Istio IstioCon 2021, 10 Mar 2021

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Istio 2021 Project Roadmap

Description

#IstioCon2021

Presented at IstioCon 2021 by Neeraj Poddar and Louis Ryan.
Closed caption and subtitles in Spanish by Pedro Galvan.

Neeraj Poddar and Louis Ryan from the Istio technical oversight committee, lead an update on the development of the project and the roadmap for 2021.

A

Yeah, good morning,, everyone. and good afternoon. good evening. If you are.

A

Coming in live and joining this, like me from india, all around the world,.

A

Excited to present the istio 2021 roadmap with louie.

B

So we're, we're talking to you about the roadmap, and as our subtitle says,.

B

Our goal is to try and make istio for operators and administrators as boring as possible. so,.

B

To that end,, this roadmap should be extremely predictable.

B

But before we get into what we're going to do in 2021,.

B

Let's talk a little bit about 2020 and how kind of 2020 ties into what is coming up. This year.

B

So, obviously, for those of you who have been actively involved or active users of istio,.

B

2020 was a pretty big year in terms of changes that we made.

B

A lot of big things, happened, and a lot of these things were laying the.

B

Groundwork for what we want to achieve in 2021,, which neeraj is going to talk, later.

B

But a lot of 2020 was about simplifying istio,, simplifying the user experience around istio,.

B

Making it easier and safer for people to do upgrades in production,.

B

Tying back into the tooling that people like to use.

B

You can see there is quite a long list of stuff here.

B

We made some pretty big architectural changes,. Istiod got introduced,.

B

Which, hopefully, has yielded some benefits for people., but we also put lots of investment.

B

Into the lifecycle, management and tooling around istio, so that it was easier to install, upgrade.

B

If you are using helm,, we introduced support for helm, v3. so,. A lot of this is focused on.

B

Just making the total experience of maintaining an istio solution in production easier for users.

B

In a parallel track,, we did add some significant new capabilities.

B

In particular,, we added the ability, in a couple of places, to improve.

B

The extension experience of istio. and one of the major reasons why we did this.

B

Is so that istio itself didn't have to pick these capabilities as built-in.

B

So, extensibility represents a safety control valve for the project.

B

That allows us to enable features that customers want without having to make those features. Part.

B

Of the technical debt or burden of all the other customers who wouldn't necessarily care about that.

B

Feature. we did some work to make security and the security posture of istio more.

B

Coherent,, more reliable. and in particular, we brought the auto mtls feature to production.

B

We also did a lot of work. To, and you'll see a continuation of this in 2021,.

B

Start to stabilize the api set that we had., so in 2019, we did a lot of work kind of revising apis,.

B

And we introduced the gateway apis,, we improved the authorization policy apis,.

B

We did a lot of api work. and so in 2020, we spent a lot of time stabilizing that work and.

B

Bringing those apis from alpha to beta and beyond.

B

We also did a bunch of work on the environmental footprint of istio.

B

We allowed for vms and integration of vms into the mesh and brought that feature forward into beta.

B

So that you can rely on this feature and consider part of the stable roadmap of istio.

B

And did a lot of work in just kind of tuning and optimizing and stabilizing various different.

B

Performance and scale aspects of the product to make it very stable., and so this is kind of 2020.

B

If you look at the roadmap slides that I gave at service mesh con at the end of 2019,,.

B

These achievements are very aligned with what we were looking to do.

B

So, overall, I can say that 2020 was actually a very successful year of the project,.

B

Both in terms of the technical direction and the architectural and product achievements, overall.

B

We see the impact of that work in the rapid and impressive adoption of istio throughout the.

B

Industry. the fact that we're talking in istiocon today, with thousands of attendees and registered.

B

Signups, people looking at this content, is also a reflection of the fact that istio is.

B

Quite widely used in production now in a lot of different places., the cncf did a survey of the.

B

Use of service mesh technologies in production, and istio is clearly leading in this category.

B

So,, that's obviously very gratifying for people who work in this project,.

B

To see this kind of growth and to see how well istio is doing in the market of open source and.

B

Commercial service mesh, products. so,. We think that the work that we did in 2020,.

B

And even in years prior,, and what we've learned to do in 2021 is going to reinforce this trend.

B

We spent a non-trivial amount of time, like gathering and listening to what our users are.

B

Saying about the usability and experience of istio and in particular, the kind of.

B

Operator experience, the day, one and day two experience in product. so,. We ran some surveys.

