mo eo e , all asks ela ed wi h gua an eeing sys em s abili y, da a backups, upda es, load balancing and he isks o ha dwa e ailu es a e delega ed o he Cloud p o ide . This model also allows o some independence om he Cloud p o ide . A e all, we a e he code owne s and hus i is possible o quickly mig a e all he sys em o ano he p o ide wi hou comp omising cus ome se ice [5]. Despi e he ad an ages o Cloud Compu ing, he e a e s ill some limi a ions o scope, such as he In e ne o Things echnologies (IoT) and p i a e sys ems, o en causing secu i y holes in he access o he p i a e ne wo k. The majo issue in he in eg a ion be ween IoT and Cloud is he in e ope abili y be ween he senso s and he web se ices. Then, i is necessa y o use a se ice o ansla e he di e en ocabula ies p esen in he a ious communica ion p o ocols and echnologies. Tha p ocess is e e ed o as da a seman i ica ion, o ma ing he da a wi h speci ic p ope ies o ags o ge a uni ied a chi ec u e [7]. The e a e some p ojec s o add ess his in e ope abili y p oblem, bu hey p esen a solu ion limi ed o a speci ic echnology, such as he Wa son IoT Pla o m on he IBM Cloud [8] o he AWS IoT Co e om Amazon Web Se ices (AWS) [9], which a e limi ed o MQTT based de ices. To p omo e ull communica ion be ween di e en ypes o applica ions, a new concep has been in oduced, he Fog Compu ing [3]. In his a chi ec u e all he agg ega ion logic o he e ogeneous de ices om he IoT ne wo k is ans e ed o a laye usually loca ed a he edge o he In e ne . Taking ad an age o his p ocessing powe , i is possible o model he da a in an accep able o ma o he Cloud business logic, and o ensu e he il e ing o he access eques s by new de ices o he IoT ne wo k. Wi h his, we can ake ull ad an age o he IoT, enjoying he p oximi y o use s and he en i onmen ha su ounds hem, and o he Cloud, wi h he i ually almos unlimi ed p ocessing and s o age capabili ies. Some p ojec s simila o VITASENIOR-MT ha e been p oposed, implemen ing Cloud sys ems o emo ely moni o heal h pa ame e s. Fo example, in [6] blood p essu e and elec oca diog am de ices a e connec ed o a pla o m ha associa es use s wi h heal hca e p o essionals (e.g., doc o s and nu ses) and le hem check he esul s in an And oid applica ion. Howe e , i does no collec home en i onmen al da a as in ou eleheal h solu ion. The majo limi a ion o hei p ojec is ha he Cloud a chi ec u e was no designed o e ol e o e ime in new unc ionali ies like as de ices and use s’ managemen . This could be achie ed by implemen ing a mic ose ices a chi ec u e and de eloping a Backo ice whe e he adminis a o s could manage equipmen , being able o add new de ice ypes o al e he h eshold ules o abno mal alues, as we a e p oposing in VITASENIOR- MT. III. GENERAL ARCHITECTURE The VITASENIOR-MT p ojec is di ided in 3 pa s (Fig. 1):  Wi eless Senso Ne wo k (WSN) and Blue oo h medical de ices: p o ides he eadings o he home en i onmen and biome ic senso s;  VITABOX: p omo es pa ien in e ac ion wi h he TV, eques s da a om he senso s ne wo k, pe o ms p e-p ocessing o he da a and sends i o he Cloud;  Cloud: analyses he da a, p o ides a web in e ace o he use s, such as ca egi e s, heal hca e p o essionals, pa ien s and adminis a o s, and no i ies hem o any i egula i ies. The WSN is o ganized in a ee opology, p omo ed by Rou ing o e Low Powe and Lossy Ne wo ks (RPL), and he communica ion be ween he nodes is based on he IEEE 802.15.4 s anda d. The oo node is assigned as Bo de Rou e (BR), which connec s he WSN