Hacker News

Top stories

Live mirror
30 storiesupdated just nowView source snapshot
  1. Human brain is two separate organs, Stanford Medicine-led research finds(stanford.edu ↗)
    54comments
  2. If math is more than proof, we need to better celebrate the rest of it(terrytao.wordpress.com ↗)
    27comments
  3. GPT-6 Astra Solves a WWI German Radio Cipher(prinzai.com ↗)
    35comments
  4. San Francisco Onion Futures Company(onionfutures.com ↗)
    61comments
  5. Android 17 is the first since 3.x to add new APIs without releasing to the AOSP(grapheneos.social ↗)
    396comments
  6. Typesafe-computer-use drives a Mac toward a goal for 1/50th of a cent per step(github.com/awlevin ↗)
    39comments
  7. Science Is Open Software(jepedersen.dk ↗)
    33comments
  8. SDCC – Small Device C Compiler(sourceforge.net ↗)
    16comments
  9. Cloudflare Quick Tunnels(cloudflare.com ↗)
    277comments
  10. You can run Git on object storage if you re-make packfiles(tigrisdata.com ↗)
    10comments
  11. Saving another 100TB of RAM(cloudflare.com ↗)
    68comments
  12. How to Write with an LLM(sockpuppet.org ↗)
    321comments
  13. Why building a Rust LSP is hard(rust-glancer.github.io ↗)
    30comments
  14. NASA-IBM Lunar Foundation open-Source Geospatial AI Model(usra.edu ↗)
    2comments
  15. Apple M6 Pro Achieves the Highest Single-Core CPU Score in Geekbench 7(geekbench.com ↗)
    3comments
  16. Ctenophores: Wonders of Biology(quantamagazine.org ↗)
    5comments
  17. How OpenAI Used Its Own LLMs to Design Its Jalapeño Chip(ieee.org ↗)
    83comments
  18. The first new cat species discovered in 100 years(nationalgeographic.com ↗)
    103comments
  19. Goroutine Leak Profiles(go.dev ↗)
    2comments
  20. OpenJev(openjev.com ↗)
    259comments
  21. Show HN: Cactus Needle 3: 8-29MB automation models can match DeepSeek V4 Flash(cactuscompute.com ↗)
    88comments
  22. Photon-Emission-Guided Laser Fault Injection Enables RP2350 Secure Debug(ledger.com ↗)
    67comments
  23. Suppress vulnerabilities applying Kubernetes context to scans(github.com/alegrey91 ↗)
    1comments
  24. Veronese's Dogs(publicdomainreview.org ↗)
    discuss
  25. Cache-to-Cache: Direct Semantic Communication Between LLMs (2025)(arxiv.org ↗)
    12comments
  26. Minimal Phone 2(minimalcompany.com ↗)
    218comments
  27. The Farnese letter(simonklee.dk ↗)
    6comments
  28. Stepfun Step 5 Preview (LLM): On AA Pareto frontier(artificialanalysis.ai ↗)
    2comments
  29. Warez: The Infrastructure and Aesthetics of Piracy (2021)(archive.org ↗)
    55comments
  30. Cyclomatic Complexity in C#(ndepend.com ↗)
    19comments

Containers and Virtual Machines at the Edge

93 pointsby 7y agoblog.stackpath.com
19 comments
7y agoHN ↗

It seems like their initial edge Engine uses the same approach as CF: plain V8 runtime.

This announcement to me sounds like the approach didn't get the expected adoption and now they're allowing regular containers in the edge.

7y agoHN ↗

We prefer both :)

We've seen amazing adoption on the EdgeEngine product, it just serves a different purpose than full on Containers and/or VM's at the Edge.

The ability to run a container/VM at the Edge enables the ability to run more complex workloads outside of scripting in a CDN like delivery pipeline.

7y agoHN ↗

Ben from the StackPath product team here. We agree containers or VM's may not always be the best available option for a workload. That's why we also offer an equivalent product to the one by Cloudflare (great product btw) - https://www.stackpath.com/services/edgeengine/

However, it should be made clear that our container and VM solution is not a "function" type offering. You can deploy a container and/or VM workload on our Edge similar to what you might find at Cloud providers. The main difference is we sit a layer above the cloud providers and make deploying worldwide simple, secure, and fast.