B

And in particular,, one of the things that we were very focused on was upgrades., even in 2020, istio.

B

Was moving quite quickly and there were a lot of architectural changes. so,. We were sensitive to.

B

The fact that those upgrades, while those changes were going on, might be disruptive to users.

B

So we wanted to kind of counterbalance that with investments and the experience around upgrades.

B

Even still though,, we were getting feedback from users that the upgrades were.

B

Hard to consume quickly enough sometimes., we release on a quarterly basis.. We intentionally match the cadence of kubernetes to also release.

B

Quarterly. so, a question for operators: we ran was:. Are you upgrading fast enough and.

B

What were the constraints around upgrades that those people were experiencing and how could we go.

B

About addressing them?, so we ran this user survey in q3 and analyzed the results in q4 last year.

B

And most users didn't feel like they were upgrading, frequently enough, or they weren't sure.

B

Some users were upgrading quickly, and often that was actually correlated with whether there they.

B

Treated their kubernetes clusters as a catalog of pets., so this was an important question.

B

If people didn't feel like, they were upgrading, frequently enough, what was holding them, back?.

B

And then obviously,, you know, with any software that is part of a security solution, which istio.

B

In many cases, is, that you want to make sure that we are pushing out the stable releases,.

B

Releases that have their cves, addressed. and istio has been very good about addressing.

B

The cves and the different components of the system and pushing patch releases.

B

We pushed out a decent number of patch releases last year as a result. so,. It wasn't very hard to.

B

Be able to go and analyze the state of user's installs and see that a non-trivial number of.

B

Them were on releases that still had known cves in them., and also that a non-trivial number were.

B

Using istio versions that had fallen out of our declared support, window. istio last year.

B

And for some time had used n-2 support policy for releases to match what kubernetes was doing.

B

Kubernetes experienced a similar kind of growing pain and feedback set from its users.

B

And they actually extended their support window in 2020.. They went to.

B

Almost a year long support window for releases. so,. We have started to match that in the project.

B

This slide and the previous slide are basically saying the same set of information, right?.

B

If you're sitting on an unsupported version,, what is holding you back from upgrading? other than,.

B

You don't have that many hours in the day to get work, done,, which is a universal problem. but,, are.

B

There specific concerns that are holding you back and what can we do about them?.

B

And then the final thing,, you know,, there's obviously quite a.

B

Lot of tooling choices in the kubernetes ecosystem for people to manage software.

B

You know, packaged solutions like istio are expected to deliver fairly packaged and low.

B

Friction ways for those solutions to be managed,, updated, upgraded in production.

B

But they're also expected to work within the ecosystem of ci/cd systems that users use.

B

For everything else that they run in their cluster., so while we put a lot of time and.

B

Effort into the istioctl and the istio operator,, which basically share a codebase for managing.

B

Istio instances,, there is still a non-trivial number of users who.

B

Use helm, like to use helm, are comfortable using helm and would like an easy and.

B

Well-Maintained solution for managing istio solutions in production with that stack.

B

And then there's a non-trivial number using a managed service., as you can see, here, you know,.

B

They might be using the ibm, managed version of istio in production, or they're, using something.

B

From rancher, argocd or something. so,, we still have a non-trivial amount kind of operational.

B

Life cycle management tool; diversity that we have to address much like any other project.

B

So,, paying down some of the architectural debt that we had, made supporting a more diverse set.

B

Of tools here, easier,- that was a big part of things in 2020., but there's still,, you know,.

B

Work and investment to be made here. and so,. With that, I'm going to hand things over to neeraj.

B

And he's going to talk about the lessons we learned in 2020,, how we're applying them to 2021.

B

And then I'll hop back in later, and we can talk a bit about some other things that are on the road.

A

Yeah, thanks louis. as louis said,. The 2020 year was a.

A

Real successful year for our project., we achieved a lot of features,. We made a lot of installations,.

A

Simplification and ux improvements,, but at the same time, we wanted to listen to our users,.

A

Listen and make sure we are adjusting our roadmap based on that feedback.

A

And the idea of theme for us is basically creating a narrative for the entire year, so that it's.

A

Easily digestible. and you can understand why the istio project is doing what it was doing, right.

A

And more than that,, creating a theme for the year helps us really prioritize. The features that we.

A

Are quarter over quarter or for minor releases and help assign resources. so based on all that.

A

Feedback the theme that we came up for 2021 was simplifying and improving the day. 2 operations.