ne wo k and he VITABOX h ough a unslip6 b idge. I allows access o new nodes, which we e p e iously de ined om he Cloud in e ace, and enables he access o he esou ces om he ex e nal ne wo k, h ough he Cons ained Applica ion P o ocol (CoAP). This way he VITABOX is able o ob ain empe a u e, humidi y, ca bon monoxide and ca bon dioxide eadings, in JSON o ma , om each node. The VITABOX collec s da a om he biome ic senso s using Blue oo h Low Ene gy (BLE). These senso s p o ide blood p essu e, weigh , oxime y o blood glucose measu emen s a he pa ien 's ini ia i e. The e is Fig. 2. Objec ela ionships in VITASENIOR-MT sys em. Fig. 3. Cloud mic ose ices a chi ec u e. also a b acele o measu ing hea a e and s eps, which a e collec ed au oma ically by he VITABOX, when i senses he b acele signal wi hin i s ange. VITABOX also cen alizes use -TV in e ac ion ea u es, allowing he use o ini ia e ce ain exams, iew his/he his o ical da a and ecei e ale s in dange ous si ua ions. All eques ed da a, ei he om WSN o medical de ices, a e s o ed in a local da abase wi h a alidi y o 24 hou s o en i onmen al eco ds and 1 mon h o biome ics eco ds. Finally, he eco ds a e o wa ded o he Cloud, whe e p o ocols and condi ions a e applied. I abno mal alues a e de ec ed, he pe sons esponsible o he elde ly and he VITABOX i sel a e no i ied. The ela ionships be ween VITABOXes, senso s, pa ien s and he di e en oles o use s, such as he doc o s and he ca egi e s, a e also de ined in he Cloud (Fig. 2). The eby, i is possible o se di e en pe missions o use s o access da a, edi p o iles, schedule exams o manage equipmen . Fo example, only he doc o s associa ed wi h a speci ic pa ien can de ine his/he clinical p o ile and schedule exams. To his end, he ca egi e needs o indica e he espec i e doc o and associa e he biome ic de ice o he pa ien . These ea u es can be pe o med h ough a web in e ace a ailable o any de ice wi h a mode n b owse . A se o senso s’ h esholds and clinical p o iles a e de ined in he Cloud o gene a e wa nings o ecommenda ions, as well as o no i y he ca egi e s when some i egula i y occu s. To p omo e a as e eac ion om house dwelle s, hese h esholds a e sen o he WSN nodes h ough he VITABOX. This way, pa ien s may ac immedia ely, o example, opening a window i he CO is oo high. As his p ocess is done locally, his ensu es he au onomous ope a ion o he VITABOX e en when no connec ion can be es ablished wi h he Cloud in as uc u e. The same is no applied o biome ic alues because o be in e p e ed and alida ed, hese da a equi e mo e compu a ional powe no o e ed by VITABOX ha dwa e and need also he in ol emen o mo e ac o s o he o e all sys em. IV. IMPLEMENTATION A. Fog Compu ing One o he p oblems when designing an IoT/Cloud a chi ec u e is he ou ing o senso da a o he Cloud, namely he lack o in e ope abili y be ween di e en echnologies used o ga he da a. Due o his eason, i was necessa y o implemen an in e media e p oxy ha could agg ega e and connec all he equipmen and p ocess he da a, so ha he da a ollowed he same schema pa e n [10]. This ask is ca ied ou by he VITABOX, because i is loca ed a he edge o he local ne wo k, acili a ing he dis ibu ion o he sys em and he collec ion o da a om au ho ized de ices. All eco ds ecei ed in he VITABOX a e modeled in o a s anda d Ja aSc ip Objec No a ion (JSON) and o wa ded o he Cloud, hus abs ac ing he Cloud om he de ices he e ogenei y and a oiding he au hen ica ion o each equipmen . Since he sou ce o he da a is he