SynapDX is involved in researching about ASD (Autism Spectrum Disorder). They are devising a blood test which can lead to early detection of this disease since it creates disorders in children which hinder their development. So if this test is detected early, doctors will be able to treat the disease in its early stages, causing many people to heal from the disease.

As this test is being developed on computers, the data it creates is stored on the computer and its analysis is considered rigorous. It involves deep machine learning and analysis of different biological data. SynapDx started using AWS since 2012 for carrying out tests which are validated and tested in the library. These tests are according to the industry and health standards, and this is the reason SynapDX considers them one of their own kind in clinical testing.

The company started research in a new area of ribonucleic acid (RNA) processing products, and this in turn led to more computing power and operational costs. The purpose was to manage the computing power along with minimizing the operational costs. They had two applications: StarCluster for clustering the work load, and Grid Engine which is related to batch queue processing systems. These applications were deployed on AWS EC2. Each piece of data was divided by sending this to Grid Engine as a new task and then processing was done on StarCluster. This approach was not simple to run, so SynapDx was looking for another solution.

SynapDX started using some more products like Amazon SQS for queueing purposes, and batch processing was done more accurately than before. Data was stored in Amazon DynamoDB, and EC2 spot instances were used for cost management. The auto scaling option was also used with Amazon EC2 instances to provide more availability. The new solution ran with approximately no errors and the cost was reduced to more than half which was a major milestone.

The environment was working on both modelling the blood samples and then diagnosing the treatment on the basis of the blood samples available after analysis. The instances were made computer optimized as well as memory optimized, and the instances had auto scaling and spot instances capabilities in order to save cost. In addition to these capabilities, all the data was stored on S3 and there were large integration servers used for enhancing the efficiency of the system. Virtual private cloud was used to manage all the networking and to make the solution more scalable and secure. Route53 was used to manage the routing between the cluster’s DNS servers. EIP and VPN connections were used to make the solution very isolated from network attacks. Subnetting and access control lists enabled the privileged users and systems to access their managed resources.

“By using Amazon SQS and DynamoDB to implement resilient workflow systems, we’re able to process days’ worth of jobs and 5 to 10 TB of data across hundreds of instances with a near zero failure rate,” says Abrams. “AWS CloudFormation is the backbone of our launch infrastructure – backed by dynamic configuration code that has proved to be a remarkably flexible and efficient tool for reproducible launch results.” All this leads to the fact that deployment was a great success and it achieved its targeted purpose. They were able to launch their instances in a secure environment with encryption and privacy standards. “The company saves time by designing software with simple interfaces for use with AWS services instead of building its own services or using complex, off-the-shelf software packages,” says Abrams. “Besides being a time-saver, the flexibility here that lets us pivot during our clinical trial was very advantageous to us.”