A

So here's a picture that I took from a nice dzone article,.

A

Which? Basically tells you what a day 2 operation looks like. so, in day, 0 you're.

A

Doing a lot of architecting designing, in day 1 you're, trying to install, set up configure.

A

Both of them are still kind of in the trial phase and learning phase. day. 2 is where you.

A

Have made your choice., so many of our users have made their choice,. They want istio in production.

A

So we want to make sure we help them be successful in running and maintaining istio.

A

You want to keep the lights on., you want to make sure you have the nice troubleshooting experience.

A

You have the right upgrading experience so that more and more upgrades are uneventful.

A

As louie said,, it was clear from the feedback that, even though we have given a lot of diverse.

A

Choices and tooling, to install and configure, we weren't really meeting our user expectations.

A

In terms of upgrades, and which was the root cause of users being stuck at older, versions.,.

A

So, in that theme,, what I want to tell you is:, as we are focusing on day 2 operations,.

A

What does it really mean for you as a user? so, louie,? Can you go to the next slide, please?.

A

So,, what it means is that the istio project in totality has matured significantly from where.

A

It was not just like 2 or 3 years, ago,, just even last year, right?, so we are moving from.

A

The phase of moving fast and breaking things, into moving slowly and fixing things. we want to have.

A

The boring predictability that louie was talking about.. We want to sustain the tremendous.

A

Production adoption that we saw in almost all the surveys that come out in the industry.

A

Istio is at the top of the list for service mesh adoption.. We are also realizing that, with all the.

A

Architectural changes that went in during 2020,, we have come up with a pretty stable core now.

A

Our foundations and the architecture is pretty solid, with istiod and extensions and telemetry.

A

Through a web assembly., so we don't want to keep changing those foundations again and again.. So our.

A

Focus is to make sure the fundamentals and the architecture remain the same.. The current istio.

A

Functionality, on top of that mostly meets our user needs., like many of our users, have often said,.

A

The functionality is raised., the functionality is there., we just want to be successful in.

A

Using and consuming it. at the same time,, we want to introduce new features,, but we want to.

A

Make sure that feature, production and the quality of the feature and direction is well-rooted in.

A

User feedback. so, how it impacts you as a user, is that you will get a lot of stability and you.

A

Will get new features,, but that new feature rate might not be as aggressive as it was in 2020.

A

And that's collectively as part of toc and the steering and the working group.

A

Members,, we are realizing that it's good for us to slow down a bit,.

A

Make sure we're stabilizing and helping our users and not breaking things like we were doing.

A

Earlier in 2019 and 2020. and lastly, the operational overhead of maintaining upgrades.

A

And debugging that has to be significantly reduced so that the man hours and the profile.

A

That users have for running, istio is less. and it has to significantly reduce.

A

Over and over for every release. so, basically,, this is our one representation of how we are.

A

Going to make istio boring. and there's a nice dog who is bored by our presentation,, I guess.

A

So, the reason I say "again" is because if you look at the 2020 blog that we wrote,, we also.

A

Wanted to make istio boring again,, but there was a realization that our architecture fundamentally.

A

Had some flaws. and because of that, we had to course correct based on the user feedback.

A

And when the architecture itself significantly changes,, it's really difficult to avoid breaking.

A

Changes. so now we are at the stage where we can really make istio into a stable,, predictable.

A

Feature and a stable product that users can continuously use in production. and.

A

Hopefully, this time for 2021,, we really want to make istio boring for all our users. so,.

A

Based on all the feedback that we got,, we were able to classify the istio users in basically.

A

Two broad categories. so, on one side, we are seeing early adopters who consume the latest.

A

And greatest istio., they want to keep up with the rapid rate of innovation in the community.

A

They want to utilize the new features, and the key for them is having the tooling to ensure that the.

A

Upgrades are frictionless, because they're, the ones who upgrade the most frequently.

A

They have to make sure they spend less and less time on it.

A

On the other hand,, we have another class of users who are maintainers.

A

They are the users who have fixed budget for infrastructure, spend overall,, which also includes.

A

Istio. they want more predictability, right?. They want larger support. Windows, like louie.

A

Was saying that kubernetes has also recognized some of these traits. and then they also want to.

A

Make sure they can go from release x to release y without going through incremental releases,.

A

Because they release less frequently., there might be a lot of istio releases in green and.

A

Going over each and every step is a big burden on them.. So the theme for us, the day, 2 operations,.

