Approved Storage Providers

Providers currently approved on-chain for the Warm Storage service. All metrics on this page pool every probe across the approved set.

Providers in this set
Cello
Criterion Value Samples Data Storage Success Storage Proving Fault Retrieval Success
No
98.6 100.0% 0.0% 99.8%
ezpdpz-main2
Criterion Value Samples Data Storage Success Storage Proving Fault Retrieval Success
No
98.5 98.8% 0.0% 99.9%
Base
Criterion Value Samples Data Storage Success Storage Proving Fault Retrieval Success
No
98.0 99.5% 0.0% 99.8%
infrafolio-mainnet-pdp
Criterion Value Samples Data Storage Success Storage Proving Fault Retrieval Success
No
97.5 99.3% 2.7% 99.6%
Mongo2Stor Mainnet
Criterion Value Samples Data Storage Success Storage Proving Fault Retrieval Success
No
96.4 96.6% 1.8% 97.2%
ezpdpz-main
Criterion Value Samples Data Storage Success Storage Proving Fault Retrieval Success
No
95.5 91.5% 2.6% 99.4%

6 providers. Aggregated over the last 7 days. Hover a value to see the underlying counts. · Score methodology

Approval Criteria

Workflow Results

Workflow Results

Each gauge shows a methodology criterion's success rate over the selected time range, pooled across all selected providers at the probe location. The arc is split at the approval threshold into a red fail zone and a green pass zone; the value reads green once the rate clears the threshold (and "no data" when there are no samples). The figure under each gauge is the number of samples the rate is based on, summed across the providers. The per-provider minimum-sample requirement is an approval criterion for individual providers and does not apply to a pooled set. Data Retention is shown as success (100 − fault), so all gauges read "higher is better".

For more information about the approval criteria, see Approval Criteria.

Whether the providers met the minimum approval criteria during the last 7 days.

Daily Performance

Daily Performance

Each cell is one day, each row an approval criterion, measured across all selected providers pooled at the probe location. A cell is green when the pooled success rate cleared the criterion's approval threshold that day (Data Storage ≥ 97%, Retrieval ≥ 97%, Storage Proving ≥ 99.8% — shown as success, i.e. 100 − fault), red when it fell short, and gray when no checks ran. Hover a cell for the day's rate and sample count.

Cells are a threshold check on the daily rate alone, the minimum-sample requirements that gate approval don't apply per day, so days with few checks can swing on a single failure.

For more information about the approval criteria, see Approval Criteria.

Daily approval-criteria status over the last 90 days

Approval Criteria Trends

Daily Data Storage Success

Daily Data Storage Success

The share of data-storage checks that succeeded each day, for the selected provider and probe location. A check succeeds only when dealbot uploads a piece, gets it added and confirmed on-chain, and confirms the data is discoverable and retrievable, all within a bounded amount of time

The bars show the daily success rate; the overlaid line shows how many checks were run that day (the samples used to compute the rate), plotted against the right axis. A high rate over very few samples is less meaningful than the same rate over many.

For more information about the data storage checks, see Data Storage.

Loading ...

Daily Retrieval Success

Daily Retrieval Success

The share of retrieval checks that succeeded each day, for the selected provider and probe location. A retrieval check succeeds when dealbot fetches a previously stored piece back from the provider without error or timeout.

The bars show the daily success rate; the overlaid line shows how many retrieval checks were run that day (the samples used to compute the rate), plotted against the right axis.

For more information about the retrieval checks, see Retrievals.

Loading ...

Daily Storage Proving Success

Daily Storage Proving Success

The share of proving periods that were proven (not faulted) each day, for the selected provider and probe location. Computed from per-poll deltas of the cumulative proving-period counters, so each day reflects the periods that came due that day. The dashed bars' overlaid line shows the number of periods (samples) each day. Approval requires ≥ 99.8% (i.e. ≤ 0.2% faulted).

For more information about the storage proving probes, see Data Retention.

Loading ...

Statistical Significance

Cumulative success rate with its 95% confidence interval. The band narrows as checks accumulate.

Data Storage Success [95% CI]

Data Storage Success [95% CI]

The cumulative data-storage success rate over the selected window for this provider and probe location, shown with its 95% confidence interval (CI). Each point is the running rate across every check up to that day; the shaded band is the Wilson score interval for that rate.

The band starts wide and narrows as checks accumulate. Once it sits entirely above (or below) the 97% approval threshold (the dashed line), the provider is statistically distinguishable from the threshold, which can happen well before any fixed sample target is reached. The dashed line on the right axis tracks the cumulative number of checks the rate is based on.

For more information about the data storage checks, see Data Storage.

Loading ...

Retrieval Success [95% CI]

Retrieval Success [95% CI]

The cumulative retrieval success rate over the selected window for this provider and probe location, shown with its 95% confidence interval (CI). Each point is the running rate across every retrieval check up to that day; the shaded band is the Wilson score interval for that rate.

The band starts wide and narrows as checks accumulate. Once it sits entirely above (or below) the 97% approval threshold (the dashed line), the provider is statistically distinguishable from the threshold — which can happen well before any fixed sample target is reached. The dashed line on the right axis tracks the cumulative number of checks the rate is based on.

