Canberra, Australia, 4th May 2022: Instaclustr is pleased to announce a significant update to the node types offered in our OpenSearch offering on Instaclustr’s SaaS Platform. Instaclustr’s OpenSearch managed service is based on the truly open source, Apache 2.0 licensed, OpenSearch project.
Following the GA release of OpenSearch, we have refreshed our node type offerings across AWS, GCP, and Azure AZ cloud service providers. Each cloud service provider is continually introducing new and improved nodes, which have been carefully curated to provide our customers with a range of options to suit their specific workloads. Across all cloud service providers, we have also added instances with local storage, to provide further flexibility for use cases requiring dense storage.
Instaclustr’s AWS customers can now confidently deploy OpenSearch on Graviton2 nodes with GP3 volumes to gain both performance and financial benefits. For an equivalent node size, customers can expect a reduction in their AWS infrastructure fees and improved performance. Graviton2 infrastructure is approximately 20% cheaper than other equivalent instances using the same number of vCPUs and the same amount of RAM.
Instaclustr also pairs the Graviton2 infrastructure with the new generation of Elastic Block Storage (EBS) GP3 disks. These GP3 disks are cheaper, and they are configured with a higher IO throughput compared to GP2 disks. The shift to GP3 disk has also included an increase in the default IOPS, which has allowed Instaclustr to introduce production nodes with smaller disks (120GB) to suit customers with smaller storage requirements.
For the new Graviton/GP3 instances Instaclustr conducted a series of benchmarking experiments using Rally for all of our EBS-backed OpenSearch offering’s node sizes. Here is a sample of the results that we produced, with the following experimental setup:
- r5.2xlarge-1600-gp2 v.s. r6g.2xlarge-1600 with GP3 disk (4500 IOPS and 1000 MiB/s)
- 3 node cluster with only data nodes
- TLS Encryption enabled
- Default Rally configuration
OpenSearch 1.2.4
OpenSearch 1.0.0
From this data, we can see that the combination of Graviton2 and GP3 outperforms the x86 and GP2 for index-append operations. For other OpenSearch operations, comparable performance was achieved, as well as improvements with latency observed with OpenSearch version 1.2.4.
Based on previous experience with performance testing of Graviton2 nodes there were a number of avenues that were tested that didn’t have a meaningful impact on performance:
- Amazon Corretto Crypto Provider (ACCP) vs JDK default
- JDK 11 vs JDK 17 (JDK 17 is used)
- Versions of OpenSearch 1.0.0 and 1.2.4
In addition to the benchmarking, Instaclustr also conducted tuning of IOPS and throughput to identify the optimal configuration for each new instance/disk combination. During this testing, Instaclustr observed a saturation point in throughput based on the maximum 1000 MB/s provisioned throughput with AWS GP3 disk. The Rally performance testing framework is write heavy and increased instance size but did not yield any increase in throughput.
Existing customers running Open Distro for Elasticsearch will need to migrate to OpenSearch to take advantage of the Graviton2 nodes. These customers can contact our Support team for assistance with migration.
For customers running on Azure, we have added the ESv4 node family to our managed platform, which we have seen is better suited to the demands on OpenSearch. In addition, we have added developer node types for Azure which will allow testing and evaluation at an improved price point. GCP customers can now take advantage of the n2-highmem family across a range of storage options depending on your workload. For both Azure and GCP customers, these nodes will also be made available to customers who are still running workloads on Open Distro for Elasticsearch.
Continuing the focus on flexibility for our customers’ varied workload requirements, we have also added nodes with a larger storage capacity across all cloud storage providers. In addition, we have also added local storage options, which can deliver cost savings when taking advantage of reserved pricing.
