What part of the world has any ipv6 limitations? In the USA An ISP will give you a /48 from their /32 if you have any colo arrangement without even a blink. That gives you 2^16 networks with essentially infinite number of hosts on each network. Zero additional charge.
About 20 years ago I was working in IT, and I was responsible for the physical build out (from a shell) of 599 North Mathilda, Sunnyvale - The foundations and outer walls/roof were up - but nothing inside.
I had responsibility (working with consultants who did this for a living) to work with the project manager, and architect/GC. All of the datacenters (back when companies put data-centers in their buildings), IDFs, MDFs. The MDF in particular was complex as it combined the floors IDF + the buildings MDF/telco connections, punchdowns, and a massive Nortel Option51c set of cabinets. We carefully laid out the room - measuring the minimal possible distance for cable techs to get in between the racks. Everything was designed down to the 1/4" in the room.
I showed up (mostly randomly) with a tape measure during construction - internal walls were up - and they were off by almost 14" - which would have made the internals almost unusable for their original purpose. They had to tear down their framing, pull everything out - thankfully before any electrics/racks/hvac had been put in place.
Having something like this would have greatly reduced that possibility. Bet they end up on every site (if they aren't already).
Right, but depending on your workload, compute might just be 1/3 to 1/2 of your spend. The remainder going on storage, networking (egress and internal between regions & AZs), LBs, and higher abstraction services (from queues to search to serverless).
Feels great to talk about 27-50% but turns out it's 9%-16% when all is said and done. You can get commitment savings on other services but you need higher spend.
Feels odd that big cloud gives better discounts to enterprise. They really don't cater to startups as much as they posture.
There are a ton of different discount options - large customers typically get between 50-60% discount based on committed spending, and AWS is pretty flexible around how that commit lands (they will allow roll overs even if they say they won't). Reserved instances get you ~70% discounts - similar to the committed spending. And my favorite - if it works for you - spot instances on EC2 come at as high as 90% off.
Nobody at commercial volume pays list to AWS - everyone gets a discount.
You can sometimes get a committed use discount within certain regions. Not as extreme as the EC2 discount, since S3 storage costs are honestly pretty low when you use storage classes correctly.
I'm intrigued. I've been using virtualenv in numerous companies for about 8 years, traditionally wrapped in virtualenvwrappers, and now in uv.
UV doesn't change any of that for me - it just wraps virtualenv and pip downloads dependencies (much, much) more quickly - the conversion was immediate and required zero changes.
UV is a pip / virtualenv wrapper. And It's a phenomenal wrapper - absolutely changed everything about how I do development - but under the hood it's still just virtualenv + pip - nothing changed there.
Can you expand on the pain you've experienced?
Regarding "things that need to be deployed" - internally all our repos have standardized on direnv (and in some really advanced environments, nix + direnv, but direnv alone does the trick 90% of the time) - so you just "cd <somedir>", direnv executes your virtualenv and you are good to go. UV takes care of the pip work.
Has eliminated 100% use of virtualenvwrappers and direct-calls to pip. I'd love to hear a use case where that doesn't work for you - we haven't tripped across it recently.
Not quite; it reimplements the pip functionality (in a much smarter way). I'm pretty sure it reimplements the venv functionality, too, although I'm not entirely sure why (there's not a lot of room for improvement).
("venv" is short for "virtual environment", but "virtualenv" is specifically a heavyweight Python package for creating them with much more flexibility than the standard library option. Although the main thing making it "heavyweight" is that it vendors wheels for pip and Setuptools — possibly multiple of each.)
Nowadays most cloud customers jam their racks full of absolutely vanilla (brand is mostly irrelevant) 128 vCores / 64 physical cores, 512 GByte servers for ~$18k - w/100 GBit NICs. That Sun 4500, maxed out (with 10 Gbit/Nics), sold for $70k. ($110K in 2025 dollars).
What's (still) super impressive was the 48 drives. Looking around -the common "Storage" nodes in rack these days seem to be 24x24TB CMR HDD + 2 7.68 TB NVME SDD (and a 960 GB Boot Disk) - I don't know if anyone really uses 48 drive systems commonly (outside the edge cases like Backblaze and friends)
Antonio Meucci invented the Teletrofono around 1849 and filed a patent for it in 1871. I know this mostly because it was a big deal in a Soprano's episode.
uv is venv + insanely fast pip. I’ve used it every day for 5+ months and I still stare in amazement every time I use it. It’s probably the most joy I’ve ever gotten out of technology.
There are roughly 20-25 major IaaS providers in the world that should have close to dependency on each other. I'm almost certain that cloud flare believe that was their posture, and that the action items coming out of this post mortem will be to make sure that this is the case.
reply