Refactor the GraphQL Module to Improve the Performance

[DimuthuMadushan]

Description

We have conducted a load test on the GraphQL module, and it shows that even though the CPU usage is high, the throughput is low. After profiling, we identified the areas that take most of the time while processing requests. To improve performance, we are planning to try out the following changes in GraphQL module:

1. Reduce the usage of cloneReadOnly() in the parser.
2. Minimize lock usage.
3. Use the readonly type wherever possible.
4. Reduce the usage of Ballerina's is checks in the parser.
5. Replace string:Char usages with string.
6. Refactor the Node modifier context.
7. Refactor Response formatter.

Version

update 9, 10, 11

[DimuthuMadushan]

I attempted tasks 1 and 6, but they did not show any improvement compared to the above changes. Currently, I am working on refactoring the response formatter.

@ThisaruGuruge has completed tasks 2, 3, 4, and 5, which have increased performance by 3X and significantly reduced latency. Refer to the performance comparison for more details.

Complex query:

Simple query:

PR: ballerina-platform/module-ballerina-graphql#2175

We have released GraphQL 1.13.2 with these changes.

Currently i'm working on task 7.