Compare / OpenSearch Service
WarpSearch vs Amazon OpenSearch Service (managed domains)
AWS's managed OpenSearch clusters. You pick instance types and counts and pay for every node-hour, plus storage, whether or not anyone is querying.
The scoreboard
Price
Cheaper on all 4 workloads OpenSearch Service can run
Monthly, list prices. Bars are drawn to the middle of a range.
Performance
No like-for-like data
No published latency to compare
AWS publishes no latency for domains; speed depends on the instances you buy and how you shard. We have not run OpenSearch Service ourselves, so we make no speed claim here.
How they bill
| WarpSearch | OpenSearch Service | |
|---|---|---|
| Billing unit | Per GB stored, per GB written, per request | Per instance-hour for data, master and warm nodes, plus EBS and managed storage |
| Smallest HA cluster | $10 a month minimum | About $390 a month for 3 x r7g.large data nodes ($0.178/hour each), before storage and dedicated masters |
| Scale to zero | Yes, after 15 minutes idle | No. Nodes run all month |
| Capacity planning | None | You size nodes, EBS volumes and shards |
| Log retention | One index, storage tiered per GB | Hot nodes plus UltraWarm (S3-backed, read-only) at $0.024/GB-month plus warm node hours |
| API | Elasticsearch, OpenSearch and Solr compatible | Full OpenSearch |
What OpenSearch Service cannot do, or does only in part: Provisioned clusters billed by the instance-hour whether you query or not. You choose instance types, node counts, EBS volumes and shards. Warm (UltraWarm) storage needs dedicated master nodes and holds read-only indexes.
Price
Monthly cost for four common workloads, at list prices in US East.
1M text chunks with 768-dim embeddings, 7 GB of JSON. 1M queries a month spread through the day. 1 GB written a month.
| WarpSearch | OpenSearch Service | |
|---|---|---|
| Storage | $0.56 (7 GB held x $0.08) | $2 to $7 (20 GB gp3 per node at $0.122) |
| Writes | $1.00 (1 GB x $1.00) | In instance cost |
| Search | $2.00 (1M requests) | $130 to $390 (1 r7g.large with no HA, or 3 across three zones) |
| Data returned | $0 (about 40 GB, inside the free 100 GB) | AWS data transfer rates |
| Minimum | $10 minimum applies (usage is $3.56) | None. Production adds about $296 for 3 dedicated masters (m7g.large) |
| Total a month | $10 | $132 to $397 |
Our sizing. 1M x 768-dim vectors need about 3.5 GB of memory for HNSW, which fits a 16 GB node.
5M products with 384-dim embeddings, 10 GB of JSON. 50M queries a month (about 19 a second on average). 20 GB written a month.
| WarpSearch | OpenSearch Service | |
|---|---|---|
| Storage | $0.60 (7.5 GB held x $0.08) | $11 to $18 (30 to 50 GB gp3 per node) |
| Writes | $20 (20 GB x $1.00) | In instance cost |
| Search | $100 (50M requests) | $390 to $1,557 (3 x r7g.large to 3 x r7g.2xlarge) |
| Data returned | $18 (250 GB returned, 100 GB free, before compression) | AWS data transfer rates |
| Minimum | Cleared | None. Production adds about $296 for dedicated masters |
| Total a month | $139 | $400 to $1,575 |
Our sizing. 5M x 384-dim vectors need about 8 to 9 GB of memory per copy; the right size depends on your peak query rate, which varies from the 19 a second average.
10 TB of raw logs kept searchable. 1 TB written a month. 10,000 dashboard aggregation queries a month.
| WarpSearch | OpenSearch Service | |
|---|---|---|
| Storage | $455 (8.5 TB held: 1 TB x $0.08, 7.5 TB x $0.05) | $277 ($204 UltraWarm for 8.5 TB at $0.024, $73 hot gp3 for the recent week) |
| Writes | $1,000 (1 TB x $1.00) | $390 (3 hot r7g.large nodes) |
| Search | $10 (10,000 requests, about 10 CPU-seconds each) | $1,042 (6 UltraWarm medium nodes at $0.238/hour) |
| Data returned | $0 | AWS data transfer rates |
| Minimum | Cleared | $296 for 3 dedicated masters, required for UltraWarm |
| Total a month | $1,465 | About $2,000 |
Estimate. 8.5 TB is 7.7 TiB, and a medium warm node addresses 1.5 TiB. Warm indexes are read-only, and aggregation buckets default to 10,000 with UltraWarm on.