VITABOX, ha is al eady au hen ica ed, i only accep s he equipmen ha was p e iously egis e ed in he Cloud. Ano he pa icula i y o ou implemen a ion is ha senso da a a e collec ed by he ini ia i e o he VITABOX, unless a senso de ec s a alue ou o he p ede ined ange. In his case, he senso igge s an e en o he VITABOX. This way i is possible o con igu e he pe iodici y o da a eques s o senso s. B. Scalabili y Cloud Compu ing is known by i s almos unlimi ed p ocessing and s o age capabili ies, howe e ha has an associa ed cos and hus i is necessa y o design he se ices a chi ec u e o ge he mos ou o he Cloud. The model adop ed was he PaaS p o ided by IBM Cloud [11] allowing ho izon al scaling o he sys em. Fi e di e en p ocessing nodes we e de eloped (see Fig. 3):  API node: Web Applica ion P og amming In e ace (API) ha p omo es he da a accessibili y, ia Rep esen a ional S a e T ans e (REST), p o iding he c ea e, ead, upda e and dele e ac ions (CRUD) o manage all da a gene a ed by he sys em;  WS node: es ablishes he WebSocke (WS) connec ions be ween VITABOXes and b owse clien s o send messages in eal- ime. These a e sen by he se e ini ia i e, used in asks such as an upda e o he VITABOX o a wa ning o he ca egi e s abou some i egula i y on he biome ic alues o a pa ien ;  Wo ke node: execu es in ensi e p ocessing ac ions such as applying il e s o da a, compa e senso s da a o p ede ined h esholds and clinical p o iles, gene a ing wa nings om abno mal alues, and upda ing VITABOXes and senso s imes amps;  Schedule node: uns all he unc ions ha execu e in loop, e.g., checking i scheduled exams we e pe o med;  Pee node: egis e s he clien s ha in end o connec h ough a Real Time Communica ion o e he Web (WebRTC). The communica ion be ween he mic ose ices is pe o med by he Rabbi MQ message queue sys em [12]. When he API node ecei es senso eco ds, hey a e o wa ded o he queue b oke . The Wo ke node is lis ening he b oke , and applies he app op ia e ules o he ecei ed da a, e i ying whe he he senso alues a e as expec ed. I a da a abno mali y occu s, an ale is egis e ed and sen o he clien s h ough he WS node. The adop ed a chi ec u e allows he sys em o scale acco ding o access ypes. I , o example, he numbe o VITABOXes inc eases d ama ically, mo e senso da a will be gene a ed, which o ces an inc easingly la ge numbe o Wo ke nodes o be ins an ia ed in o de o analyze hem. I , on he o he hand, an inc ease o accesses by he ca egi e s occu s, hen mo e API nodes can be alloca ed o no comp omise he da a access. I e en mo e use s a e associa ed o VITABOXes, hen mo e WS nodes a e equi ed o ensu e eal- ime wa nings. The eby, we a e su e ha he Schedule node will no need he balance because he loop ime will dynamically change as he numbe o scheduled ac ions g ows. No only he message b oke unc ions as a b idge be ween he mic ose ices, bu i also p o ides he publish/subsc ibe (Pub/Sub) used by he WebSocke s. This allows hem o scale ho izon ally wi h he WebSocke node, wi hou es ic ing access o ale s by he clien s. Tha is, ega dless o he WebSocke node ype o which he clien connec s, i will always ecei e he messages add essed o i sel . The message channels ha he clien s subsc ibe a e ela ed o all VITABOXes o which hey ha e pe mission o access he da a, gua an eeing he p i acy in he sys em. C. Laye ed s uc u e One majo conce n in Cloud a chi ec u e is o gua an ee he bes la ency o eques s. We planned and implemen ed se e al mechanisms o comply wi h ou sys em pe o mance