A

And the focus areas in that is mapped to make sure we meet the needs of both of these user classes.

A

So one of the first focus areas that I am going to talk about is stability and maintainability, and.

A

Then the remaining two focus areas, louie is going to cover in the next few minutes., so stability and.

A

Maintainability is one of the key areas for us. and within that, improving the upgrade experience.

A

For our users is the key., so we realized we were not meeting the mark. so right around.

A

Late last year, we tried to figure out what we can do collectively.

A

So that we can ease this burden from the users, right?, so we created an upgrade working group.

A

Whose responsibility is to make sure upgrades collectively work as we expect.? We were looking.

A

At,, we were kind of realizing that, even though some of the features are tested.

A

In isolation,, when they come together as part of a new release,, sometimes the upgrades break.

A

And these are the type of holes that the upgrade working group is trying to fill. both by making.

A

The overall infrastructure robust,, but also creating some policies that can be used by.

A

All the working groups in creating features., so our future releases are much more seamless.

A

Promoting the revision risks upgrades, and making sure that is the recommended approach for most of.

A

Our users is a priority for this year.. We want to support skip-level upgrades and that skip-level.

A

Upgrades will be primarily supported through revision based upgrades and promoting skip-level.

A

Will allow some of the users who upgrade less frequently to still consume the latest istio.

A

We want to have the pre and post upgrade checks so that, if you have enabled some features that.

A

Might be experimental, or if you have some configuration,, which is incompatible with.

A

The newer version, we are able to flag it before you go through that pain, yourself.

A

And then lastly,, we found some gaps in our testing,, where we were doing upgrade testing,.

A

But they weren't really mirroring some of the production use. Cases,.

A

Which was leading to escapes., so we are trying to come up with better test metrics. and in that.

A

I would encourage people from the community,, especially users, who have large production.

A

Deployments to come forward and say,, what are the type of things that are breaking your.

A

Environment when you upgrade? and then hopefully we can incorporate it in our testing. Strategy.

A

Another big part of making sure that day 2 operations are better, is the enhanced.

A

Troubleshooting. we have made significant investment and improvements in the istioctl.

A

Tool to add analyzers,, which look at cross config, cross multiple configuration issues.

A

And it has been a really great tool for people to understand when they make misconfigurations. That.

A

Are not just in one resource,, but when multiple resources work in conjunction.

A

So we'll continue to add them.. We are working on an events api to surface in a more programmatic.

A

Way, issues that might be happening in your infrastructure., so you can consume them.

A

Through other means and actually flag and understand what went down. similarly,.

A

Api validation will continue, as we have continuously made improvements in that. and then.

A

The istioctl bug report tool that was added in 2020,. We can even make more improvements so.

A

That you can collect the information you need to debug and troubleshoot your istio service.

A

Mesh easily in production,, looking over a bunch of different information and trying to hunt for that.

A

Is a big time consuming effort for most of our users., and we want to minimize that.

A

Lastly, another issue that often happens with the stability and misconfigurations is users.

A

Don't understand which istio apis are mapped to which personas and roles, right? we have.

A

Developers using istio,, we have mesh operators and we have mesh admins, and we want to make sure.

A

That istio apis are well aligned with these user roles and that we give them the capabilities to.

A

Configure the responsibility and configure istio the way they want., so that clarity, we will try to.

A

Make sure it's there, this year,, so that less accidents happen in mesh and different people.

A

Wearing different hats in your organization can successfully run this year in production.

A

And as louie mentioned,, the feature graduation has been a theme in 2020, and that will continue in.

A

2021. primary,: we want our users to run istio with most of the features, which are battle tested,.

A

Which are stable or beta., so the working groups came together and created a nice enhancement.

A

Workflow., so we go through a sequence of steps for each of these features when we go from.

A

Experimental to alpha, from alpha to beta or beta to stable., so I have listed just a few of these.

A

Features that we plan to make sure that these go to the next level of stability.. This includes cni,.

A

Ipv6,: dual stack: support for ipv4 and ipv6,.

A

And then virtual machine expansion,, which has seen significant investment in 2020,,.

A

We'll continue to invest and make sure we reach much more high levels of stability.

A

Similarly,, the multicluster mesh will have improvements, not only from the user experience.

A

Point of view,, but also adding clusters from multiple administrative domains and helm. V3 is one.

A

Of those questions that has cropped up quite a bit in istiocon., so we introduced helm v3 in the later.

