# ignacio(jsign) - Update 9

This week I focused on a plan we talked with Guillaume and Kevaunray in a call we had the previous week. We're trying to figure out if we can have a concluding answer around the limits of performance comparing MPTs and VKTs. This was asked by Dankrad.

In our discussion we talked about two main paths:
- Use the `replay-block` benchmark that Guillaume is working on, which provides a "gas per second" throughput metric.
- Build a benchmark that does an in-memory comparsion of insertion performance of MPTs and VKTs (e.g: inserting 10k key-values and comparing performance). Do this with the real production implementation of MPT in Geth, and the current VKT implementation (to be used in Geth).

The ideal metric is the first bullet, but this benchmark will be ready to be shared soon (still wrapping some work). In the meantime, I foucused in doing the latter which is also useful.

I did a PR in `gballet/go-ethereum` geth fork with these benchmarks: https://github.com/gballet/go-ethereum/pull/146

Instead of diriving conlusion right away (see results in the PR), I noticed by looking at the profiling information that there were more opportunities to improve the performance between both. I've already worked on some of these improvements, and still have to share them with Guillaume... but they're looking promising. Probably, more new on this soon.