Loading skill
Install any skill in seconds. Free to start, no credit card required.
Get Started Free →Guides Qdrant multi-tenant scaling. Use when someone asks 'how to scale tenants', 'one collection per tenant?', 'tenant isolation', 'dedicated shards', or reports tenant performance issues. Also use when multi-tenant workloads outgrow shared infrastructure.
.claude/skills/qdrant-qdrant-tenant-scaling/SKILL.md| Test case | Without → With | Effect | Δ tokens | Δ turns |
|---|---|---|---|---|
| case-01 | ✗→✓ | ▲ Improved | -46% | 0% |
| case-02 | ✗→✓ | ▲ Improved | -3% | 0% |
| case-07 | ✗→✓ | ▲ Improved | -41% | 0% |
| case-13 | ✗→✓ | ▲ Improved | -20% | 0% |
| case-19 | ✗→✓ | ▲ Improved | -30% | 0% |
Do not create one collection per tenant. Does not scale past a few hundred and wastes resources. One company hit the 1000 collection limit after a year of collection-per-repo and had to migrate to payload partitioning. Use a shared collection with a tenant key.
Here is a short summary of the patterns:
Use the default multitenancy strategy via payload filtering.
Read about Partition by payload and Calibrate performance for best practices on indexing and query performance.
At this scale, the cluster may consist of several peers. To localize tenant data and improve performance, use custom sharding to assign tenants to specific shards based on tenant ID hash. This will localize tenant requests to specific nodes instead of broadcasting them to all nodes, improving performance and reducing load on each node.
If some tenants are much larger than others, use tiered multitenancy to promote large tenants to dedicated shards while keeping small tenants on shared shards. This optimizes resource allocation and performance for tenants of varying sizes.
Use when: legal/compliance requirements demand per-tenant encryption or strict isolation beyond what payload filtering provides.
is_tenant=true on the tenant index (kills sequential read performance)payload_m instead)| Case | Status | Duration (ms) | Turns | Tokens | Tool calls | ||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Without | With | Δ | Without | With | Δ | Without | With | Δ | Without | With | Δ | ||
case-01 | fail→pass | 31,304 | 11,806 | -62% | 1 | 1 | 0% | 4,493 | 2,437 | -46% | 0 | 0 | — |
case-02 | fail→pass | 23,159 | 20,088 | -13% | 1 | 1 | 0% | 4,299 | 4,165 | -3% | 0 | 0 | — |
case-03 | pass→pass | 18,643 | 11,590 | -38% | 1 | 1 | 0% | 3,253 | 2,623 | -19% | 0 | 0 | — |
case-04 | pass→pass | 17,142 | 10,268 | -40% | 1 | 1 | 0% | 2,965 | 2,293 | -23% | 0 | 0 | — |
case-05 | pass→pass | 29,100 | 7,539 | -74% | 1 | 1 | 0% | 1,861 | 1,729 | -7% | 0 | 0 | — |
case-06 | pass→pass | 10,265 | 5,436 | -47% | 1 | 1 | 0% | 1,681 | 1,264 | -25% | 0 | 0 | — |
case-07 | fail→pass | 19,778 | 6,942 | -65% | 1 | 1 | 0% | 2,883 | 1,701 | -41% | 0 | 0 | — |
case-08 | pass→pass | 24,687 | 11,537 | -53% | 1 | 1 | 0% | 3,434 | 2,369 | -31% | 0 | 0 | — |
case-09 | pass→pass | 16,373 | 9,916 | -39% | 1 | 1 | 0% | 2,720 | 2,004 | -26% | 0 | 0 | — |
case-10 | pass→pass | 18,030 | 9,068 | -50% | 1 | 1 | 0% | 3,085 | 2,234 | -28% | 0 | 0 | — |
case-11 | pass→pass | 15,737 | 8,656 | -45% | 1 | 1 | 0% | 2,593 | 1,783 | -31% | 0 | 0 | — |
case-12 | pass→pass | 15,271 | 10,003 | -34% | 1 | 1 | 0% | 2,297 | 1,894 | -18% | 0 | 0 | — |
case-13 | fail→pass | 17,534 | 13,409 | -24% | 1 | 1 | 0% | 3,078 | 2,470 | -20% | 0 | 0 | — |
case-14 | pass→pass | 23,913 | 17,753 | -26% | 1 | 1 | 0% | 3,448 | 3,208 | -7% | 0 | 0 | — |
case-15 | pass→pass | 19,964 | 12,394 | -38% | 1 | 1 | 0% | 3,333 | 2,290 | -31% | 0 | 0 | — |
case-16 | pass→pass | 21,101 | 14,638 | -31% | 1 | 1 | 0% | 3,397 | 2,624 | -23% | 0 | 0 | — |
case-17 | pass→pass | 7,714 | 3,557 | -54% | 1 | 1 | 0% | 1,079 | 1,007 | -7% | 0 | 0 | — |
case-18 | pass→fail | 22,715 | 9,954 | -56% | 1 | 1 | 0% | 3,205 | 2,299 | -28% | 0 | 0 | — |
case-19 | fail→pass | 29,553 | 7,261 | -75% | 1 | 1 | 0% | 2,324 | 1,636 | -30% | 0 | 0 | — |
case-20 | pass→pass | 17,787 | 10,795 | -39% | 1 | 1 | 0% | 2,507 | 2,079 | -17% | 0 | 0 | — |
case-21 | pass→pass | 12,857 | 6,189 | -52% | 1 | 1 | 0% | 1,864 | 1,475 | -21% | 0 | 0 | — |
case-22 | pass→pass | 16,240 | 12,084 | -26% | 1 | 1 | 0% | 2,729 | 2,135 | -22% | 0 | 0 | — |
case-23 | pass→pass | 23,467 | 14,871 | -37% | 1 | 1 | 0% | 3,702 | 3,116 | -16% | 0 | 0 | — |
case-24 | pass→pass | 9,487 | 4,272 | -55% | 1 | 1 | 0% | 1,381 | 1,133 | -18% | 0 | 0 | — |
DecimalAI ran this skill against gemini-3.6-flash twice over the same eval suite — once with the skill loaded and once without — and compared the two runs case by case. 24 cases were attempted. The headline lift of +17 percentage points is the difference between those two pass rates over the 24 comparable cases. 1 case got worse with the skill loaded, and it is included in that figure.
Without the skill loaded, the model failed this case. With it loaded, the same prompt on the same model passed. This is one improved case from the latest verified run; every case, including any that regressed, is in the table above.
Other measured skills in the registry, with their headline benchmark lift.