equi emen s. The de elopmen pla o m adop ed was NodeJS [13]. Al hough Ja aSc ip (JS) is an in e p e ed language, NodeJS uns on Google's V8 engine ha compiles JS di ec ly in o machine code and op imizes pa s o he code in eal ime [14]. To achie e he mos ou o he compile , he code ollows a laye ed pa e n [15] enabling he euse o code, a be e o ganized s uc u e and op imized unc ions ins an ia ion in hidden classes [14], main aining hem o a minimum. This p omo es be e long- e m op imiza ion in he JIT (jus -in- ime) o he mos eques ed pieces o code (ho unc ions). Fou backend applica ions we e de eloped o di e en pu poses (Fig. 3). The de ined laye s we e:  HTTP Se e : ep esen s he s a ing se ice o lis en clien s’ eques s;  Middlewa e: i s code execu ed on e e y eques , ha de ines accep ed heade s, alida es he au hen ica ion okens and ex ac s he clien idiom;  Rou e : in e p e s he eques ed ou e and e sion and o wa ds he eques o he espec i e con olle ;  Con olle : il e s he access acco ding o he use 's pe missions, collec s he pa ame e s sen , and calls he unc ions o he business logic;  Business: de ines CRUD unc ions and da a analysis o business logic;  B oke : es ablishes communica ion wi h a message queue se e , and pe o ms he messages sen and ecei ed;  Model: de ines he da a models and es ablishes he connec ion o he da abases;  WebSocke se e : es ablishes he WebSocke s wi h he clien s;  B oke -WebSocke in e ace: es ablishes he connec ion be ween he WebSocke s and he messages queues;  Pee Se e : suppo s he pee disco e y by he clien s egis e . D. Da abase The Cloud in as uc u e is con inuously ecei ing senso s’ da a gene a ed by he a ious VITABOXes. Thus, a da abase se ice designed o high equency eads and w i es o la ge olumes o da a is equi ed. Fo his eason, i was decided o use a non- ela ional da abase, since hese ha e a linea sea ch capabili y and allow o scale ho izon ally. The non- ela ional da abase adop ed was he MongoDB [16], a documen -o ien ed da abase ha allows o deal wi h he cons an egis a ion o use logs, senso alues and wa nings. The d awbacks o his solu ion a e he lack o da a o ganiza ion, since NoSQL se ices a e schemaless, no equi ing a s uc u e o he sa ed da a, and he lack o ma u i y in ACID (A omici y, Consis ency, Isola ion, Du abili y) ansac ions. These cha ac e is ics may ep esen a law o he use s’ da a sa e y. Thus, he addic ion o a ela ional da abase, such as MySQL [17], is jus i ied o gua an ee he accoun ing p ocess and ela ionships o he VITABOXes wi h use s, senso s and pa ien p o iles. E. Access Fo he use o access he sys em, he API Web se ice p o ides he documen a ion iles and he web applica ion in e p e ed by he b owse s. The documen a ion is gene a ed h ough he apiDoc lib a ies [18], o ming a web page based on API sou ce code anno a ions, de ining in each REST ul ou e he me hod o be used, he pa ame e s and heade s o send and he possible esponses. To ee some p ocessing powe om he Cloud, ins ead o ende ing he pages on he se e (se e -side iew), he web applica ion is being de eloped wi h Vue.js [19], a p og essi e Ja aSc ip amewo k o on end Fig. 4. Load es wi hou sepa a ion o se ices. Fig. 5. Load es wi h senso s da a b oke ing app oach. de elopmen . This app oach allows o a be e p ojec s uc u ing, making he de elopmen o he on end independen o he de elopmen o web se ices in he Cloud. This way, all da a p ocessing o g aphics displaying and in e aces c ea ion a e ans e ed o he clien , eeding on JSON s uc