A

Half of 2020, we'll continue to promote it,, make sure that it goes through the feature. Enhancement.

A

Workflow and make it through beta and stable., so helm v3, along with the istioctl and operator, will.

A

Continue to be the main means by which you can primarily configure and install istio., we realize.

A

Many organizations have made investments in helm, and it's important that we meet their demands,.

A

Right? and then another effort that is going on is to make sure we evaluate our current feature.

A

Status and fix the gaps we realized early on in the community. We might have moved things too.

A

Quickly and we want to make sure that whatever gaps that we identify, we fix and we follow our.

A

Own process to make sure the feature status accurately reflects what's in it., so with that,.

A

I think I'm going to hand it back to louis who's, going to talk about the remaining focus. Areas.

B

Thanks neeraj. so,, you know,. Obviously, what we've seen up to this point is very much a.

B

Focus on tooling and stabilization of the core features of istio very much.

B

Targeted around that day, 2 operations theme that neeraj introduced.

B

Istio itself is obviously going to keep moving, right. you know,. There is still demand for new.

B

Features and capabilities. and neeraj has described what we do now is we put any of.

B

Those new features through a very well-defined and very measured process for graduation.

B

I encourage people to go and take a look at how we progress features through the workflow. and also,.

B

You know, that workflow has a very well-defined pacing to it, right?, so features will be.

B

Introduced, will be progressed and will graduate,, but they will do it at quite a measured pace.

B

So, as I mentioned earlier in the introduction,, we did put a lot of focus in 2020 on enabling.

B

Extensions to istio. there are a lot of really important and good reasons why we did this.

B

One obviously is to act as a pressure relief valve for us not having to build features.

B

Into istio itself., and this allows those features other ways to get qualified.

B

By the people who care about them and use them, possibly prior to them being incorporated.

B

Into istio itself., but also extensions, are super important because there are integrations into.

B

Ecosystems within a given deployment that istio itself couldn't actually incorporate as a feature.

B

A common example that I see is large enterprises, and they often have policy systems for access.

B

To services and data that they have developed in-house and have been maintaining for many.

B

Years and use in their production systems prior to the introduction of istio.

B

And so to bring istio into that ecosystem, they need to be able to integrate with those policies.

B

To make sure that their posture,, whether it's a security posture or an ordering posture or.

B

Requirements around logging are still being met as they make the transition, and decouple.

B

Their kind of infrastructural management aspects of that from their applications,, which is,.

B

One of the core value propositions of istio., so in 2020, we put a lot of effort into webassembly,.

B

And I'm not going to go into all the details of webassembly, but you know,. We.

B

Put a lot of work into stabilizing webassembly as part of envoy, and the istio team, eventually.

B

Completed the upstreaming of webassembly support into the mainline envoy, itself. and.

B

That included stabilizing an api for webassembly extensions that execute in that context.

B

So it was a huge contribution. and then,. You also saw.

B

An evolution in the ecosystem around developer tooling for webassembly extensions to make.

B

It easier for people to build and test webassembly extensions and there's some.

B

Good talks about that already at istiocon,, which I personally was very excited to see.

B

And so the other thing that we're doing is we're going to provide the other kind of.

B

Last mile of infrastructural support within istio to make sure that extensions can be distributed.

B

And managed within the mesh. so that the value that you're getting by writing and the value.

B

That you get out of the development workflows can be delivered into live production, systems.

B

And there's obviously, an ecosystem of developer tooling,, whether it's sdks build tooling support,.

B

Testing support, and also you know, work being done in the community and not just as part of istio,.

B

But kind of as the broader ecosystem, to enable discovery of wasm extensions that.

B

Other people have written that you might find useful in your context.

B

So that was a big focus for us. another thing,, and this kind of ties very much into the integration.

B

With policy systems,, there is another form of integration that envoy standardized and that we.

B

Have been using for a while,, which is the ability to extend the authorization mechanisms built.

B

Into envoy, by calling out at policy evaluation points in envoy to external systems, using a kind.

B

Of pre-canned interface. and there are two forms, one in grpc,, one in http. and so we're providing a.

B

Better api experience in istio to allow operators of istio to configure that extension point.

B

And then bind it into the kind of policy experience of istio itself., so you decide,.

B

I know if I have my external policy system,. I want to cause in this context, right,.

B

That context could be for these workloads, for these workloads, for this traffic pattern, right?.

B

So that's now being introduced and in the last release we actually shipped it in.