1 PB of raw logs kept searchable. 50 TB written a month. 100,000 heavy aggregation queries a month.
| WarpSearch | OpenSearch Service | |
|---|---|---|
| Storage | $31,280 (850 TB held, tiered) | $23,255 ($20,400 UltraWarm for 850 TB, $2,855 hot gp3 for the recent week) |
| Writes | $6,450 (1 TB x $1.00, 9 TB x $0.25, 40 TB x $0.08) | $3,114 (6 hot r7g.2xlarge nodes) |
| Search | $1,330 (100,000 requests, about 133 CPU-seconds each) | $76,300 (39 UltraWarm large nodes at $2.68/hour) |
| Data returned | $0 | AWS data transfer rates |
| Minimum | Cleared | $296 for 3 dedicated masters |
| Total a month | $39,060 | About $103,000 |
Estimate. 850 TB is 773 TiB, and a large warm node addresses 20 TiB, so 39 nodes. Warm CPU for 100,000 heavy aggregations is not sized, so this is a floor.
Performance
What OpenSearch Service publishes
| Query latency | Not published. Depends on the instance types and shard layout you choose |
|---|---|
| AWS-commissioned benchmark (Trail of Bits) | OpenSearch 2.17 vs Elasticsearch 8.15 on big5: geometric mean 12.1 vs 18.8 ms. Not a comparison with us |
| UltraWarm query capacity | A medium warm node has 2 CPU cores and addresses 1.5 TiB; AWS warns broad queries across many shards can strain it |
| Cold start | None. Nodes are always on |
AWS publishes no per-query latency for managed domains, so a number-for-number comparison is not possible. The one public benchmark compares OpenSearch with Elasticsearch, not with WarpSearch. Our own medians on 2 to 4 CPUs are listed below.
What we measured
| Test | Result |
|---|---|
| Keyword search, 2.8M docs (LoTTE) | 2.1 ms median, 3,700 queries/s |
| Keyword search, 10M passages (MS MARCO v2) | 2.4 ms median (5.4 ms with an exact total hit count) |
| Keyword search, small BEIR sets | 1 to 6 ms median |
| Vector search, 10M x 128-dim (BigANN SIFT) | 0.964 recall@10 at 6.2 ms median, 162 queries/s |
| Vector search, 1M x 1536-dim (arXiv) | 0.961 recall@10 at 15.7 ms median |
| Filtered vector search, 10M (YFCC) | 0.959 recall@10 at 10.4 ms median |
| Log filters, 22M events (http_logs) | 1 to 2 ms median |
| Log aggregations, 22M events | 190 to 430 ms median (sort, daily histogram, avg by status) |
| Ingest, short text | 63,000 to 97,000 docs/s |
| Ingest, logs (247M events) | 76,000 to 81,000 events/s |
| Ingest, 768 to 1536-dim vectors | 2,200 to 4,000 vectors/s |
| Time to searchable | Seconds, or immediate with commit=true |
| Cold start | Not measured yet. First query on a 2.3 GB index: 17 to 19 ms vs 2.7 ms warm (page cache only) |
Measured October 2026 through our Elasticsearch-compatible API, warm cache, client in the same region, single runs. Every test ran on deliberately minimal hardware, 2 to 4 CPUs (a 4 CPU container or an AWS i8g.large with 2 vCPUs), so these are the least favorable results we can produce. More hardware only improves them. Numbers exclude internet round trips.
Bottom line
Choose WarpSearch for
- Search apps: $10 vs $132 to $397 for RAG, and $139 vs $400 to $1,575 for the catalog workload.
- Log retention, by our estimate: about $1,465 vs about $2,000 at 10 TB, and $39,060 vs about $103,000 at 1 PB, in one index with no hot and warm tiers to manage.
- No instance types, EBS volumes, dedicated masters or shards to size, and no paying for idle hours.
Choose OpenSearch Service for
- Full OpenSearch: every plugin, Index State Management, security analytics and Dashboards.
- Always warm: no cold start on any query.
- Native AWS: IAM, VPC and CloudWatch, with reserved instance and savings plan discounts on steady load.
WarpSearch costs a fraction of a provisioned cluster for apps and catalogs, and less for log retention by our estimate. Pick OpenSearch Service when you need the full OpenSearch plugin set on infrastructure you control and your load is steady around the clock.
Sources
- OpenSearch Service pricing
- AWS Price List API (OpenSearch Service, us-east-1)
- UltraWarm storage
- Trail of Bits benchmark
- WarpSearch pricing
Rates are on-demand, US East (N. Virginia), from the AWS Price List API on October 21, 2026. Node counts and sizes are our estimates, not AWS sizing guidance. Reserved instances and savings plans lower the instance price. Prices checked October 20, 2026. List prices in US dollars, before discounts. Estimates are marked. Tell us if something is out of date: contact us.