u es p o ided by he web API. This inc eases Cloud pe o mance by a oiding he was e o compu a ional esou ces in ende ing web pages and dec eases ne wo k a ic, since JSON da a packe s a e smalle han ans e ing ull HTML iles. The choice o Vue.js acili a es bidi ec ional communica ion be ween componen s, as well as lis ening o e en s be ween componen s wi hou he need o di ec dependence. The on end also suppo s WebRTC, allowing ideocalls be ween clien s (pa ien s om VITABOXes and ca egi e s/doc o s om he web applica ion) o e a Pee - o-Pee (P2P) communica ion [20]. This solu ion educes he la ency ypically in oduced by an in e media y se e , since he clien s only need o egis e in a Pee Se e o coo dina e he connec ion be ween pee s and o cope Ne wo k Add ess T ansla o s (NAT) and i ewalls. Ano he posi i e ea u e o his a chi ec u e is he abili y o deploy in e na ionaliza ion lib a ies, local s o age and access o o he web se ices wi hou ha ing o o e load he Cloud, since hese unc ionali ies a e all suppo ed by he clien . F. Secu i y Acco ding o he Gene al Da a P o ec ion Regula ion (GDPR), all da a ha di ec ly o indi ec ly iden i y a use mus ollow a se o echnical and o ganiza ional measu es o gua an ee access le els secu i y [21]. The e o e, se e al mechanisms we e implemen ed, om communica ion o s o age, o a oid undue access o c i ical in o ma ion. All communica ion connec ions a e es ablished o e secu e channels (h ps, wss, ampqs and au hen ica ion in he MySQL and MongoDB d i e s), ensu ed by he Cloud p o ide . In he p oposed eleheal h solu ion, se e al use oles a e implemen ed, such as adminis a o , home sponso , heal hca e p o essional and he pa ien himsel . The adminis a o has access o all equipmen , e iews he equipmen s a e and loca ion, bu canno access da a ob ained om senso s o sa egua d he pa ien p i acy. The use esponsible o a VITABOX has access o all da a o ela ed de ices and can g an access o o he use s. The heal hca e p o essional has access only o he da a ela ed o he pa ien s ha we e associa ed o him. The VITABOX is seen by he Cloud as a clien . Use ole alida ion is ob ained by he JSON Web Token (JWT) au hen ica ion me hod, whe he i is a VITABOX o a web applica ion use . JWT p o ides he maximum po en ial o a s a eless se ice ha gua an ees use au hen ici y, in which he se e does no sa e clien s' sessions. This oken is also used o alida e WebSocke connec ions. As he s o age se ices a e hos ed in he Cloud, we mus assume "Hones bu Cu ious" beha iou o he espec i e p o ide [22]. Since he enc yp ion p ocess implies an ex a compu a ional cos , i is necessa y o unde s and wha da a mus eally be enc yp ed, o a oid unnecessa y CPU expenses. The eby, only use s' a ibu es as names, add esses, emails, con ac s and pho os ilenames we e enc yp ed using a symme ic key based mechanism, lea ing unenc yp ed he da a collec ed by he senso s. In ac , he da a collec ed by he senso s can also be conside ed as pe sonal da a, bu al hough unenc yp ed, hese da a canno be associa ed wi h i s owne . So, e en ual a acke s only know ha he da a exis , bu hey canno know hei owne ship. This app oach esul s in a signi ican esou ce sa ing, wi hou comp omising he use ’s p i acy. The da abases a e accessed by web se ices h ough an Objec -Rela ional Model (ORM) and an Objec -Documen Model (ODM), acili a ing he o ganiza ion o he business logic in he de elopmen p ocess and a oiding possible sou ces o a acks such as he case o SQL injec ion. V. RESULTS Du ing he