loading...

Introducing Hole a new serverless technology for Node.js

matuzalemsteles profile image Matuzalém Teles ・3 min read

Introducing New Serverless Technology for Node.js

Publication originally posted at https://hole.build/blog/2020/02/16/introducing-hole-serverless.html

Introducing the new generation of serverless technology for Node.js, efficient and cold start to zero with hole.build.

Managing servers with a complex infrastructure a professional team focused only on monitoring and consuming several hours to decide how to scale and support large peaks in access and use of the application's APIs, has always been the problem to deal with high availability and fast growth projects.

Over time, several technologies and standards have been created to deal with this and they are one of the most critical things of a product when not well thought out and orchestrated can be a big headache for the rapid growth of a company but it requires professional demand qualified staff for monitoring and security. At the beginning of a small startup this can be a big cost because they need to grow quickly, they need to worry about their product, validate, win over their first customers and start selling.

Maintaining an infrastructure, monitoring and servers team to maintain the product can be very expensive, paying for services that are not used or are idle in times of low access can impact the company's balance sheet.

A few years ago, the "serverless" (FaaS) movement and technologies began to emerge, with a view to solving these types of problems:

  • auto-scaling,
  • scaling to down,
  • zero servers,
  • without complex infrastructure,
  • pay only for resources when used,

This is beautiful and it looks like the best of all worlds to start building the product on top of that, but with that it came with a main disadvantage: performance, functions that are not executed with high frequency can suffer a higher response latency than the code being runs continuously on a server.

Provisioning of the functions from other serverless platforms

On other serverless platforms, when a request arrives, the platform provisions a container with its function runtime, there is a waiting time until its function can actually start to be executed and process the request, called a cold start. Your container is kept on "hold" for some time so that can process another request but without cold start, when new requests arrive and some of these containers are already occupied, other containers will have to be sized to process the new requests starting with the cold start.

At Hole we built our technology to solve some of the main problems of serverless: performance, security, monitoring and debuging. Our functions are executed with cold starts to almost zero, we limit and add more layers of security in the environments of execution of the function, we show metrics of requests made successfully and failed in more details and insights on the performance of your code. In addition to improving serverless technology, we are very concerned with the experience of using the technology, console, design and friendly docs.

Asynchronous requests for Hole functions

Our functions can be performed with cold starts to almost zero and functions can be configured to handle more than one asynchronous request increasing the limits of provisioning your function. You can read more about how our technology works in our documentation.

These are some crucial points that we are attacking but we want to improve even more how companies interact and work with serverless technologies is just the beginning, and we have many things that we want to show. It will be a long journey and we are excited to share our learnings and thoughts as we progress.

Today, we are starting to accept teams and companies for our private alpha. If you are interested in joining early and influencing the direction of Hole, sign up here and follow us on Twitter.

Discussion

pic
Editor guide
 

After reading "How it works" an isolate sounds like Node's VM module. Could you please give us more details on how running several functions in 1 process is more secure than traditional approach with 1 function per 1 process?

 

Hey @the_spyke , good question, it may sound like the VM module but we are not necessarily using it, because we would have problems of resource struggle between the functions (CPU, Memory. We have other approaches to address this.). Isolate is a concept and the V8 mechanism to run sites with contexts separately, we use this same mechanism to perform the functions but it is not necessarily the only layer we use to create the sandbox environment it does not mean that it is safer than a traditional environment running the container with only one function. In addition we have other layers of security on top of these processes, we also rely on the isolate context of the V8 that has long been guaranteeing the security of the web with this approach, similar approaches that are appearing with WebAssembly with the implementation of WASI.

The tradeoff we have when we go for this approach is that we may lose the possibility of running functions in different languages ​​but we are studying the possibility of running WebAssembly in our functions.

I hope I have answered your question!

 

it does not mean that it is safer than a traditional environment running the container

In the article you wrote that "At Hole we built our technology to solve some of the main problems of serverless: performance, security", so I was interesting in how you did this. Thank you for your answer.