With a few clicks or a single API call you can deploy a micro-service all over the world (even add an anycast IP if you need one) in under 60 seconds.

7y agoHN ↗

None, it's StackPath under the hood. We have over 65Tbps of capacity all self built/managed at the major IX's in the world.

With a lot of our experience coming from the original founding team at Softlayer, we've built an incredible Platform.

More information on our network can be found at https://www.stackpath.com/platform/network/

7y agoHN ↗

Are you guys running Knative or similar to handle the serverless workloads in your edge ?

7y agoHN ↗

We are not. Knative is a great project and something we're definitely looking into for more "function" type workload offerings down the road.

Today our container/vm solution does not have a warmup concept other than the initial deployment of your container. You simply specify your image, some attributes, and it's deployed to the locations specified on our Edge. Once deployed, you have the ability to delete the workload at any time, but it's not elastic based on requests to the workload.

7y agoHN ↗

Cool, so what's your container/vm orchestration layer then?

7y agoHN ↗

Seems pretty pricey, smallest container is about $70/month for each region and each instance

7y agoHN ↗

Just saw the EdgeEngine post recently too, looks like some good competition to Cloudflare along with Fly.io.

The pricing is higher though and some early testing showed inconsistent performance, might have to wait until it matures a bit. Also missing IPv6.

7y agoHN ↗

Is there an open source alternative to this shared runtime? I can't quite recall seeing this process-level isolation in the FaaS sphere, maybe except for gVisor. Thanks.

7y agoHN ↗

Given my experiences with Stackpath's CDN, I wouldn't trust them with any sort of workload. They bill themselves as the "secure" CDN, yet were not validating SSL certs on the origin.

7y agoHN ↗

Sorry to hear about this experience, would love to know more as we take security very serious.

Our CDN certainly will validate SSL certificates at the origin if the setting is enabled.

With that said, you may be on a legacy CDN product often referred to as "Secure CDN" and that setting may differ from our current CDN offering.

I'd be more than happy to loop our CDN team in to clarify. Feel free to reach out directly to product at stackpath.com and I'll get it setup :)

7y agoHN ↗

Not really the "edge". Just a data center nearer the user, like Akamai.

Is it worth the hassle? All you're saving is latency in the backbone between the point of presence and your home data center. Does this beat directing requests to the nearest AWS region? As long as you're in roughly the right hemisphere and time zone, you're probably close enough.

7y agoHN ↗

Of course every "edge" is just another data center, but there's a massive rise in functions/serverless computing platforms focused on moving compute to more PoPs globally. Some have now graduated to running full containers as well.

If you have a stateless app, or state that can be deployed in multiple locations, then these are good solutions compared to doing it yourself and are relatively easier to deal with than the IaaS options of the clouds. Having it built into the CDN layer only helps with performance.

7y agoHN ↗

It probably doesn't matter if you're in the US or Western Europe, but could matter a lot in other regions. I don't know about Stackpath's PoP but Cloudflare and Akamai have PoP with much lower latency in many regions than the next data center by a cloud provider.

7y agoHN ↗

Edge can be defined as many things, and is currently being defined by the world in many ways.

At StackPath our definition of Edge is basically being as close as possible to the eyeballs aka users. Think of it as the front door to the internet.

Today our Edge expands across major IX's around the world and that's just the beginning. 5G is approaching us quickly along with container data centers.

The way we built our orchestration system it can deploy and manage workloads anywhere. In the future, that will include 5G container data centers which gets workloads even closer to things like self-driving cars, smart cities, IoT devices, <insert your idea here>.

7y agoHN ↗

"workloads even closer to things like self-driving cars, smart cities, IoT devices"

Oh, please. If you really need those last few milliseconds of lag reduced, you need local computation. If you don't, an AWS datacenter on the same continent is probably good enough.