5 ms·
So what ? It's not the first time it happened with Coinbase, and probably not the last
by maxencecornet 9y ago
So what ? It's not the first time it happened with Coinbase, and probably not the last
- astrowilliam 9y agoThat's what happens when a business is trying to scale to meet crazy demand.
- scurvy 9y agoCan't they just click the auto scale button in AWS? Isn't that the entire point of cloud services?
- s0rce 9y agoProbably not a infrastructure capacity issue but a limited amount of cryptocurrency in a hot wallet for transactions.
- ihattendorf 9y agoIf that were the case it would be withdrawals that were temporarily disabled. There aren't any on-chain transactions made when you buy/sell on Coinbase, only when you deposit/withdraw.
- dragontamer 9y agoCold wallets don't scale, and that's the entire point of them. "Hot Wallets" can be hacked, because they're connected to the whole system. "Cold Wallets" are offline, and require human intervention to transfer money into and/or out of. Figuring how much money to keep in the "Hot Wallet" is a tradeoff in services, convenience, and security.
- scurvy 9y agoThis doesn't sound like a wallet problem. That would only break withdrawals. They're not doing any transactions at all.
- wmichelin 9y agoCan't tell if you're trolling or not... ;p
- SilasX 9y agoThat works, so long as you've refactored your infra to the point that the computational resources (whether server nodes or hard disks or or database shards) are modular enough that you can just add more stuff and it will play nice with everything else. That objective -- not a cute UI that lets you change one of the numbers -- is the hard part. And yes, it is to Coinbase's discredit that they haven't done so, but (assuming you're not joking), they could very well not be at that stage. It's typical for startups to have technical debt like that.
- scurvy 9y agoCan't they just get a bigger boat? Scaling via hardware is a good fix until systems can be made modular for scaling.
- cwkoss 9y agoMost scalable systems have some 'singletons' somewhere in the stack. At a certain scaling point, the singletons must be made scalable to multiple instances, which often entails making a new singleton to manage these instances. In this way, scaling often gives you an order of magnitude or two of available additional capacity, but periodically requires new dev work to support further increases.
- SilasX 9y agoIt's what happens when some businesses fail to act fast enough and use their loads of money to buy actual expertise. They had four outages in June. That's plenty of warning.