sys em de elopmen , se e al app oaches we e conside ed o make he se ices a ailable. Ini ially he Wo ke and API unc ionali ies we e in eg a ed in o he same se ice, bu du ing load es s, a signi ican delay was no iceable in he endpoin o senso s da a collec ion. The es ool used was Vege a [23], wi h which we we e able o ealize ha he esponse ime was g ea e in p ocessing he da a in a single se ice, han sending hem o a message b oke and analyze hem in a sepa a e se ice, no blocking he esponse. Sending 50 senso alues pe eques wi h a equency o 50 eques s pe second esul ed in a se ice ou age, leading o a la ency o mo e han 30 seconds pe eques (Fig. 4), which ha med no only he ecep ion o da a bu also he access o o he endpoin s. A e es ing he b oke implemen a ion, wi h he same eques s pe second and he same amoun o da a pe eques , he API node showed a signi ican imp o emen , a ely exceeding he esponse ime o 60 ms Fig. 6. Example o an au oma ed es ou pu . (Fig. 5). In his app oach he API node jus alida es he sou ce o he da a (VITABOX) and o wa ds hem o he b oke . The esul s o he es s we e ob ained om es ing jus one API node and a Wo ke node, showing ha he mic ose ices app oach gi e us e iden imp o emen s. As p e iously s a ed, his pe o mance can e en be imp o ed wi h he launching o new se ice ins ances (nodes) as needed. In addi ion, au oma ed unc ional es s we e also implemen ed, h ough he Ja aSc ip es ing amewo k Mocha [24], o check i all he endpoin s we e pe o ming he desi ed asks, alida ing he ou pu s o i egula si ua ions and es ima ing he execu ion ime o each endpoin (Fig. 6), allowing he p og amme o eshape he code o ob ain be e pe o mance. VI. CONCLUSION AND FUTURE WORK A solu ion was c ea ed o emo ely moni o biome ic and en i onmen al da a o pa ien s a hei homes. These da a a e p ocessed and analyzed in he Cloud, o issue ale s and no i ica ions, as well as o p o ide his o ical da a o amily membe s and duly au ho ized heal h p o essionals. As showed, he adop ion o mic ose ices in he Cloud, ollowing he concep o Fog Compu ing in an IoT solu ion, p o ides a be e pe o mance and a sus ained g ow h o he sys em. In addi ion, his helps a oiding d as ic changes in he p ojec a chi ec u e in he u u e, when new unc ionali ies a e added, ocusing only in he de elopmen o he se ice ha is in ended o be changed/in oduced wi hou penalizing he ope a ion o he emaining. As his eleheal h solu ion is s ill an ongoing esea ch p ojec , he e is s ill wo k o be done. The main change ha may be expec ed is he eplacemen o he module o senso s’ da a analysis in he Wo ke se ice, by an a i icial in elligence (AI) module, since cu en ly he Wo ke is jus based on a se o ules, wi h adjus able h esholds, ha alida e he collec ed da a. Rega ding he AI module, ou goal is o ha e a classi ie based on his o ical biome ic and en i onmen al da a ha may p edic abno mal e en s in newly collec ed da a. We also in end o add a mic ose ice o suppo a eal- ime communica ion sys em o emo e con igu a ion o he VITABOXes by a echnical eam. ACKNOWLEDGMENT This wo k has been inancially suppo ed by he IC&DT p ojec VITASENIOR-MT CENTRO-01-0145- FEDER-023659 wi h FEDER unding h ough p og ams CENTRO2020 and FCT. REFERENCES [1] A. Chłoń-Domińczak, I. E. Ko owska, J. Ku kiewicz and A. Ab amowska-Kmon, "Popula ion ageing in Eu ope. Fac s, implica ions and policies," Eu opean Commission, Di ec o a e- Gene al o Resea ch and Inno a ion, 2014. [2] E. Wilkins e al., "Eu opean Ca dio ascula Disease S a is ics 2017," Eu opean Hea Ne wo k, B ussels, 2017. [3] A. Bo a, W. d. Dona o, V. Pe sico and A. Pescapé, "In eg a ion o Cloud compu ing and In e ne o Things: A su ey," Fu u e Gene a ion Compu e Sys ems, pp. 684-700, 2016. [4] G. Pi es e al., "VITASENIOR-MT: a eleheal h solu ion o he elde ly ocused on he in e ac ion wi h TV," in 20 h IEEE In . Con . on e-Heal h Ne wo king, Applica ion & Se ices, Os a a, 2018. [5] A. M.