Cloud Native Computing Foundation Cloud Native Wasm Day EU 2022, 19 May 2022

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: Lightning Talk: WAMR, Intel - Le Yao, Intel

Description

Lightning Talk: WAMR, Intel - Le Yao, Intel

Lightning Talk: WAMR, Intel - Le Yao, Intel

This session introduce the WAMR - a hign performance and small footprint WebAssembly runtime in Envoy. The advantages can get by building envoy with this wasm runtime.

A

Hello, everyone welcome to my topic about walmart and why uh I'm loyal from intel as a cloud australian software engineer, and today I would like to introduce walmart and from three aspects. First, I will give a brief explain about enguava sim. Then I will introduce the web assembly my current time, which is developed by intel mama here and uh finally, I will give the benefits we can get from this work.

A

You know why we enable obama and why yeah uh anyway, integrity was same technology, because the same can bring accessibility and flexibility to involve without bringing up and what is done. We can learn about the envirosim as these four modules.

A

First nua has a external internal c platas filter to encapsulate what same code and offer more api implementation, and uh it also have a project named because they were same host. If you encapsulate, what's in runtime, extract sandbox api and request the info api to sandbox.

A

uh We all know web assembly is a stack based, virtualbox virtual machine yeah and uh another project is approximate same sdk yeah. It will explore export the sandbox api and declare the reference to invol api and uh also in catalyst. Encapsulate my api and control the sandbox api, so we can use the ssdk to develop our waxing filter and the implant mentions uh interface we defined in the envy was in filter. Then we can integrate this filter into the invoice service mesh yeah.

A

um Currently we must pause the volume five was in runtime in its latest release, and one more is here: yeah the warmer is what we bring into and why the develop build of the enviro binary is a wave around them. It is based on the ray bar engine in browser yeah and over the same time I know.

A

Waim here is a other runtime which is with different features. Part yeah and the runtime noon uh is default, compatible modules linked into and one yeah. So what is warmer here?

A

This last shows the web assembly macaron temp warmer over real, and we can see we can see it here. One more itself supports many cpui texture and at least here, under the platform support listed here, uh we can see in ramona itself support many uh cpu architecture and the platform yeah, and it had a successful community.

A

uh Walmart itself target is a small list and the first standalone same runtime and the module designed to cover the uses from low end device to cloud yeah and the supporting interpreter ahead of time and the just in time. Compilation is in the warmer itself is a sea-based implementation and iom-based compilation.

A

um One advantage it has is it has a small flipping and a memory consumption. The ram call of it shows the like this yeah uh under there is a self-implemented ahead of time, audio loader. uh It also supports the intel sjx no need for the sdx label as other things, and it can mature the application for using its application framework and uh it's very easy to post new platform.

A

You can get the code detail from this link. Yeah.

A

Here we give the one more performance number uh we have tested on the uh x86 platform, architecture and platform, and uh it says the performance number compared with the native gcc compiler. We can say the ahead of time. Performance is from 0.4 to 1.5 of the gcd calculation among the workloads. This is the benchmark of it, and this is the performance number yeah yeah, it's gcc generic sse code, yeah.

A

And uh here we will give the uh sgx or kubernetes related adaption under engagements.

A

First most of the intel project, such as project private data object, open source project can get the benefits from using wama and uh ali and financial, from alibaba, adopt obama on sdx for the uh china leading blockchain platform and also kubernetes under the open source process and the ali live adoption.

A

The walmart, with sdx on multipatters computer protector and also by dual x level, mesa te also um integrator and warmer and uh stitchers.

A

Other customers, like microsoft, alibaba and our under team yeah.

A

So wait, oh, I have. I have gave you a brief introduction about walmart, so why we bring walmart in why? Because itself, wrong or similar time may have caused the binary may increase the binary size and the performance number is not real in our text, so we bring one more into enviro and do some performance tests to get the this advantage.

A

The first is the memory consumption we use the stalker commands to get the noise and peak, so we can see in this table uh using one more interpreter model or walmart. Just in time model we can say the memory consumption uh have a largely deal decrease yeah later.

A

Also, the binary says, based on way bar on time, is larger than murmur using warmer to build the binary. We can get the y binary and decrease the set of about 50 and the ships set about 10 yeah and the high performance uh way using the lighthock to test the performance will generate some hp requests to avoid a gateway and get the performance number.

A

We can say in most benchmarking uh we get about five to ten improvement, yeah yeah, also uh using warmer. We may enable sdx into ny hp filter which will increase the safety of the hp filter chain. Yeah.

A

So here's the reference I will give you all to about our work working uh and why uh a pull request to and while proxy have been merged on the the proxy was in uh project. And this is our one more time and if you have any question you can contact me using this email. Yeah thanks.