Blogs & News
How the Aridhia DRE Addresses Historic Researcher Experiences with Trusted Research Environments
With the wider adoption of the now myriad varieties of trusted research environments (TREs), a recent blog from Richard Welpton (Head of Data Services Infrastructure, Economic and Social Research Council) looks at the experiences of surveyed researchers using these platforms. While the final results of the survey will be used to make recommendations for addressing current concerns as part of ESRC’s Future Data Services programme, his thorough blog details some early headline opinions on engaging with various TREs. Let’s see how the Aridhia DRE measures up in response.
Accessing Secure Data
In what may sound like a familiar story in the healthcare research domain, the time taken to access data in TREs for many of the respondents was reportedly over 2 years. Combined with difficulties in understanding data access request forms, poor communication as to request progress, and the constraints (in some cases) of being forced to use on-site options without the possibility of remote access.
With the work put into the DRE over the past several years on simplifying and streamlining approval workflows, data access is easier and more controlled than ever, from both the requester and data owner’s perspectives. The DRE enjoys the immediately obvious circumventing of one problem: being a 100% cloud-based service where there is no “on-prem” version of deployment to contend with.
The DRE provides a fully configurable, orchestrated Data Access Request mechanism. Allowing data owners to supply custom request forms, configure data usage conditions and automate the approval process – including individuals who are not necessarily members of the DRE. Each request and approval is audited and users are notified of progress. Upon approval, users can immediately access their workspace and work with the data. This process not only reduces the time taken, it also reduces administration and paperwork and provides a standard approach and framework for all data requests.
Safe Researcher Training
Training and other learning and support materials will always be an integral part of any TRE experience. How obligatory that training is will always be at the discretion of the platform owner. While few of our customers stringently enforce user training, Aridhia itself provides a series of training modules and a comprehensive series of learning articles and tutorials. These give users a broad introduction to the platform, schooling them in specific lessons and best practice methods. Aridhia’s support team closely liaises with the customer success team so that they are best placed to advise users on the best course of action to take regarding frequently asked questions.
Onboarding and Tooling
The number of researchers who struggle with setting up and then logging into a TRE session looks to be unacceptably high: though it is noted that those who do manage to log in then easily find both the software and data they plan to work with. Obviously, given the nature of the Aridhia DRE, setup of an instance of the environment is an issue that is entirely bypassed: leaving users to only concern themselves with logging in.
Once inside, workspaces provide plenty of initial tooling to get them started. Surveyed participants were most likely to use R and Stata and appreciated GUI functionality as well: with an abundance of support for R and RStudio built into each workspace, Stata optionally available within workspace VMs, and an extensive knowledge base site featuring instructions on bringing in preferred tooling, the design of the DRE is already aimed at solving these user concerns before they’re encountered.
Computing Power
Flexible compute options are a definite boon of an online TRE model. Many complaints were raised about insufficient computing power for research needs, with as many as a third reporting that this led to an inability to complete their desired research activities. The Aridhia DRE provides power workspaces with in-built tooling making use of dedicated compute resources. However, compute power can be increased on request by deploying virtual machines to the workspace. Each user of the workspace has access to the virtual machines and there is no limit to the number of machines that can be attached to a workspace. Linux and Windows machines are available, including Data Science Virtual Machines with over 50 in-built tools and resources used frequently by data scientists. The DRE can also make use of Azure serverless compute resources, such as Azure ML for machine learning.
Security vs Usability
A common question we get relates to how we’re able to strike a balance between meeting the security requirements expected when dealing with sensitive health data and ensuring that the locked down nature of the platform doesn’t hamper productivity. Other existing TREs are apparently still struggling with this, as several respondents list this as an issue, particularly in the area of code resolution. DRE users can create private Git repositories (via Gitea) that are local to their workspace for version control and other collaborative coding activities. Virtual machines come with automatic restrictions in place for external internet access. Allowed domains typically include sites used for statistics software and package management, but can be easily modified by administrators.
Import and Output Requests
While many respondents found problems with the import of code and other data, this tends to not affect DRE users. In many instances of the DRE, customers are beginning to request that uploads, as well as exports, are subject to administrative approval (which Aridhia is more than happy to facilitate). For others, import of data is at the discretion of the user, providing they follow audited guidelines of what they do and do not have permission to upload: in these examples, there is no need to wait for the approval of an administrative agent. The important thing is that organisations should always be presented with a choice in the matter.
While it is true that waiting for dataset approval or federated analysis credentials is a necessary reality that researchers and data scientists will have to live with, much of the DRE’s development of approval pipelines and the simplification of the request process has ensured that admin-related waiting times are kept to a minimum (or in some cases, instantaneous).
The Future
We thank ESRC and Richard for their work in helping identify ways to improve the TRE experience and look forward to seeing the results and finalised recommendations when they are published in the coming months.