-H. Kuo, "Oppo uni ies and Challenges o Cloud Compu ing o Imp o e Heal h Ca e Se ices," J. o Med.In e ne Res., 2011. [6] S. Babu e al., "Sma eleme y ki o p oac i e heal h moni o ing in u al India: The jou ney so a and he oad ahead," in 20 h IEEE In . Con . on e-Heal h Ne wo king, Applica ions & Se ices, Os a a, 2018. [7] I. Villanue a-Mi anda, H. Naze an and R. Ma inek, "A Seman ic In e ope abili y App oach o He e ogeneous In e ne o Medical Things," in IEEE 20 h In . Con . on e-Heal h Ne wo king, Applica ions and Se ices, Os a a, 2018. [8] IBM Co p., "MQTT connec i i y o de ices," 2018. [Online]. A ailable: h ps://console.bluemix.ne /docs/se ices/IoT/de ices/mq .h ml. [Accessed Sep embe 2018]. [9] Amazon Web Se ices, Inc., "How AWS IoT Wo ks," 2018. [Online]. A ailable: h ps://docs.aws.amazon.com/io /la es /de elope guide/aws-io - how-i -wo ks.h ml. [Accessed Sep embe 2018]. [10] F. Bonomi, R. Mili o, J. Zhu and S. Addepalli, "Fog Compu ing and I s Role in he In e ne o Things," MCC '12 P oc. o he i s ed. o he MCC wo kshop on Mobile cloud compu ing, pp. 13-16, 2012. [11] IBM Co p., "IaaS, PaaS and SaaS – IBM Cloud se ice models," 2018. [Online]. A ailable: h ps://www.ibm.com/cloud/lea n/iaas- paas-saas. [Accessed Sep embe 2018]. [12] Pi o al So wa e, Inc., "Rabbi MQ is he mos widely deployed open sou ce message b oke .," 2007. [Online]. A ailable: h ps://www. abbi mq.com. [Accessed Sep embe ]. [13] Node.js Founda ion, "Abou Node.js," [Online]. A ailable: h ps://nodejs.o g/en/abou /. [Accessed Augus 2018]. [14] T. Lau ens, "How he V8 engine wo ks?," 2013. [Online]. A ailable: h p:// hibaul lau ens.gi hub.io/ja asc ip /2013/04/29/how- he- 8- engine-wo ks/. [15] A. K. Ba czynski, "Assessing he Impac o Using Design Pa e ns o En e p ise Applica ion A chi ec u e" MSc hesis, Uni . o Oslo, 2014 [16] MongoDB, Inc, "In oduc ion o MongoDB," 2008. [Online]. A ailable: h ps://docs.mongodb.com/manual/in oduc ion/. [Accessed Augus 2018]. [17] O acle Co po a ion, "Wha is MySQL?," 2018. [Online]. A ailable: h ps://de .mysql.com/doc/ e man/8.0/en/wha -is-mysql.h ml. [Accessed Augus 2018]. [18] P. Ro mann, "APIDOC Inline Documen a ion o REST ul web APIs," 2015. [Online]. A ailable: h p://apidocjs.com/. [Accessed Sep embe 2018]. [19] Vue.js Team, "Wha is Vue.js?," 2018. [Online]. A ailable: h ps:// uejs.o g/ 2/guide/. [Accessed Sep embe 2018]. [20] S. Du on, "WebRTC in he eal wo ld: STUN, TURN and signaling,". [Online]. A ailable: h ps://www.h ml5 ocks.com/en/ u o ials/web c/in as uc u e/. [Accessed No embe 2018]. [21] Eu opean Union, "Gene al Da a P o ec ion Regula ion," [Online]. A ailable: h ps://gdp -in o.eu/. [Accessed Sep 2018]. [22] S. Yu, C. Wang, K. Ren and W. Lou, "Achie ing Secu e, Scalable, and Fine-g ained Da a Access Con ol in Cloud Compu ing," in 2010 P oc. IEEE INFOCOM, 2010. [23] T. Sena , "Vege a," 2018. [Online]. A ailable: h ps://gi hub.com/ sena / ege a. [Accessed Oc obe 2018]. [24] Mocha Backe s, "Mocha simple, lexible, un," 2018. [Online]. A ailable: h ps://mochajs.o g/. [Accessed Oc obe 2018].