Rockset for Real-Time Analytics at Snowflake Summit
10x Faster Queries
50% Lower Compute Cost
100% Guaranteed Fresh Data
Tackling the Challenges
Challenge #1:
Compute costs are rapidly growing
Snowflake is not compute optimized
Snowflake organizes data into its compressed, columnar format. This is great for minimizing storage footprint and budget-friendly for analysts running occasional queries on batch data. However, querying data stored in columnar format requires computationally intensive scans, making it too expensive to run sub-second queries on fresh data.
Rockset is compute optimized
Rockset indexes all fields, including nested fields, in a Converged Index, which combines an inverted index, a columnar index and a row index. This translates to a slightly bigger storage footprint in exchange for faster queries, lower data latency, and less compute costs.
Challenge #2:
Query speed is too slow
Snowflake does full scans
Snowflake has to scan through large portions of data to run each query, which means queries can take tens of seconds to run, especially as data size or query complexity grows. And each query requires a minimum of 100s of milliseconds of start-up time. Some try to accelerate performance by adding more costly compute, but even then, hit an upper bound for performance and cannot increase query speeds for true real-time analytics.
Rockset uses indexing to minimize scans
Rockset’s cost-based query optimizer leverages our Converged Index to automatically find the most efficient way to run low latency queries by exploiting selective query patterns within the indexed data and accelerating aggregations over large numbers of records. Rockset does not scan any faster than a cloud data warehouse. It simply tries really hard to avoid full scans altogether.
Challenge #3:
Data latency is too high
Snowflake data is stale
Snowflake loads data in batches to minimize compute processing, resulting in a delay before new data can be queried. Some data warehouses try to reduce this latency by continuously loading small data batches, such as Snowpipe on Snowflake. However, though continuous, these solutions are both not real-time, as data might not be available for querying for many minutes, and incredibly expensive to run. This can be compounded by throughput constraints as the writes queue up if too much data is pushed through at one time.
Rockset makes data queryable within a second
Rockset has built-in real-time data connectors that guarantees data freshness, which no data warehouse has. By using RocksDB LSM trees and a lockless protocol, Rockset enables writes to be visible to existing queries within a second of data being generated. In addition, Rockset separates compute needed for indexing from compute needed for queries to deal with bursty writes.
As you modernize your data stack to build more data applications, use Rockset to increase analytics speed and decrease costs.
Here are four reasons to use Rockset for real-time analytics:
Reduce compute costs by 50%
Increase query speeds by 10x
Reduce data latency to one second
100% serverless and built in the cloud
See Why Companies Choose Rockset for Real-Time Analytics
Modern companies are building real-time logistics tracking, security analytics, predictive maintenance and more in record time.
See Our Customers
Ritual uses Snowflake for ad-hoc analysis, periodic reporting and machine learning model creation, but knew that Snowflake would not meet their sub-second latency requirements for personalization and looked to Rockset as a potential speed layer.
Learn more
More from Rockset
Compare Rockset and Snowflake
Connect with our solutions team to dig deeper into the architecture, indexing, data ingestion and query processing.
Let's Talk