What’s actually slowing this PC down?
Pick the symptom - the matching free tool is one click away.
There was no single winner in Ahmed Amer’s August 27, 2026 benchmark of five managed graph databases. Memgraph had the lowest reported one-hop traversal latency and led the benchmark’s lookup tests; Neo4j AuraDB was fastest at a full-graph citation aggregation. In a mixed read/write test, ArangoDB’s throughput changed little as concurrency rose from 10 to 40 clients. These are results from the providers’ particular free-tier or trial configurations—not a resource-matched verdict on the databases themselves.
Which database was “best” depends on the workload
The rankings changed with the query. Memgraph led the reported one-hop traversal result, while Neo4j AuraDB led the full-graph aggregation. The mixed workload produced a different comparison: four services handled substantially more operations per second at 40 clients than at 10, while ArangoDB’s reported throughput was nearly flat.
As an Amazon Associate I earn from qualifying purchases.
| Benchmark result | Service | Reported result |
|---|---|---|
| One-hop traversal, p50 latency | Memgraph | 69.4 ms |
| One-hop traversal, p50 latency | Neo4j AuraDB | 77.4 ms |
| One-hop traversal, p50 latency | CognoDB Cloud | 139.9 ms |
| One-hop traversal, p50 latency | ArangoDB Oasis | 173.8 ms |
| One-hop traversal, p50 latency | FalkorDB Cloud | 193.0 ms |
| Full-graph citation aggregation, p50 latency | Neo4j AuraDB | 185.2 ms |
| Full-graph citation aggregation, p50 latency | Memgraph | 266.7 ms |
| Full-graph citation aggregation, p50 latency | FalkorDB Cloud | 402.0 ms |
| Full-graph citation aggregation, p50 latency | CognoDB Cloud | 1,799.1 ms |
| Full-graph citation aggregation, p50 latency | ArangoDB Oasis | 4,058.0 ms |
These latency figures are the benchmark author’s reported medians (p50), not independent or industry-wide measurements. The article also reports that Memgraph led its primary-key and indexed/filtered lookup tests, but the results summarized here do not provide lookup timings to compare.
Recommended Free Tools
What happened when the benchmark added concurrent clients?
The mixed workload combined 80% reads and 20% writes. The benchmark ran each concurrency level for ten seconds; the figures below are reported throughput, in operations per second.
#1 Best Overall
| Service | 10 clients | 40 clients | Change, approximately |
|---|---|---|---|
| Memgraph | 136.4 ops/sec | 497.1 ops/sec | 3.6× |
| Neo4j AuraDB | 111.4 ops/sec | 442.6 ops/sec | 4.0× |
| CognoDB Cloud | 63.4 ops/sec | 246.7 ops/sec | 3.9× |
| FalkorDB Cloud | 50.0 ops/sec | 203.2 ops/sec | 4.1× |
| ArangoDB Oasis | 15.8 ops/sec | 16.6 ops/sec | 1.05× |
ArangoDB’s near-flat result is a measured outcome in this setup. The author checked that the edge index was used and reported no query-planner warnings. A connection-pool limit, HTTP/REST overhead, or an instance resource ceiling were proposed as possible explanations, but the test does not establish the cause. It would be a mistake to treat this observation as proof that ArangoDB generally fails to scale.
What did the benchmark actually test?
Ahmed Amer’s comparison used one client machine and one shared dataset across CognoDB Cloud, Neo4j AuraDB, Memgraph Cloud, FalkorDB Cloud, and ArangoDB Oasis. The dataset was SNAP’s cit-HepTh high-energy-physics theory citation network: 27,770 papers and 352,807 directed citation edges covering January 1993 through April 2003. The graph represented papers as Paper nodes and citations as CITES relationships.
Rank #2
Because the raw dataset did not have a second attribute for filtered lookups, the benchmark added a synthetic bucket property calculated as id % 100. Its workload categories included ingestion, one-, two-, and three-hop traversals, primary-key and indexed/filtered lookups, a full-graph aggregation that counted citations per paper and returned the top 20, and the mixed concurrent read/write test.
For read tests, the author ran ten warm-up iterations followed by 100 measured iterations. The concurrent test ran for ten seconds at each of its two client counts. The reported one-hop result is one depth-specific latency measure; it should not be read as the result for every traversal depth or every graph query.
Rank #3
Why these numbers are not a controlled engine shootout
The services were tested in their respective no-cost configurations, not on matched hardware. The repository reports CognoDB at 0.5 vCPU and 512 MB RAM; Memgraph on a 2-CPU, 2-GB-RAM 14-day trial; FalkorDB with a documented 100 MB free-tier memory limit; and ArangoDB on a 4-GB trial deployment. Neo4j’s free-tier CPU and memory allocation was not disclosed in the benchmark. Differences in available resources can affect performance, so the results compare the tested offers and configurations as well as the underlying systems.
The deployments were not placed in deliberately matched regions. CognoDB and Neo4j happened to run in us-east4, Memgraph in Frankfurt, and FalkorDB in AWS ap-south-1. The author notes that regional latency may have affected query times. With one client machine and deployments in different locations, the latency results are not geography-neutral.
Rank #4
The benchmark also reports two provider-specific complications that should be kept in context:
Do these 3 things before closing this tab:
1Repair Windows errors before they cause bigger problems2Fix the driver behind crashes, sound loss and screen glitches3Clear out junk files and repair common Windows errors- FalkorDB’s Bolt endpoint did not connect in this environment, so the author used its native RESP client. That is an environment-specific report, not evidence that FalkorDB generally lacks Bolt support. The author also noted that the documented 100 MB free-tier limit seemed inconsistent with loading the dataset, but did not independently verify that apparent mismatch.
- For CognoDB, the author reported that the same Neo4j driver code worked after changing the connection credentials and URI. That is observed compatibility in this benchmark, not a universal compatibility guarantee.
How to use the results when choosing a graph database
Use the findings to decide what to test, not to skip testing. If your application is traversal-heavy, the one-hop result gives Memgraph a reason to be on your shortlist; if it depends on graph-wide citation-style aggregation, Neo4j AuraDB led this particular test. If concurrency matters, reproduce the mixed workload and inspect both throughput and latency under the load your application expects. A ranking on one query shape does not settle performance on another.
Before selecting a managed service, rerun representative queries against your own graph and compare equivalent conditions where possible. Keep the dataset, client location, resource tier, protocol and driver, warm-up, iteration count, concurrency, query shape, and result size consistent—or record differences explicitly. Include ingestion and the indexes your application will actually use, not just the fastest read query.
Amer’s benchmark article and its linked repository contain the scripts, queries, caveats, and instructions for rerunning the tests. The repository also makes the setup easier to inspect than a table of timings alone. Still, the published figures come from one benchmark run and have not been independently replicated in the evidence available here; they should be treated as a starting point for a workload-specific evaluation.
Quick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
The Tool Desk
Outbyte PC Repair FREEClear out junk files and repair common Windows errorsFree Scan →Outbyte Driver Updater FREEFix the driver behind crashes, sound loss and screen glitchesFind Drivers →
