DataStax Enterprise (DSE) already offers a plethora of solid capabilities to make your distributed database dreams become more real than The NeverEnding Story. But are you aware of all of the crazy, quality of life updates and new features added in DataStax Enterprise 6? These include: significantly improved performance; anti-entropy improvements with DSE NodeSync; quality updates for DSE Search, Graph, Analytics, OpsCenter, Advanced Security, and Studio; metrics collection; and Kafka and Docker integrations. We’ll take a look at all of it, plus give you a sneak peek at some of the foundational changes coming in DataStax Enterprise 6.8 that will rock your world.
This is for C* workloads. 2x read/writes on both throughput and latency.
If asked: Not much TPC perf gain with how lucene works. DSE Search and TPC are not aligned yet and we are working on a solution to provide TPC/Advanced Performance for Search in the future.
This is testing a higher stress workload. Where 5.1 starts showing distress at 40,000/ops/s 6 easily handles 80,000/ops for the same workload.
Double the performance helps our customers on a couple dimensions: You can do more stuff, and you can do it more quickly. You can serve double the users on the same hardware -- or cut your hardware in half for existing user workloads. So there’s a cost-effectiveness component and the ability to wring more work out of the hardware.
moderate to small workload just to illustrate differences in latency when the system is not overly stressed
Not only does SEO account for faster compaction, but it has a direct effect on…. (next slide for TPC)
SEO implements AIO (AsyncIO)
To help mitigate some of these problems, complex tools were built to help orchestrate and add some structure and resiliency to repair. But only so much you can do with tooling.
https://www.datastax.com/2018/04/dse-nodesync-operational-simplicity-at-its-best
Increased unification with DSE Analytics allows you to perform analytics queries on graph data without a traversal (SQL)
With DSE Graph 5.1, we quickly learned that some items like transactional analytics were much faster using DSE Graph Frames vs. Gremlin OLAP. That’s why with the release of DSE 6, graph users will no longer have to choose which analytical implementation to use to perform a graph analytics operation. With DSE 6, the DSE Graph engine will automatically route a Gremlin OLAP traversal to the correct implementation, resulting in the fastest and best execution for end users.
https://www.datastax.com/2018/05/whats-new-in-dse-graph-6
Not saying SQL is the best language for analytics, but a lot of folks know SQL and want to enable them to join the fun
https://www.datastax.com/2018/05/whats-new-for-datastax-enterprise-analytics-6
Search optimization determines which is better path, Search index or not
full production support for structured streaming, this was experimental prior
https://www.datastax.com/2018/05/whats-new-for-datastax-enterprise-analytics-6
Backup service - natively leverage Amazon S3 API’s to improve speed and efficiency
https://www.datastax.com/2017/04/announcing-opscenter-6-1
https://www.datastax.com/2018/05/announcing-opscenter-6-5
Upgrade service - downloads software updates onto servers or staging area, applies updates in rolling restart fashion so no downtime
https://www.datastax.com/2017/04/announcing-opscenter-6-1
https://www.datastax.com/2018/05/announcing-opscenter-6-5
OLD Data access = Gremlin, Spark (SQL, DataFrames, etc…) :::: NEW Data access = CQL, DSE Search, Gremlin, Spark (SQL, DataFrames, etc…)
OLD data loading = Gremlin, graph loader, Spark - DSEGF :::: NEW DSE unified ingest = Gremlin, CQL load utils, DSBulk, Spark - DSEGF
OLD schema storage = Graph + CQL :::: NEW schema storage = CQL only
We currently see 2 paths for DataStax’s One Model story in DSE 6.8
Path 1: Seamlessly mixing models in the same or different application
Write data in any model you choose and read data in any model you choose
Path 2: Layer models as a development strategy
Start with the easy, proven, known CQL data model and have the ability to add graph functionality easily over time, without rewriting/migrating data.
hopefully this year, maybe in the fall
hopefully this year, maybe in the fall
DataStax Labs provides the Apache Cassandra™ and DataStax communities with non-supported previews of enhancements that may or may not be included in future DataStax production software, as well as tools, aids, and partner software designed to increase productivity
So be our guest, have fun with DataStax Labs Previews, try it out, but DO NOT put any Labs technologies into production. It almost seems like we shouldn’t say that, but we know how exciting this can be. Just resist the temptation.
As you are trying out some of our new Labs technologies, tools, and experimental features we would love your feedback. Good or bad, let us know!
You can connect with us through DataStax Community <community.datastax.com>