talawahtech

8 ms·

1. 31.▲
   by talawahtech 5y ago

   Thanks, that actually means a lot. It took a lot of work, not just on the server/code, but also the writing. I asked a lot of people to review it (some multiple times) and made a ton of changes/edits over the last couple months. T
2. 32.▲
   by talawahtech 5y ago

   It is a statically generated site created with vitepress[1] and hosted on Cloudflare Pages[2]. The only dynamic functionality is the contact form which sends a JSON request to a Cloudflare Worker[3], which in turn dispatches the message to
3. 33.▲
   by talawahtech 5y ago

   This comprehensive level of extreme tuning is not going to be directly useful to most people; but there are a few things in there like SO_ATTACH_REUSEPORT_CBPF that I hope to see more servers and frameworks adopt. Similarly I think it is
4. 34.▲
   by talawahtech 5y ago

   There is no published limit, but based on my tests the network device for the c5n.xlarge has a hard limit of 1.8M pps (which translates directly to req/s for small requests without pipelining). There is also a quota system in place, so
5. 35.▲
   by talawahtech 5y ago

   Yea, I looked it wrk2 but it was a no-go right out the gate. From what I recall the changes to handle coordinated omission use a timer that has a 1ms resolution. So basically things broke immediately because all requests were under 1ms.
6. 36.▲
   by talawahtech 5y ago

   Thanks! > I always disable C-states deeper than C1E AWS doesn't let you mess with c-states for instances smaller than a c5.9xlarge[1]. I did actually test it out on a 9xlarge just for kicks, but it didn't make a difference. Onc
7. 37.▲
   by talawahtech 5y ago

   Thanks! The c5 instance type is about 10-15% faster than the m5, but the m5 has twice as much memory. So if memory is not a concern then switching to c5 is both a little cheaper and a little faster. You shouldn't need the c5n, the regu
8. 38.▲
   Extreme HTTP Performance Tuning(talawah.io)

   976 points by talawahtech 5y ago|145 comments
9. 39.▲
   by talawahtech 5y ago

   Interesting, seems to be the AWS equivalent of Cloudflare Workers, though with much stricter restrictions. Max execution time: 1ms, Max memory: 2MB. On the surface they are undercutting Cloudflare in terms of price ($0.10 vs $0.15 per milli
10. 40.▲
    by talawahtech 6y ago

    Check out Cloudflare Pages: https://pages.cloudflare.com/
11. 41.▲
    by talawahtech 6y ago

    All the 16 vCPU "n" instances (m5n, c5n, r5n, etc) are capable of hitting the 25 Gbps limit easily. In your report, all of the AWS results are limited to either 5Gbps or 10Gbps, but this is because of a very specific test conditio
12. 42.▲
    by talawahtech 6y ago

    The network throughput benchmark is pretty misleading. https://news.ycombinator.com/item?id=25816596
13. 43.▲
    by talawahtech 6y ago

    There are number of issues with this report. The AWS networking section is particularly problematic and in need of extensive disclaimers or changes to the test methodology. On the throughput side, all this test does is demonstrate the docum
14. 44.▲
    by talawahtech 6y ago

    I suspect that RedHat is actually planning on making a no-cost version of RHEL available without support. I think they are still in the process of deciding exactly how to pitch it without cannibalizing existing sales. They are using this fa
15. 45.▲
    by talawahtech 6y ago

    Another useful data point that will probably help force RedHat's hand if/when they decide to offer their own no-cost version of RHEL.
16. 46.▲
    by talawahtech 6y ago

    Thx, I'll check it out!
17. 47.▲
    by talawahtech 6y ago

    Perfect timing, I was planning on launching a new personal blog next week. Was gonna use GitHub Pages fronted by Cloudflare but this is even better since it has preview and password capabilities. If anybody from Cloudflare is listening I wo
18. 48.▲
    by talawahtech 6y ago

    I am looking forward to it because it is an LTS release that has all the latest io_uring enhancements. io_uring looks like it will be a big deal, it will probably be a while before it really hits mainstream, but I was waiting on 5.10 before
19. 49.▲
    by talawahtech 6y ago

    FYI 5.10 will be the next LTS kernel. https://twitter.com/gregkh/status/1320745076566433793?s=19
20. 50.▲
    by talawahtech 6y ago

    Another relevant point made in the video is that they restrict cells to a maximum size which then makes it easier to test behavior at that size. This would have also helped avoid this specific issue since the number of threads would have be
21. 51.▲
    by talawahtech 6y ago

    Multi-AZ doesn't protect against a software/OS issue like this, Multi-AZ would be relevant if it was an infrastructure failure (e.g. underlying EC2 instances or networking). The relevant resiliency pattern in this case would be wh
22. 52.▲
    by talawahtech 6y ago

    Thanks! That is exactly what I was looking for.
23. 53.▲
    by talawahtech 6y ago

    What is used to secure/encrypt the connection between the clients and the workers? I did a quick search in the GitHub repo for WireGuard and didn't get any results so I guess you aren't using it.
24. 54.▲
    by talawahtech 6y ago

    Looking at the screenshot of the IG story, it indicates that the compromised account is now followed by @blood, @murder and @dead. Looks like whoever took over the account is not some random kid with a friend at Facebook, but a "profes
25. 55.▲
    by talawahtech 6y ago

    This is awesome and long overdue. Providing a simple way to go directly from an HTTP endpoint to SQS, Event Bridge or Kinesis makes it much easier to build more robust webhooks or tracking endpoints. I have tried doing similar things in the
26. 56.▲
    by talawahtech 6y ago

    AWS has non-virtualized "metal" instances as well. Customer may be be running container workloads on those instances with SMT disabled.
27. 57.▲
    by talawahtech 6y ago

    AWS has non-virtualized "metal" instances as well. Customer may be be running container workloads on those instances with SMT disabled.
28. 58.▲
    by talawahtech 7y ago

    Ok now Microsoft just needs to acquire what remains of Docker and their Developers, Developers, Developers, Developers collection will be complete.
29. 59.▲
    by talawahtech 7y ago

    I kinda wish Microsoft would just acquire the rest of the company and be done with it. It fits in very well with their Developers, Developers, Developers, Developers focus. Especially in the era of Open Source friendly MS. VSCode, GutHub an
30. 60.▲
    AWS Launches G4 Instances with Nvidia T4 Tensor Core GPUs(aws.amazon.com)

    3 points by talawahtech 7y ago|0 comments