CSE 5095 Cloud Computing and Software as a Service in BMI Timoteus Ziminski Computer Science & Engineering Department The University of Connecticut tbz@engr.uconn.edu Spring 2011 CC1 Overview CSE 5095 Background Definition Providers Technologies BMI in the Cloud Application Fields Example: Virtual Patient Chart Limitations Conclusion Critique Research topics CC2 Overview CSE 5095 Background Definition Providers Technologies CC3 Background CSE 5095 What is the Cloud? CC4 Definition CSE 5095 NIST definition: “Cloud computing is a model for enabling ubiquitous, convenient, on-demand network access to a shared pool of configurable computing resources (e.g., networks, servers, storage, applications, and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction.” Bottom line: computing paradigm that hides complexity through a notion of abstract resources Although this paradigm is still considered to be under development, several very mature platforms are apparent. CC5 Providers CSE 5095 Google Mail, Calendar, Health, Maps, Docs, Google Apps Amazon Amazon Web Services (AWS), Simple Storage System (Amazon S3) IBM SmartCloud Services, LotusLive, Blueworks, Collaborative Care Microsoft Azure, WindowsServer Hyper-V, Office 365 And many, many more… Enormous range of products covering a variety of tasks: development, testing, training, archiving, analysis, and collaboration CC6 Service Models CSE 5095 Heterogeneous types of service can be embedded into the cloud All options hide underlying infrastructure to a certain level Software as a Service (SaaS) Provides applications Cross-device and cross-platform capabilities Platform as a Service (PaaS) Access to the cloud hosting environment Deployment of applications Infrastructure as a Service (IaaS) Provides access to basic resources CPU, network, storage CC7 Characteristics CSE 5095 On-demand self service Services are mostly automatic Broad network access Usage of standard mechanisms Allows thin clients Resource pooling No control about physical instances Rapid elasticity Resources are presented as unlimited Capacities grow and shrink automatically according to requirements Measured service Pay for what you use Little up-front capital needed CC8 Deployment Models for Cloud Infrastructure CSE 5095 Private Access by one organization Management in-house or by third party Community Shared across multiple organizations Public Accessible by general public or large number of organizations Typically used for selling cloud applications Hybrid Composite of multiple clouds CC9 Importance to IT Market CSE 5095 Citigroup estimates Amazon will generate $650 million revenues through AWS in 2011. IBM intends to generate revenues of $7.0 billion from cloud computing by 2015 (over 5% of total revenues.) Microsoft will be spending 90% of its $9.6 billion research and development budget on cloud strategies in 2011. Gartner predicts the cloud market to grow to $102.1 billion net this year from $68.3 billion in 2010. CC10 Technologies: On the Hard Side CSE 5095 Cloud computing requires reliable, efficient and powerful data centers General hardware progress gives advantages Disc space, memory, energy efficiency More important, networking and virtualization technologies are fueling the development of cloud applications Omnipresent access to low latency broadband internet Multiplexing architectures Hardware virtualization support Clustering techniques CC11 Technologies: On the Hard Side (ctd.) CSE 5095 Lightweight, always connected end devices are increasingly popular Tablet PCs, Smartphones, Laptops (Cr-48) Usage of multiple platforms simultaneously creates synchronization issues CC12 Technologies: On the Soft Side CSE 5095 On the server side virtualization software plays an important role VMware, VirtualBox, Virtual PC, Xen From an organizational perspective, Service-oriented Architectures allow the decomposition of IT processes Encapsulation into service with interfaces Create potential for replacement of local services with options from the cloud Possible implementations REST: HTTP/S or SOAP RPC Web Services: XML, WSDL, UDDI, SOAP CORBA CC13 Technologies: On the Soft Side (ctd.) CSE 5095 Development of operating systems which heavily rely on cloud applications Android, iOS, Chromium OS Increasingly sophisticated web browsers IE9, Firefox 4, Chrome, Safari 5 HTML5, JavaScript/AJAX, Flash, Silverlight To some extent also programming language concepts and libraries Java, Python, Perl, Ruby, Scala CC14 Overview CSE 5095 PaaS SaaS IaaS Use Cloud Resources CC15 Overview CSE 5095 BMI in the Cloud Application Fields Example: Virtual Patient Chart Limitations CC16 Application Fields CSE 5095 In which BMI areas can we leverage the benefits of Cloud Computing? Clinical Informatics Imaging Informatics Bioinformatics BMI Clinical Research Informatics Public Health Informatics HIT Training CC17 Clinical Informatics Challenges CSE 5095 Patient records Electronic Medical Records, Personal Healthcare Records Inherit need for efficient and reliable large-scale data storage and exchange Collaboration between Healthcare professionals Team-based collaboration Transition of care Healthcare organizations Practices, Labs, Pharmacies, Clinics, Hospitals Domains Providers, Payers, Researchers CC18 CC Benefits for Clinical Informatics CSE 5095 Personal Healthcare Records Google Health, Microsoft Health Vault, WebMD Electronic Medical Records First attempts to organize EMRs in private clouds Modular approaches with plug-in architectures Health information exchange Decision support Automatic note processing, voice recognition Patient education Opens the door for stepwise integration via a cloud of clouds Hybrid clouds regulate the amount of exchanged data CC19 CC Benefits for Clinical Informatics CSE 5095 Implementing HIE in the context of small practices Application service provider model Limits requirements and costs Technical knowledge can remain on provider side Maintenance reduces to thin clients Compliance with regulations becomes provider task Data warehousing Example: Data from the pathology laboratory In-house solutions create further data silos The cloud concept opens the doors for affordable generic solutions CC20 Virtual Patient Chart