[HN Gopher] The Fundamental Mechanism of Scaling
___________________________________________________________________
The Fundamental Mechanism of Scaling
Author : yagizdegirmenci
Score : 23 points
Date : 2021-07-17 11:20 UTC (1 days ago)
(HTM) web link (brooker.co.za)
(TXT) w3m dump (brooker.co.za)
| kureikain wrote:
| This is what I came to conslusion that a system without any
| dependencies can scale easiser.
|
| For my email forwarding app (https://hanami.run), our web tier
| can down, and even the database can down but the mail servers to
| handle incoming emails and forward out can tolerate database
| outage.
|
| For my particular use case, the mail servers boot then load
| entire state that they need into its cache, then write this to a
| local LevelDB. Incoming email is persisted into this LevelDB too.
| Then another thread will flush mail to Postgres if needed.
|
| By this simple design, we can even survive data center outage
| because the mail servers has all the state it need. Granted that
| during the outage, no one can signed up(web tier and control
| panel are down), but at least existing user won't be affected
| Jare wrote:
| > a system without any dependencies can scale easiser
|
| This is true for teams and projects too.
| michaelfeathers wrote:
| > The fundamental tool of cloud scaling is coordination
| avoidance.
|
| Having many systems rather than one is the ultimate scaling hack.
___________________________________________________________________
(page generated 2021-07-18 23:01 UTC)