Managed OpenSearch, sized for your index
Elasticsearch as a service on OpenSearch, with shards, node roles and retention sized for your index: dedicated clusters in OSStend's cloud, inside your own cloud account, or co-managed in your data center, run 24/7.
45 minutes with an engineer: we review your workload and sketch the design.
- ✓Operated by AceMQ: 160+ customers, 10+ years
- ✓24/7 on-call engineers
An example design file for a log search service, with illustrative values:
# Managed OpenSearch service: opensearch-logs engine: opensearch masters: 3 data_nodes: 6 target: aws/dedicated snapshots: hourly monitoring: included on_call: 24/7
Managed OpenSearch from OSStend is Elasticsearch as a service on the open-source fork: dedicated master and data nodes sized for your index, with keyword and vector search together. Engineers run it 24/7 in OSStend's cloud, your cloud account or your data center, including on-premises for regulated data.
- PostgreSQL
- MySQL
- MongoDB
- Redis
- Valkey
- OpenSearch
- Kafka
- RabbitMQ
- AWS
- Google Cloud
- Azure
- Kubernetes
- OpenShift
- Prometheus
- Grafana
- Datadog
- Dynatrace
- Splunk
What OSStend tailors for OpenSearch
An OpenSearch cluster is designed around one index estate: its size and growth, the shard and node layout, the search features it needs and how long data is kept. The design is written down and signed off before anything is built.
- Snapshots, hourly
- Metrics to your Prometheus or Datadog
- AceMQ engineers on call 24/7
| Area | Set for your workload |
|---|---|
| Size and shape | Dedicated master nodes (called cluster manager nodes in current OpenSearch) and data nodes, with the primary shard count fixed before the first document, because an existing index's primary shard count is hard to change. AWS's guideline is 10-30 GiB per shard where search latency matters and 30-50 GiB for write-heavy log analytics, with no more than 25 shards per GiB of heap (checked 9 Nov 2026). |
| Topology | Multi-zone clusters with replica shards, three dedicated master nodes for a stable quorum (the same guidance AWS gives), and hybrid links between cloud and data center. Streams from Kafka feed the index directly; see managed Kafka. |
| Index lifecycle | Rollover and retention policies set for how long each kind of data stays searchable, using Index State Management policies that roll indices over, snapshot them and delete them at the age you choose. |
| Search features | Keyword plus vector (hybrid) search for AI retrieval. See vector database for RAG. |
| Versions and upgrades | Version pinning to the release your applications are validated on. OpenSearch forked from Elasticsearch 7.10.2, and the OpenSearch support matrix maps Elasticsearch 7.0 to 7.10.2 to OpenSearch 1.x. Later Elasticsearch versions are not on it, which is why the design step checks the source version first. Upgrades follow your change calendar. |
| Where it runs | OSStend's cloud, your cloud account, your data center for regulated data, Kubernetes or OpenShift, or an air-gapped site. |
| Monitoring and snapshots | Metrics and alerts forwarded to your Datadog, Dynatrace, Splunk or Prometheus, with snapshot and restore on a schedule that suits your recovery needs. On-call engineers watch cluster status, unassigned shards, JVM heap and shards per node. |
There is no fixed menu here: the plugins, settings and cluster size your workload needs are written into the design you sign off before anything is built, and that design is what OSStend runs.
Where turnkey search services stop
Plan-based search services ship their own index defaults, shard ceilings and plugin rules. Amazon OpenSearch Service documents each of them, and they differ from the open-source defaults. An index estate that does not match them has to be reshaped to fit.
| Area | What a plan fixes | What OSStend designs instead |
|---|---|---|
| Index defaults | A new index on Amazon OpenSearch Service gets five primary shards and one replica by default, where open-source OpenSearch defaults to one primary and one replica (checked 9 Nov 2026). | The primary shard count set per index from its size and growth, before the first document goes in. |
| Shard ceilings | The limit is 1,000 shards per node on OpenSearch 1.1 to 2.15 and up to 4,000 from 2.17, tied to JVM heap (AWS docs, checked 9 Nov 2026). | A shard plan set from your heap and index sizes, with too many small shards treated as a design fault. |
| Plugins | Custom plugins need OpenSearch 2.15 or later, and AWS caps them at 25 per account per Region, 1 GB each, and 20 plugins of any type per domain (checked 9 Nov 2026). | The plugins your workload needs, agreed in the design. |
| Where it runs | One provider's cloud: Amazon OpenSearch Service runs on AWS. | OSStend's cloud, your cloud account or your data center, on one contract. |
Plan menus differ by provider and change over time. Each figure above is from AWS's current documentation with the date checked, and the Amazon OpenSearch Service alternatives page adds its price lines and quotas dated 17 Nov 2026, and any other provider's limits should be read from its own documentation.
