It’s Christmas Day Observed, so no curation today. Instead, here’s a classic paper which has shaped the industry: Amazon’s Dynamo paper. Abstract:
Reliability at massive scale is one of the biggest challenges we face at Amazon.com, one of the largest e-commerce operations in the world; even the slightest outage has significant financial consequences and impacts customer trust. The Amazon.com platform, which provides services for many web sites worldwide, is implemented on top of an infrastructure of tens of thousands of servers and network components located in many datacenters around the world. At this scale, small and large components fail continuously and the way persistent state is managed in the face of these failures drives the reliability and scalability of the software systems.
This paper presents the design and implementation of Dynamo, a highly available key-value storage system that some of Amazon’s core services use to provide an “always-on” experience. To achieve this level of availability, Dynamo sacrifices consistency under certain failure scenarios. It makes extensive use of object versioning and application-assisted conflict resolution in a manner that provides a novel interface for developers to use.
This paper shaped Amazon’s Dynamo system, their DynamoDB offering on AWS, and a host of competitors, including Riak and Cassandra. That makes it worth reading.
Modeled using principles from the “AWS Dynamo” paper, Riak KV buckets are good for scenarios which require frequent, small data-sized operations in near real-time, especially workloads with reads, writes, and updates — something which might cause data corruption in some distributed databases or bring them to “crawl” under bigger workloads. In Riak, each data item is replicated on several nodes, which allows the database to process a huge number of operations with very low latency while having unique anti-corruption and conflict-resolution mechanisms. However, integration with Apache Spark requires a very different mode of operation — extracting large amounts of data in bulk, so that Spark can do its “magic” in memory over the whole data set. One approach to solve this challenge is to create a myriad of Spark workers, each asking for several data items. This approach works well with Riak, but it creates unacceptable overhead on the Spark side.
This is interesting in that it ties together two data platforms whose strengths are almost the opposite: one is great for fast, small writes of single records and the other is great for operating on large batches of data.