Managing Schema Migration in NoSQL Databases: Advisor Heuristics vs. Self-adaptive Schema Migration Strategies

Managing Schema Migration in NoSQL Databases: Advisor Heuristics vs. Self-adaptive Schema Migration Strategies

연구 분야: Databases

학회: International Conference on Model-Driven Engineering and Software Development, International Conference on Model-Driven Engineering and Software Development

Schema-flexible NoSQL databases are increasingly popular backends in the agile application development as they allow developers to write code assuming a new database schema that is different from the current one. If the application is in production already, non-functional requirements for application performance and cost efficiency are routinely part of service-level agreements (SLAs). Co-evolving the schema with the application code then requires subtle management decisions regarding the migration of variational legacy data that is persisted in the production database. Eventually, project managers have to deal with the repercussions of schema evolution in order to comply with SLAs, especially if stipulated metrics compete with each other in tradeoffs. To this end, we present a NoSQL Schema Migration Advisor that supports the schema migration management in NoSQL databases in two distinct ways: If the migration situation can be elicited, a heuristic is offered to estimate the impact of schema evolution by means of choosing a migration strategy and pace code releases accordingly. If this information is not sufficiently or not readily available, self-adaptive schema migration strategies are presented that can automatically curate variational data such that competing metrics can be balanced out in order to comply with SLAs, if possible, making management interventions superfluous.