Design. Move. Run. for OpenSearch
For OpenSearch, an engineer designs the node layout and shard plan and you sign it off, OSStend picks the data-copy method for your source version and owns the cutover, and named engineers then run the cluster 24/7.
Design
An engineer reviews your index estate (document counts, growth, query and ingest rates, retention, source version and where it must run) and writes the cluster design: node layout, shards, storage, search features and version. You sign it off before anything is built.
Move
OSStend runs the migration from Elasticsearch or a self-run cluster and owns the cutover. Snapshot and restore or reindexing moves the data, as the design step decides for your source version, and applications are tested against OpenSearch before traffic switches. Both sides stay live until you do.
Run
Named engineers run it 24/7: snapshots, upgrades on your calendar, shard and capacity changes, lifecycle policy tuning, heap and shard-count watching and incident response, with an SLA written into your contract.
Deployment and service levels
OpenSearch runs dedicated in OSStend's cloud or inside your own cloud account, both fully managed, or co-managed in your data center. Regulated data can stay on premise.
Dedicated in OSStend's cloud
Single-tenant OpenSearch clusters on Google Cloud or AWS, with master and data nodes spread across zones and replica shards placed away from their primaries.
Fully managedIn your cloud account
OpenSearch inside your VPC, so index data stays in your account together with the node bill and the network boundary. AWS first, then Azure and Google Cloud.
Fully managedIn your data center
Your hosts, Kubernetes or OpenShift clusters, with air-gapped sites and regulated data kept on premise, run by OSStend engineers through an operator.
Co-managed
Fully managed OpenSearch clusters carry an uptime commitment, and a data center deployment carries a 24/7 response-time commitment because the nodes and disks are yours. The service levels page sets out the tiers, and deployment shows each option in detail.
Uptime and response figures are set per plan and written into your contract. AceMQ, which operates OSStend, sets its support response targets from 15 minutes, by plan.
Proof from AceMQ's operating record
The shard, snapshot and upgrade work for OpenSearch is done by AceMQ's engineers. AceMQ operates OSStend with 160+ customers, 10+ years of running open-source infrastructure and engineers on call 24/7, which matches acemq.com.
Customer names are shared only with AceMQ's written permission, so the OpenSearch record here is anonymous. A reference with a similar index size or source can be raised in the 45-minute review.
Pricing for managed OpenSearch
Managed OpenSearch is priced from four drivers: the node layout, the index size and retention, the SLA tier and where it runs. An engineer prices the design after a workload review, and there is no free tier or card sign-up for OpenSearch.
- Node layout
- three dedicated master nodes plus the data nodes, and replica shards on top, set how many instances have to run.
- Cluster size
- master and data node count, instance size and storage, which follow index size and retention.
- SLA tier
- the uptime commitment on a fully managed cluster, or the 24/7 response-time commitment on a co-managed one.
- Where it runs
- OSStend's cloud, your cloud account or your data center, which decides who pays for the nodes and disks.
Send the index size, daily ingest, retention and where search must run, and the engineer prices that cluster rather than a plan size.
The pricing page describes the model that applies across every engine.
When to contact OSStend, and when not to
Contact OSStend when an index estate has grown past what a search plan lets you configure, or has to run in your account or data center. If a plan still covers the index, stay on it.
Contact us if
- ✓Your Elasticsearch or search service has outgrown a plan: more nodes, more storage or settings the plan does not expose.
- ✓You need keyword and vector search together for AI retrieval.
- ✓Search has to run in your cloud account or in your data center for regulated data.
- ✓You are leaving Elasticsearch and want the migration to OpenSearch run for you.
