On Fri, Jul 19, 2024 at 9:15 AM Phil Endecott via Boost < boost@lists.boost.org> wrote:

I worry that you have over-engineered this in a way that has made it much more expensive than it needs to be.

That may be true, and our principle for solving these problems is to first optimize for getting something in place quickly, with the highest possible level of service, with cost as a secondary consideration. And then once things are stable we look at other available, less expensive solutions. This lets us evaluate alternatives for reducing cost, at our own pace and after other more pressing matters are resolved, without pressure. Thanks

Phil Endecott via Boost writes:

Louis Tatta wrote:

...

Server Hosting $23,000 [per quarter]

That's a lot of money.

Indeed. For perspective, for $7K/month you could rent about 45 Xeon Gold 24-core servers on Hetzner. So over 1000 cores of compute or about 7 dedicated cores per Boost library. Imagine the CI system you could build from that. (I appreciate that managing those 45 servers will cost something, potentially even more than the servers themselves. Still, for something like CI which doesn't have strict reliability/availability requirements, a lot of this management cost can be pushed upfront and made fixed.)

It's really none of my business how you spend your money, [...]

I had a related question for while now so while at it, let me ask it: Where does this money come from and how long will it last? I looked around the C++ Alliance website for the answer but all I could find was that "The Alliance is currently funded by a private endowment". Are you (as in, Boost developers) not concerned where this non-trivial amounts of money come from? Equally important, are you not concerned about what happens if/when it runs out? For example, there seem to be multiple, full-time developers working on the new Boost website. Based on that I think it's reasonable to assume that the result will require a non-trivial amount of ongoing maintenance. What happens when/if there is no more funding for this maintenance? Are the Boost developers prepared to step in and do web development? I realize am asking pointed questions while not really being part of the project (other than packaging Boost for build2), so if the answer is "this is none of your business", then that's fair enough.

you could rent about 45 Xeon Gold 24-core servers on Hetzner.

Hi Boris, The $7k was affected by a one-time charge on AWS Cloudfront that will be reduced in future quarters. Much of the current billing now is CI. Hetzner is a great hosting company. We have an account there, and this is an interesting idea. Still, analyzing the number (45 servers), cut that by 50% or more, since the expected billing amount is less. And it's going towards other costs also. Let's say 23 Hetzner servers. The CI jobs we're hosting burst up to 500 simultaneous jobs. Is it enough? The solution is autoscaling (on AWS). I explored autoscaling on Hetzner also. The problem is that you want to create a disk image (an AMI) with all the preinstalled CI software, and be able to quickly launch and destroy the instances, using the custom AMI. However, Hetzner allows that to happen on a restricted number of officially provided AMIs, and then when it comes to a customized image, that must be transferred across their network. It was taking more than 10 minutes. Not an acceptable delay. A CI job should be launching instances in 1-2 minutes. I confirmed the delay issue with their tech support. No fix. Nevertheless, in spite of that, I do occasionally think the same thought as you: "Let's launch 100 dedicated machines on Hetzner". Maybe at some point, yes.

On Sat, Jul 20, 2024 at 12:50 AM Boris Kolpackov via Boost < boost@lists.boost.org> wrote:

I had a related question for while now so while at it, let me ask it: Where does this money come from and how long will it last? I looked around the C++ Alliance website for the answer but all I could find was that "The Alliance is currently funded by a private endowment".

The Alliance funding could last many decades. Long enough, that we need a plan for transitioning the organization to others in the case of the loss of its principals. This is not such an easy problem to be solved as my network of contacts are around my age, and ideally the executor of any legal entity which would outlast the current principals should be significantly younger. Happy for any volunteers to reach out to me at vinnie@cppalliance.org. Are you (as in, Boost developers) not concerned where this non-trivial

amounts of money come from? Equally important, are you not concerned about what happens if/when it runs out?

This is a concern, yes. Note however that this is not a new concern as Boost already has exposure to the loss of volunteers and resources: * When an author abandons a library * When a maintainer stops working on a library * When a release managers retiresn * When Boost infrastructure is not maintained * When the custodian of Boost's shared resources turns its attention to other projects These problems have existed and will continue to exist for the lifetime of the project. The solution will be the same as it has always been. New volunteers and contributions are necessary. there seem to

be multiple, full-time developers working on the new Boost website. Based on that I think it's reasonable to assume that the result will require a non-trivial amount of ongoing maintenance. What happens when/if there is no more funding for this maintenance? Are the Boost developers prepared to step in and do web development?

The new website is an ambitious project which will require non-trivial resources over the next few years to implement, as we have many interesting features planned with the hopes of increasing qualified participation in the project. Once the site matures, the ongoing maintenance will be considerably less. We chose Python/Django for its enormous popularity and for the ease of which it is to find talented individuals who can perform the work, whether volunteer or paid. An emphasis was placed on using existing technologies that follow best practices. Boost's existing website, running on the "wowbagger" server, is already an existential risk, and essentially is unfixable. See: https://cppalliance.org/pdf/Boost-Server-Report.pdf What I am saying is that while the hypothetical problem of C++ Alliance lacking funds may be a problem in the future, the current website's lack of resources is a real problem today. My usage of the term website also encompasses the mailing lists, which run on the same machine and get served from the same domain. To mitigate the risk of short-term problems with the Alliance, we have deployed a static copy of the old website which is very cheap to maintain and runs on modern software and services. This will be available at a separate URL even with the new website up, allowing a grace period where folks can experience the old site for whatever reason. Including the case where we forgot a page or some information, and we need to refer to the old site to update the new one. This static duplicate of the current boost.org also serves another function. It is a safety mechanism if the new website goes lights-out for any reason. This way the old site can be pressed back into service. It is true that the community will suffer a loss of aesthetics and functionality but this is not much different than when someone abandons a library. Boost can and will deal with such eventualities as it always does. I realize am asking pointed questions while not really being part

of the project (other than packaging Boost for build2), so if the answer is "this is none of your business", then that's fair enough.

I think these are fair questions. And for the health of Boost I think more people should be asking these questions, and related questions. Of course, volunteering time and resources would be better still :) Thanks