B

I believe in development and tying that user experience at the api level is really the.

B

Kind of powerful thing that we're doing. the capability had been there in envoy for some.

B

Time and many users using it through the envoy, filter api,, but the envoy filter api.

A

Being this kind of a breakglass experience has a number of downsides for operations in production.

B

It's harder to operate and so, refactoring that and providing a more stable api base for people.

B

And integrating into the overall api experience for authorization is a.

B

Much more solid experience for what is a very important use. Case.

B

And then, finally, we're doing a kind of similar thing again around the telemetry experience.

B

So, istio kind of shipped with a number of pre-baked telemetry experiences,.

B

And envoy had an even kind of larger range of available, telemetry integrations.

B

And so we wanted to improve the api experience in istio, for enabling those telemetry integrations.

B

And enable it in a way that made sense in the context of istio itself, rather than kind.

B

Of falling back to the break glass mechanisms that we were offering, whether it was, you know,.

B

Create your own build of envoy do funny things at injection, time, or use the envoy filter api to get.

B

Break glass experience., so basically what you see us doing, is, you know,, looking at the places where.

B

People were going through some pretty serious hoops to get the integration that they want.

B

And then elevating the user experience so that they had a much easier and much more maintainable.

B

Extension experience for critical use, cases, making those part of.

B

The kind of api experience of istio, itself. and so you'll see that trend continue in 2021, as we.

B

See the critical extension points that are being used in production,, elevating user experience.

B

And lastly, you know, we're obviously doing a lot of work in the development experience.

B

Within the community itself, around istio to make the product itself more stable.

B

We've already done a lot of work around how quickly we can qualify a release.

B

Our releases have become much more predictable in 2020.. We've had to do fewer patch releases,.

B

Our developer efficiency has improved., we have fewer test, flakes. the code coverage.

B

And feature coverage,, and it's really important to kind of distinguish between those two things,.

B

Has improved by leaps and bounds. we've gotten better at the process of feature promotion.

B

We have better tooling support for that. we've, even gotten, you know,.

B

We put a lot of work into making sure that there's a strong correlation between.

B

The documentation that we're shipping for any release and the actual features in the product.

B

And that includes,, you know, actually having tests that test, that correlation. and so,.

B

From any kind of product maturity standpoint, ensuring that the developer experience for istio,.

B

What we, the community members experience when we build features, is a lot smoother than it would.

B

Have been at the beginning of 2019., and we intend to keep investing in this because it's very clear.

B

That it's yielded not just a better experience for developers of the project,, but also for.

B

Users, because they see more consistent releases,, they see fewer bugs,. They see a more efficient.

B

Product and they have more predictability about the outcomes that they're going to get.

B

And so, while that's,, you know, a little bit about the sausage making and maybe as a user.

B

You don't care so much about this, within the community. This is actually really important.

B

Then, because the world is moving,, there are always some important concerns that are.

B

Still somewhat orthogonal to the main theme that we talked about.

B

There are sort of within kubernetes itself,. There are some api evolutions going on that overlap.

B

With the use cases that we express in istio. and so we're hoping, the kubernetes community.

B

And working hard to kind of track those feature and api proposals and make sure that istio is well.

B

Positioned to deliver a really good experience and a very capable experience., if those are the set.

B

Of apis, you want to use for certain networking functions. and in particular,, I'm referring to.

B

The kubernetes gateway api proposal and also their multicluster api proposals.

B

I certainly encourage people to go and take a look at those and we can provide things for them.

B

Envoy is also a moving target, right? envoy.

B

Is a pretty impressively large project at this point.? It accumulates features.

B

Some of which are very tied to what we implement in the control plane like delta xds.

B

Which, hopefully, will make the control plane more efficient at scale.? And then there are.

B

Some other kind of more transport oriented things that are occurring on envoy. you know,.

B

I didn't mention here quick support,, but it's certainly something they're kicking the tires.

B

And they're doing a bunch of work with http2 tunnels, the ability to hold tunnel protocols over.

B

Http2, which we expect to make use of for a variety of functions within istio., because a lot of what.

B

Istio is actually tunneling various protocols over tls today and doing that on http2 over tls.

B

For those protocols might actually do a better job. and then there's some other things around.

B

Stateful sets and zero config that we are going to take a hard look at for very specific.

B

Application workload types,, particularly ones that have built in service discovery, mechanisms.

B

So with that,, I think we're about on time for questions.