Not a fit if
- ✕A standard search plan holds your index estate, and it costs less than a designed cluster.
- ✕You want a free tier, or to sign up with a card. For OpenSearch, OSStend begins with a conversation with an engineer.
- ✕Your applications depend on features that only Elastic provides, which OpenSearch does not.
Questions about Elasticsearch hosting on managed OpenSearch
These are the questions search and platform teams ask most before moving to managed OpenSearch.
What is Elasticsearch as a service, and where does OpenSearch fit?
Elasticsearch as a service means a provider runs search clusters for you. OpenSearch is the open-source fork of Elasticsearch, and it covers keyword search, log analytics and vector search. OSStend runs OpenSearch clusters designed around your index, with 24/7 on-call engineers.
Can OSStend run hosted Elasticsearch?
OSStend provides Elasticsearch hosting on OpenSearch, the open-source fork, and does not host Elastic's own distribution. If your applications depend on features that only Elastic provides, OpenSearch will not cover them, and the design review checks that before any move.
Which Elasticsearch versions can move to OpenSearch directly?
OpenSearch forked from Elasticsearch 7.10.2. The OpenSearch support matrix covers Elasticsearch 5.x to 7.10.2, with 7.0 to 7.10.2 going to OpenSearch 1.x, and says nothing about later Elasticsearch releases. The design step reads your source version and picks the route, and a version past 7.10.2 is tested rather than assumed.
How does a move from Elasticsearch to OpenSearch work?
The design step checks the source version and the applications' queries and clients. The data moves by snapshot and restore or by reindexing, applications are tested against OpenSearch, and traffic switches only after the new side is verified. OSStend engineers own the cutover. The Elasticsearch to OpenSearch migration guide covers the methods and the version limits.
How many shards should an index have?
Enough that each shard stays in a sensible size band, and no more. AWS's published guideline is 10-30 GiB per shard for search workloads and 30-50 GiB for write-heavy log analytics, because too many small shards waste CPU and memory. The primary shard count is set per index before data arrives, since changing it later is hard.
Can OpenSearch run in my own cloud account or data center?
Yes. OSStend operates OpenSearch inside your VPC, fully managed, with the index data kept in your account. In a data center it is co-managed on Kubernetes, OpenShift or your own hosts, including air-gapped sites for regulated data.
Can OpenSearch do vector search for RAG?
Yes. OpenSearch supports keyword plus vector (hybrid) search, which retrieval for AI applications often needs. See vector database for RAG for the pattern and the tuning.
What does managed OpenSearch cost?
The price follows the node count, index size and retention, the SLA tier and where it runs. The pricing page sets out the model, and an engineer prices the specific design after a workload review.
Related pages
- KafkaManaged Kafka in your cloud or data centerManaged Kafka with the broker count, partition plan, KRaft, Connect and schema registry designed for your estate, run 24/7 in your cloud or your data center.
- AI data · RetrievalVector database for RAG on your own infrastructureA vector database for RAG on your own infrastructure: when PostgreSQL with pgvector is enough, when OpenSearch hybrid search fits, and how to size retrieval.
- AlternativeAmazon OpenSearch Service pricing and alternativesAmazon OpenSearch Service pricing and alternatives: what the domain, storage and UltraWarm lines cost, and running dedicated OpenSearch outside AWS.
- Service levelsService levels: two tiers, a written SLAOSStend service levels: fully managed and co-managed tiers, what each one covers, who is on call and how incidents are handled, with a written SLA for each.
- DeploymentManaged databases: our cloud, yours or on-premManaged databases in OSStend's cloud, your AWS, Azure or Google Cloud account, or your data center. PostgreSQL, Kafka and RabbitMQ, run 24/7 by engineers.
- PricingDatabase as a service pricing: dedicated, tailoredDatabase as a service pricing from OSStend: dedicated, tailored clusters priced by engine, size, SLA tier and location. Enterprise only, with no free tier.
Talk to an engineer about your OpenSearch index
45 minutes with an engineer: bring your index sizes, daily ingest and retention, plus the Elasticsearch version you run, and the engineer sketches the cluster design.
- ✓A design sketch for your workload, not a plan menu
- ✓A zero-downtime cutover path for what you run today
- ✓A price for that design, after the review