5 ms·
It's really not that hard (or even expensive) to run your own. Here's[0] an example doing it on k8's. I had something similar running on GCP in a couple hours.
by mritchie712 5y ago
It's really not that hard (or even expensive) to run your own.
Here's[0] an example doing it on k8's. I had something similar running on GCP in a couple hours. It's been running for a month with no issues.
0 - https://messari.io/article/running-an-ethereum-node-on-kubernetes-is-easy https://messari.io/article/running-an-ethereum-node-on-kuber...
- sv123 5y agoBut one of the main points of article is that people don't want to run servers, developers included. Even being easy, letting someone else do it will always be easier.
- throw_nbvc1234 5y agoHow many (large) companies, governments, etc... run their own email servers? If there's a strong enough need, people will run their own servers even if they'd rather not. "people don't want to run servers" arguably could be rephrased as "people don't have a reason (today) to run their own servers". I'd argue this is a key difference between web1 and cryto centralization and the web2 centralization. If Google announced tomorrow that anyone can buy the gmail contents of any gmail address, you'd bet a lot more individuals would either switch to alternatives or start running their own severs.
- threeseed 5y ago> How many (large) companies, governments, etc... run their own email servers Every year a decreasing number as everything moves to SaaS and the cloud.
- tsimionescu 5y ago> How many (large) companies, governments, etc... run their own email servers? Office 365 financials alone suggest that the answer is "very few, and rapidly decreasing". I work for a ~30k employee technology company that doesn't run it's own email servers.
- mattl 5y agoShould be pretty easy to find the top 100/1000/10000 companies and look at their MX records. I’d imagine it’s a large number of Office 365, GSuite by Google and Barracuda/ProofPoint which may point to a SaaS thing or an internal server.
- dcolkitt 5y agoBut the question was how hard is it to run a competitor to Infura. And the answer is trivially easy. Infura is just an Ethereum node API that's publicly exposed. Building an Infura competitor literally is nothing more than $100/month it costs to run a Geth node on AWS.
- matthewdgreen 5y agoThis is true today. But the standard approach in this industry is to start by offering access to an open service and then quickly build in value-add services that aren’t available in the open service. So for example, the smart move would be for Infura to offer a proprietary chain or rollup that gets widely used but isn’t available outside of Infura. If they can pull that off, competition could get much harder.
- des1nderlase 5y agoI second this. If history has thought us anything is that every web3 company will work toward increasing the competitive gap.
- mritchie712 5y agoRight, this was my point. People don't usually run Postgres themselves (e.g. set up Postgres in a docker container), but it's not very hard to do. The article makes it sound like Infura has a moat. There's no moat, it's as easy to switch as it is to switch Postgres clouds. To be clear, I agree with most of their findings, this on is just a bit off.
- biorach 5y ago> People don't usually run Postgres themselves (e.g. set up Postgres in a docker container), but it's not very hard to do. It's easy to do a basic install. It's quite hard to do it right, at scale, with workload-appropriate configuration, replication, backup etc. My point... neither Postures nor Indira, or any other blockchain solution are easy to install and maintain in a fully scaled-up, fault-tolerant, multi-node deployment
- wbl 5y agoAs someone who has run nodes, no it is hard and expensive. Every time a geth node dies it has to resync and no persistent volume mounts and stateful sets are not solutions. They are problems. If you need to scale horizontally you get strange consistency issues with the API. All of this makes for a very unpleasant experience. It's built for TLC on a beefy box not a herd.
- mritchie712 5y agoWhat version of geth were you using? How many CPUs? When one of my geth nodes dies, another spawn without issue.
- wbl 5y agoAnd that's the rub. The new node doesn't have the same state as the old one. So clients making requests assuming that latest is the same start having problems. If you haven't seen them you just haven't been running a production quality service.
- throaway46546 5y agoIn a discussion about people not wanting to run their own servers the fact that your first instinct was to use GCP is telling.