Etermax at RedisConf 2019
RedisConf is the main conference for anything Redis related. It is hosted by RedisLabs, the main contributor behind Redis, and as always included the presence of Salvatore Sanfilippo, creator of Redis. The event took place between April 1st and 3rd at Pier 27, one of the many waterfront venues in San Francisco.
What were we looking for?
We went there to learn about the latest Redis announcements, to exchange experiences and also to empower our business relationships with different actors on the Silicon Valley scene.
Also, having the opportunity to talk to other gaming companies, especially with developers, is always an enriching experience because you learn what they normally don’t tell you “on stage”, and that’s really cool.
We were also invited to close the keynote with a brief interview together with Avram Lyon, VP of Technology at Scopely, sharing some ideas and commenting on our experience with Redis from the gaming industry perspective.
What did we find?
There were more than 1500 Redis developers, innovators and contributors who met to share use cases, get hands-on training and learn how to build innovative applications that scale. There were approximately 60 technical sessions, plus hands-on exercises. All of the cloud big players were sponsoring the event and that shows the importance Redis has for the industry nowadays.
Finding a free spot to seat was our first challenge in this big building where different sessions took place simultaneously, only separated by “separating posts”, with colored headphones for each talk. We were a bit skeptical about everyone being on the same physical space, but the headphones strategy turned out very well.
Highlights
Optane memory support: reached GA
The main story here is that Redis Enterprise is now generally available for any Intel ® Optane™ DC Persistent Memory hardware.
We will probably need to explain a little bit about Redis tradeoffs and the RedisLabs “Flash technology” for this to make sense, so here we go.
With Redis, you have an extremely fast database, mainly because all the data is stored in RAM. The obvious tradeoff here is that RAM is more expensive than common storage. This is when Redis Flash comes into play.
When you have a big set of data, and by taking advantage of the fact that normally only a subgroup of values will be intensively accessed, what happens is that these “hot values” will be kept in RAM, while the “cold values” will be moved to an NVMe-based SSD disk. It is important to note that this disk is treated as a RAM extender and does not replace the existing data-persistence mechanism of Redis. We have been using it for more than a year now, and you can learn more about our story by watching our talk in last year’s redisconf.
Finally, this is where Intel Optane technology comes into play. This technology is supposed to replace the previously mentioned NVMe-based SSD disk, with an expected huge gain in performance (some benchmarks here).
The only missing thing is that, for the cloud world, it’s currently only available on Google Cloud Platform. However, if the performance gain proves to be that huge, we should probably just sit and wait for the other cloud providers to adopt it.
New module: RedisGears
This new module is an in-database serverless engine for multi-model and cluster operations supporting both event-driven and batch operations. From a reductionist point of view, it appears that Redis is now offering its own “AWS-Lambda” to communicate modules, allowing you to build custom pipelines of data processing.
Python is the language to use for these scripts, with more languages to be added in the future.
This seemed to be one of the top announcements since it was present in most of the sessions and it is the enabler of the new multi-model approach that they seem to be taking.
New Module: RedisTimeSeries
The module name basically explains itself. It adds time-series data structure to Redis. In the documentation, you can find interesting use cases for this, such as using it as the backend database for a decentralized Prometheus server or just integrating it directly to grafana in order to have a quick-and-easy metrics visualization system.
RDBTools: visualizations
Redis Labs announced the acquisition of RDBTools from HashedIn. This tool enables you to explore and interact with your redis data, find ways to reduce memory usage, monitor commands in realtime, do basic cluster management and view the slow log.
Any Redis user knows that exploring a big Redis database is not an easy or safe task. At Etermax, we had no option but to develop our own exploring tool, and it does not meet all our needs.
Not all the data structures are supported yet, but over the next few months, they will be adding support for Redis Streams as well as RedisJson, RediSearch, Redis Graph, RedisTimeSeries and RedsiAI, and they will also be adding monitoring and alerts later on this year.
We are eager to give it a try to see if we can deprecate our custom browser tool.
Conclusions
New Redis modules were announced, new visualization offerings appeared, a faster & affordable approach is on the way, and the “multi-model” approach was presented.
About this new approach, the idea seems to be to transition from being just a really fast no-sql database or cache, to become a multi-purpose database where you can do anything from storing time-series data, doing machine learning, performing complex indexing & search, and even using it as a kafka-like streaming platform.
We have to admit that we are a bit skeptical about this new direction, but since they have surprised us before, we’ll have to wait and see.
Finally, we enjoyed meeting Salvatore Sanfilippo in person and having the chance to discuss technical issues with him. We were glad to see that such a passionate and talented person can also be so humble and nice to talk to!
