MemSQL 201: Tips

In a recent webcast , we shared tips and tricks for understanding MemSQL, best practices for implementation, and demoed MemSQL with a real-world use case.

When moving an application to, or creating one on MemSQL, start by thinking about whether rowstore or columnstore storage (or both) is ideal

If you have high throughput requirements, consider using MemSQL Pipelines

Use EXPLAIN and PROFILE operators to identify query bottlenecks

Take advantage of reference tables to eliminate distributed joins

Using and monitoring Management Views within the information schema is a great way to profile your cluster workload

We also took questions from webcast attendees, and here’s what we said:

Question: What if my use case doesn’t fit into rowstore or columnstore? What if it has some transactional elements and some analytical?

Answer: Some customers have a notion of “active” data that fits in memory, and once that data is expired or finished updating, it is moved to a columnstore table with the same schema. Ultimately, this is workload dependent. With our customers, we scope test workloads and look at the performance of queries on each model. Our sales engineering team members are experts in this domain.

Question: What are the benefits of the MemSQL columnstore compared to other Massively Parallel Processing columnar systems?

Answer: MemSQL columnstore facilitates real-time streaming data loading through in-memory optimized data structures, and caching metadata allows greater performance via segment elimination during query processing. We also use query vectorization techniques

本文来自电脑杂谈，转载请注明本文网址：
https://www.pc-fly.com/a/jisuanjixue/article-81538-1.html