6 ms·

Amazon MemoryDB for Redis – A Redis-Compatible, Durable Database

53 points by NathanKP 5y ago

skyde 5y ago
When they say "Redis compatible" I wonder if this include Redis keyspace notifications, Redis Pub/Sub and server assisted client side caching. Those are the only reason I am using Redis because not other database easily let you subscribe to notification on change to a specific set of keys.
- bdcravens 5y ago
  In the launch podcast, the question was asked about specific features. They didn't mention the ones you listed, but the response was that it was 100% compatible with v6, so I suspect so. https://aws.amazon.com/podcasts/467-introducing-amazon-memorydb-for-redis/ https://aws.amazon.com/podcasts/467-introducing-amazon-memor...
pull_my_finger 5y ago
Is this the project that triggered Antirez to change the license on Redis?
- _msw_ 5y ago
  Disclosure: I work for AWS where I build infrastructure services The license to Redis never changed. See his blog post [1] around the time when his _employer_ announced some license changes on modular extensions to Redis. See also threads on HN [2]. [1] http://antirez.com/news/120 http://antirez.com/news/120 [2] https://news.ycombinator.com/item?id=17818923 https://news.ycombinator.com/item?id=17818923
  - sitkack 5y ago
    When will it pass Jepsen?
kevinburke 5y ago
How is this better than Hadoop?
- PeterCorless 5y ago
  Redis and workalikes are in-memory key-value datastores, more appropriate for real-time transactional workloads. They serve entirely different purposes than Hadoop, which is a data analytics platform. Hadoop is targeted towards large "slow" data processing jobs, using a function called "map reduce." Redis and its workalikes are targeted to quick atomic CRUD operations (create, reads, updates and deletes). It's like asking what's the difference between a truck and a Ferrari. They both have engines and steering wheels, right? But totally different workloads.
- p0rkbelly 5y ago
  How is Redis different than Hadoop?
jonathan_olson 5y ago
This is pretty cool. I'm hoping to see a stress test write up.
seg_lol 5y ago
I'd much rather use OSS like Aerospike or ScyllaDB for this workload and still retain some agency over my deployment. Getting a system like this right is super hard, and not being OSS, you are at the mercy of the cloud provider to fix issues. Aerospike recently released their own Redis protocol proxy for Redis [1]. Scylla has their own nascent Redis api driver [2], would love to get folks feedback on it. [1] https://github.com/aerospike/skyhook https://github.com/aerospike/skyhook [2] https://siddharthc.medium.com/redis-on-nvme-with-scylladb-5e12afd38dbc https://siddharthc.medium.com/redis-on-nvme-with-scylladb-5e...
- tomnipotent 5y ago
  I'm really impressed with Seastar [1], the library ScyllaDB wrote to solve for writing non-blocking, asynchronous code in C++. It's also used by Redpanda [2], a Kafka-compatible service in the same vein as MemoryDB/Skyhook. [1] https://github.com/scylladb/seastar https://github.com/scylladb/seastar [2] https://github.com/vectorizedio/redpanda https://github.com/vectorizedio/redpanda
  - seg_lol 5y ago
    The arch of ScyllaDB blew my mind gotta say. Great talk [1] by Avi Kivity, co-founder of Scylla that goes over the design and architecture. Glommio [2] (believe it used to be called Scipio) is a thread-per-core framework for Rust modeled after Seastar. There isn't much in the hn hive mind about it, this one [3] stands out. [1] https://www.youtube.com/watch?v=0S6i9BmuF8U https://www.youtube.com/watch?v=0S6i9BmuF8U [2] https://github.com/DataDog/glommio https://github.com/DataDog/glommio [3] https://news.ycombinator.com/item?id=24976533 https://news.ycombinator.com/item?id=24976533
    - tomnipotent 5y ago
      Great recommendation, thank you.

53 points by NathanKP 5y ago

When they say "Redis compatible" I wonder if this include Redis keyspace notifications, Redis Pub/Sub and server assisted client side caching. Those are the only reason I am using Redis because not other database easily let you subscribe to notification on change to a specific set of keys.

bdcravens 5y ago
In the launch podcast, the question was asked about specific features. They didn't mention the ones you listed, but the response was that it was 100% compatible with v6, so I suspect so. https://aws.amazon.com/podcasts/467-introducing-amazon-memorydb-for-redis/ https://aws.amazon.com/podcasts/467-introducing-amazon-memor...

pull_my_finger 5y ago

Is this the project that triggered Antirez to change the license on Redis?

_msw_ 5y ago
Disclosure: I work for AWS where I build infrastructure services The license to Redis never changed. See his blog post [1] around the time when his _employer_ announced some license changes on modular extensions to Redis. See also threads on HN [2]. [1] http://antirez.com/news/120 http://antirez.com/news/120 [2] https://news.ycombinator.com/item?id=17818923 https://news.ycombinator.com/item?id=17818923
- sitkack 5y ago
  When will it pass Jepsen?

kevinburke 5y ago

How is this better than Hadoop?

PeterCorless 5y ago
Redis and workalikes are in-memory key-value datastores, more appropriate for real-time transactional workloads. They serve entirely different purposes than Hadoop, which is a data analytics platform. Hadoop is targeted towards large "slow" data processing jobs, using a function called "map reduce." Redis and its workalikes are targeted to quick atomic CRUD operations (create, reads, updates and deletes). It's like asking what's the difference between a truck and a Ferrari. They both have engines and steering wheels, right? But totally different workloads.
p0rkbelly 5y ago
How is Redis different than Hadoop?

jonathan_olson 5y ago

This is pretty cool. I'm hoping to see a stress test write up.

seg_lol 5y ago

I'd much rather use OSS like Aerospike or ScyllaDB for this workload and still retain some agency over my deployment. Getting a system like this right is super hard, and not being OSS, you are at the mercy of the cloud provider to fix issues. Aerospike recently released their own Redis protocol proxy for Redis [1]. Scylla has their own nascent Redis api driver [2], would love to get folks feedback on it. [1] https://github.com/aerospike/skyhook https://github.com/aerospike/skyhook [2] https://siddharthc.medium.com/redis-on-nvme-with-scylladb-5e12afd38dbc https://siddharthc.medium.com/redis-on-nvme-with-scylladb-5e...

tomnipotent 5y ago
I'm really impressed with Seastar [1], the library ScyllaDB wrote to solve for writing non-blocking, asynchronous code in C++. It's also used by Redpanda [2], a Kafka-compatible service in the same vein as MemoryDB/Skyhook. [1] https://github.com/scylladb/seastar https://github.com/scylladb/seastar [2] https://github.com/vectorizedio/redpanda https://github.com/vectorizedio/redpanda
- seg_lol 5y ago
  The arch of ScyllaDB blew my mind gotta say. Great talk [1] by Avi Kivity, co-founder of Scylla that goes over the design and architecture. Glommio [2] (believe it used to be called Scipio) is a thread-per-core framework for Rust modeled after Seastar. There isn't much in the hn hive mind about it, this one [3] stands out. [1] https://www.youtube.com/watch?v=0S6i9BmuF8U https://www.youtube.com/watch?v=0S6i9BmuF8U [2] https://github.com/DataDog/glommio https://github.com/DataDog/glommio [3] https://news.ycombinator.com/item?id=24976533 https://news.ycombinator.com/item?id=24976533
  - tomnipotent 5y ago
    Great recommendation, thank you.