4 ms·

I automated my work as a distributed systems consultant by inventing new theory

I've built software which can provide distributed global ordering and mutexes over even geographically distributed kafka topics. I believe I can bring together the monolith-people with the microservice-people by reconciling the benefits of both approaches. This is an "easy to access" article on my project: https://andrasgerlits.medium.com/distributing-consistency-6a1d11edc1cd This is one on CAP Theorem and how it applies to us https://andrasgerlits.medium.com/cap-and-us-9e0734cbfe1e AMA

Andys 5y ago
Thanks for posting. Does this new database support SQL? I saw a few mentions but not sure if it was explicitly stated It sounds like the bottleneck is going to be the transaction controller in each node, which has to serialize the processing of transactions to ensure they don't clobber each other. I guess that could still scale pretty well as long as you can scale out the number of nodes?
- andras_gerlits 5y ago
  Yes, it does. Well, it's not a database, but it does :). The original whitepaper describes each mechanism through which different aspects of the SQL standard can be implemented, it's linked on the bottom of our website and from the article. http://dianemodb.com http://dianemodb.com
- andras_gerlits 5y ago
  Also, you can scale the number of transaction-handling nodes to your hearts content. Besides, it supports fully consistent follower-reads by following its own definition of time and potentially skewing the client API's transaction time-window (thereby fitting even strict consistency). I also talked about this at a conference https://www.youtube.com/watch?v=uCu27LEW2UU https://www.youtube.com/watch?v=uCu27LEW2UU
- andras_gerlits 5y ago
  Also, this isn't entirely true. The transactions can clobber each other and they don't need to be serialized by the transaction-coordinator, but each node can only be single-threaded. The presentation explains most of it, the whitepaper explains everything, if you're curious.

Andys 5y ago

Thanks for posting. Does this new database support SQL? I saw a few mentions but not sure if it was explicitly stated It sounds like the bottleneck is going to be the transaction controller in each node, which has to serialize the processing of transactions to ensure they don't clobber each other. I guess that could still scale pretty well as long as you can scale out the number of nodes?

andras_gerlits 5y ago
Yes, it does. Well, it's not a database, but it does :). The original whitepaper describes each mechanism through which different aspects of the SQL standard can be implemented, it's linked on the bottom of our website and from the article. http://dianemodb.com http://dianemodb.com
andras_gerlits 5y ago
Also, you can scale the number of transaction-handling nodes to your hearts content. Besides, it supports fully consistent follower-reads by following its own definition of time and potentially skewing the client API's transaction time-window (thereby fitting even strict consistency). I also talked about this at a conference https://www.youtube.com/watch?v=uCu27LEW2UU https://www.youtube.com/watch?v=uCu27LEW2UU
andras_gerlits 5y ago
Also, this isn't entirely true. The transactions can clobber each other and they don't need to be serialized by the transaction-coordinator, but each node can only be single-threaded. The presentation explains most of it, the whitepaper explains everything, if you're curious.