Hi I'm Josh Lambert, a product manager for gate lab, and today I'd like to talk about what we're working on for the memory group in 12.9.

We are continuing to work on the three main areas that are working on previously and 12.8. This release as well. Those are working to enable Puma on get lab comm we're about 28% of the way they're rolling out across the fleet and we're working to continue to roll out across the entire fleet. This release we're also working to be able to more confidently bring Puma to all of our self-managed customers as well, so they can realize those benefits and we're continuing to also improve on project import exports as well.

Let's dive into the particular issues here, you'll see quite a lot of them here, our flag with import/export. This is a critical area for memory team and you can see we're working to improve the performance and reliability. This includes things like making sure public imports are consistent memory. We can use new line, delimited JSON, zwey to more efficiently process. These imports and exports with reduced memory, consumption and again just continuing to make further performance improvements as well. Then that result is.

We want to make sure that imports and exports are reliable and faster than they are today. Currently, in some cases. Well previously, before I started this work, it was potentially possible for a very large import to fail and continue to be retried over and over again, because how much memory considered plus, of course it would take a number of hours. So we working to make sure that doesn't happen, that large imports reliably work and significantly improving the performance.

That said, sometimes this can be extremely large, especially for a large monetary post, and so they still will take some time, but significantly less and the Pumas side. You can see we're also to working on a couple of the edge cases here that identified some pool, slides that are working to reduce. There was an outcome of another issue and also continuing to work on a few other areas of Puma like improving documentation and as defaults for omnibus.

Essentially, we have identified some good worker and thread counts for get babcom, but we have fairly large instance sizes here on.com, and so we want to make sure we have strong documentation and the faults and omnibus that we'll set up the most number of people for a success and if they need to, they can then further have information on how to effectively tweak the settings for their own instance sizes.

So those are the two and we really three key areas: Puma on both calm and for ourself and as customers, that's available an experimental status right now. If it curious about trying Puma, please read the documentation. We saw significant improvements to both memory, consumption and performance with kuma uncom.

Details are available in our total eight release post and also, of course, I'm continuing to iterate on improving and making this input exports more reliable in performant. Thank you so much and looking forward to bringing these great enhancements in 12.9, which we release on March 22nd.