CSE 5095 CC21 CC benefits for Clinical Informatics (ctd.) CSE 5095 Virtual chart Cloud computing services can provide very abstract storage resources Space and performance is presented as limitless and automatically adjustable Highly accessible and easy to synchronize across platforms We are starting to have experience real-life proof platforms for real-time collaboration on documents Is that enough? CC22 A Word of Warning… CSE 5095 Cloud Computing is another useful tool, not the silver bullet! CC23 Bioinformatics Challenges CSE 5095 Heavy computational analysis tasks DNA sequencing Genomics Protein microarrays Mutation analysis In addition to the specific hardware for the analysis, we need Storage Petabyte scale CPU cycles CC24 CC Benefits for Bioinformatics CSE 5095 The cloud starts to offer elegant solutions for largescale data storage Know-how about hardware and data management can cumulate at service providers Elastic nature and “pay as you go” payment models increase accessibility for institutions Do research rather than worrying about backups Algorithmic tasks are the more at home in the domain of Grid computing However, the cloud can increase discoverability and streamline the access to Grid resources CC25 Medical Imaging Informatics CSE 5095 Essentially very similar arguments as discussed for the field of Bioinformatics CC26 Public Health Informatics Challenges CSE 5095 Tracking trends Large scale data warehousing Regional State-wide Population wide Disease Control and Prevention Emergency response CC27 CC Benefits for Public Health Informatics CSE 5095 Collection of public health data Virtual chart would a valuable source Inter-cloud communication The nature of the cloud facilitates: Storage of public health data Analysis of public health data Outcomes can be provided through cloud services CC28 Clinical Research Informatics Challenges CSE 5095 Repositories and Databases Data acquisition De-identification Data sharing Cohort building Support of clinical trials Data collection Data management Data analysis CC29 CC Benefits for Clinical Research Informatics CSE 5095 Outsource responsibility for regulation conform deidentification Data management services Break open data silos Reuse collected data Template based solutions for trial management Promotion of best practices Supporting tools Exploit mobile devices and adequate apps for data collection during trials Mobile phones, tablets CC30 HIT Training Challenges CSE 5095 Training is an important and often underestimated component Who requires training? Healthcare professionals Students of related fields In what areas can we train Medical education Transfer of new knowledge Skills in using software components EHR, Practice Management, e-Prescribing Adherence to best practices CC31 CC Benefits for HIT Training CSE 5095 Virtualization has proven to be a powerful tool to provide training environments and programs Training in a Cloud is a popular product Scalable Flexible Cost efficient Deployment and management of training environments is largely simplified by the “disposable” nature of virtual platforms Service providers can offer cost efficient programs and cover a larger variety of topics CC32 Limitations CSE 5095 Cloud computing and SaaS has some risks and limitations that we need to recognize in the context of BMI Trust and Security Reliability and Availibility Performance Standardization Overhead CC33 Trust and Security CSE 5095 Loading data off sensitive outside of organizational boundaries is by definition a security risk that has to be weighted against potential benefits Third parties are gaining access Service providers, governmental institutions Limitations through policies and regulations have to be taken into consideration Cloud computing does not make any assumptions about physical locations Data can travel outside of country boarders Employ suitable secure data transfer and encryption techniques Currently still performance problems CC34 Reliability and Availability CSE 5095 BMI systems are frequently of critical importance Clinical care systems can directly influence decisions that can have potentially crucial impact on the well being of patients Risk system outages and performance problems Recent outage of AWS Not an acceptable scenario for many BMI settings Absorbing the impact Local failover Sacrifices advantages of low complexity CC35 Performance CSE 5095 SaaS solutions have matured over the past years Sophisticated Frontends Low latency broadband internet However, the healthcare domain tends to be very unforgiving towards workflow disruptions Potential fluctuations in QoS can have harsh impact on the acceptance of services CC36 Standardization CSE 5095 Cloud Computing does not give any immediate advantages regarding interoperability Data models, standards and ontologies are a completely separate issue. Even worse, there are binding standards for cloud interfaces Potential vender lock Over 10 national and international organizations are currently working on establishing standards CC37 Overhead CSE 5095 Technologies such as Web Service and SOAP RPC which are used in Cloud Computing might be still too heavy weight for building a system that is aiming to span over the complete healthcare system From this ultra large scale system perspective we might need to rely on even more abstract concepts such as HTTP/S REST. CC38 Overview CSE 5095 Conclusion Critique Research topics CC39 Critique CSE 5095 Just another hype? A lot of advertisement Similar buzz-word overload as with SOA In fact, often no solid distinguishing of the patterns Nothing new? Similar middleware concepts and ideas go far back Existent services get tagged with the Cloud label But noteworthy Aligns with the current development of clients Conceptually targets many BMI problems Hardware and network progress enabled well convincing platforms and application suites CC40 Research Topics CSE 5095 Improving the Cloud Trust and Security Scheduling and Virtualization Mobile Cloud Computing Inter-cloud communication (cloud of clouds) Leveraging the Cloud Data Analysis Data Management Data Modeling Text and Voice recognition Conferences IEEE CLOUD, Cloud Expo, Cloud Connect CC41 CSE 5095 Thank you! Questions? CC42