The Effect of Continuous Encryption of Data in Cloud Native Architecture

Enterprise cloud-native software systems, handle a diverse set of customer data that can be of variable volume with variable throughput. With increasing emphasis on privacy, locality and security ([1]) of user data; data encryption (at-rest and in-transit) becomes an essential part of enterprise applications. Data encryption uses system resources for converting plain-text to cipher-text
^1
1
Cipher-text is encoded data that has been encrypted. The process of encoding plain-text data into cipher-text is called encryption. and this utilization varies with the strength of the cipher suite. In a low latency software system serving real-time user traffic with a non-deterministic workload, data security can become a bottleneck when there is a surge in both throughput and volume of encryption-eligible transactions. This can result in a 99
^th
percentile (P99) of CPU and network IO resource utilization at 94 % and a 95
^th
percentile (P95) data encryption latency of 700ms per request. The bottleneck can evolve into a single point of failure if this traffic is sustained for a longer duration and overwhelms the system resources. Thus, maintaining peak availability and responsiveness of the software system at all times becomes a complex task with data security. In addition, migration of software systems to a cloud computing environment brings in another dimension of managing cost- to-serve during surge periods
^2
2
Surge period refers to a sudden short term spike in workload or traffic to the system, In this paper, we study the impact of encryption and data security on the availability and cost- to-serve of a cloud-native enterprise application and discuss a novel approach to spread out the surge in resource utilization for peak availability and responsiveness with an improved P99 of 45 % in system utilization and P95 data encryption latency of 20ms per request.