Elasticsearch 5.0 Now Available on ObjectRocket

The Elasticsearch community has reached a huge milestone with the release of Elastic Stack 5.0.

We’re ready to join the party and are now offering Elasticsearch and Kibana 5.0 as an Early Access option for new builds on the ObjectRocket platform. This means speed, new features and peace of mind knowing you only need to remember one version number from now on, all on the dedicated and fully managed ObjectRocket hosting platform.

What’s New in Elasticsearch 5.0?

You can get the full list of new features in latest version of the Elastic Stack from the fine folks at Elastic, but given our focus on operating and managing your cluster, there are a few things in 5.0 we want to highlight. I’ll provide some more detail below, but my favorites are:

  • The ingest node
  • Performance improvements
  • Timelion in Kibana core

With all of the great new things, there are a couple of casualties in the new version. The largest of these is site plugins, which enabled the ElasticHQ, Kopf and elasticsearch-head dashboards we’ve offered on all other versions of Elasticsearch on the ObjectRocket platform. As of today, these plugins don’t work on Elasticsearch 5.0 clusters, but we’re working hard to bring them back on the ObjectRocket platform, so stay tuned.

Ingest Node

All Elasticsearch clusters on ObjectRocket split out node roles: we have dedicated master nodes, data nodes, client nodes for load balancing and Kibana nodes. Adding to that list is the ingest node, a preprocessor that will allow you to transform and operate on your data as it’s being loaded into the Elasticsearch cluster.

Today, when you spin up an Elasticsearch 5.0 cluster on ObjectRocket, you’ll get client nodes configured to carry both the role of ingest node and client node. However, we’re working on a couple of potential extra options for ingest nodes, like dedicated ingest nodes and the ability to split your current client nodes into specific roles. We’re interested to see the use cases and types of adoption for the ingest nodes, so let us know what you think and how you’d like to use them for your application.

Performance

Performance has been a common topic throughout the development of 5.0 and in our initial testing on the ObjectRocket platform, it delivers. To test it out, we’ve used ESRally as an objective benchmark to compare the performance of version 5.0 and 2.3.4 on two similarly configured ObjectRocket for Elasticsearch clusters.

Using the PMC track in Rally, we first compared latency, service time and throughput metrics for each test. That’s a lot of data, but since all three of those metrics show a similar trend, I’m only showing the 90 percent service time data below. The main trend is that most tests are +/-10 percent between the two versions, except for expression and index-append tests, which show massive improvements on 5. Note that in the charts below, all numbers are normalized and lower numbers are better.

Elasticsearch results

Beyond the per-test metrics, the most compelling metrics are the total time metrics gathered for the whole test run. Across the board, we saw huge improvements for Elasticsearch 5. Except for flush time, every metric took half of the time or less than Elasticsearch 2.3.4.

Elastic search total operation time

The last result I want to point out is the heap metrics. In this case, there are huge savings in the heap used for terms and segments. What this means is that you get all of that performance goodness without chewing up a whole lot of your usable memory.

Elasticsearch heap

It’s really promising to see Elasticsearch 5 come out of the gate with such significant performance improvements. We’ll definitely do more analysis and dig into the numbers from these runs a little deeper, but what we’ve seen so far looks quite good. I’d like to extend a big thank you to Seth Sanchez on our engineering team for getting these performance numbers together so quickly.

Timelion

I had to throw in a Kibana feature, since we include Kibana with every ObjectRocket for Elasticsearch cluster, so Timelion makes the list. We regularly get requests to install Kibana plugins and Timelion by far leads that list. Data with a time component, like that from our Twitter connector, always presents new challenges in indexing and visualization, so Timelion just kills it on the visualization side. It’s great to now have that as a standard part of Kibana.

Try Elasticsearch 5.0 on ObjectRocket now!

There’s so much more than the few items listed above, so go create an Elasticsearch 5.0 with Kibana instance today in the ObjectRocket UI by selecting 5.0 as the version during the creation process. We’re going to continue to build upon the product to take advantage of new features and to find creative ways to bring back old ones (long live site plugins!), so you can expect to see a constant stream of updates and improvements in the coming weeks.

It is also worth noting that since this is a brand new release, we have version 5.0 marked as Early Access on our platform. The Early Access label is a light warning that the technology may not be 100 percent production ready yet, and that it hasn’t been available long enough for us to guarantee it’s as stable and reliable as previous versions. Our customers had some challenges with 2.0 when it launched, and though we see no red flags in 5.0, we would like to be a little cautious until it’s been in the wild a bit longer.

For more details on the implications of Early Access, please see our test terms.

Steve Croce joined Rackspace in 2015 and currently works in the Datastores group as product manager for Rackspace Cloud Databases and ObjectRocket Elasticsearch. He has 15 years experience in the tech industry with roles in hardware and software ranging from development to product management and marketing. He holds a Master’s degree in Electrical Engineering and an MBA from MIT and now lives in Austin, TX with his wife and two children.

LEAVE A REPLY

Please enter your comment!
Please enter your name here