For more information about the retrieval checks, see Retrievals.

Loading ...

Storage Proving Success [95% CI]

Storage Proving Success [95% CI]

The cumulative data-retention success rate over the selected window for this provider and probe location, shown with its 95% confidence interval (CI). Each point is the running share of proving periods that were proven (not faulted) up to that day; the shaded band is the Wilson score interval for that rate. Retention is reported as cumulative proving-period counters, so the daily counts are derived from per-poll deltas.

The band starts wide and narrows as periods accumulate. Once it sits entirely above (or below) the 99.8% approval threshold (the dashed line is equivalent to ≤ 0.2% faulted), the provider is statistically distinguishable from the threshold, which can happen well before any fixed sample target is reached. The dashed line on the right axis tracks the cumulative number of proving periods the rate is based on.

Caveat: the confidence interval assumes independent trials, but retention faults tend to cluster. This means a single outage faults many consecutive periods at once. The effective sample size is therefore smaller than the raw period count, so the band is somewhat optimistically narrow and should not be read as exact at the boundary.

For more information about the storage proving probes, see Data Retention.

Loading ...

Data Storage

Data Storage Pipeline

Data Storage Pipeline

Each column is a single data-storage check for the selected provider and probe location, ordered oldest (left) to newest (right). Each row is a stage of the storage pipeline:

  • Upload — dealbot uploaded the piece to the provider.
  • On-chain — the piece was added and confirmed on-chain.
  • Discoverability — the piece's root CID was verified findable via IPNI.
  • Retrieval — the freshly stored piece was fetched back from the provider using the /ipfs endpoint.

A cell is green when the stage succeeded, red when it failed, and gray when it was skipped or never reached (for example, on-chain confirmation is skipped when the upload itself failed). Retrieval checks started being recorded later than the rest of the pipeline, so probes from before then show a gray (skipped) retrieval cell rather than a failure. Hover a cell for the probe timestamp and stage detail.

For more information about the data storage checks and the state progression, see Data Storage.

Stage outcomes of the data storage pipeline

Latencies

Upload Latency

Upload Latency

Daily p50, p90 and p95 of how long the upload took, in seconds, for the selected provider and probe location. Measured from when the dealbot starts uploading the piece to the provider until the upload finishes.

The dashed line shows the daily sample count. Tail percentiles (p90/p95) capture the slow uploads that median latency hides.

For more information about the data storage checks, see Data Storage.

Upload latency from Europe for a 10MiB file

Download Latency (TTFB)

Download Latency (TTFB)

Daily p50, p90 and p95 of the retrieval time to first byte (TTFB) or time to last byte (TTLB) for the selected provider and probe location, in seconds. Computed over successful retrieval checks only. Failed or timed-out retrievals have no first-byte nor last-byte time.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slow retrievals that median latency hides.

For more information about the retrieval checks, see Retrievals.

Download latency from Europe for a 10MiB file

IPNI Verification

IPNI Verification

Daily p50, p90 and p95 of how long IPNI verification took, in seconds, for the selected provider and probe location. Measured from when the piece is advertised on IPNI until its root CID is confirmed findable via IPNI [details].

The dashed line shows the daily sample count. Tail percentiles (p90/p95) capture the slow verifications that median latency hides.

For more information about the data storage checks, see Data Storage.

Loading ...

Pull Acknowledgement Latency

Pull Acknowledgement Latency

Daily p50, p90 and p95 of the pull acknowledgement latency for the selected provider and probe location, in seconds. Measures how long the provider took to accept the pull request after dealbot submitted it, the time until the provider confirms it will fetch the synthetic test piece. Computed over pull checks where the request was acknowledged.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slow acknowledgements that median latency hides.

For more information about the pull checks, see Pull Check.

Loading ...

Pull Started Latency

Pull Started Latency

Daily p50, p90 and p95 of the pull started latency for the selected provider and probe location, in seconds. Measures how long after the request the provider began pulling, the moment it first reaches out to fetch the synthetic test piece from dealbot. Checks where the provider never started pulling are excluded.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slow starts that median latency hides.

For more information about the pull checks, see Pull Check.

Loading ...

Pull Completion Latency

Pull Completion Latency

Daily p50, p90 and p95 of the pull completion latency for the selected provider and probe location, in seconds. Measures the full duration from the dealbot submitting the request to the provider reporting that the pull reached a terminal state (finished or failed). Computed over pull checks that reached a terminal state.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slow completions that median latency hides.

For more information about the pull checks, see Pull Check.

Loading ...

Sampled Retrieval Latency (TTFB)

Sampled Retrieval Latency (TTFB)

Daily p50, p90 and p95 of the sampled retrieval time to first byte (TTFB) or time to last byte (TTLB) for the selected provider and probe location, in seconds. Computed over successful sampled retrieval checks only.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slow retrievals that median latency hides.

For more information about sampled retrieval checks, see Sampled Retrieval.

Sampled retrieval latency from Europe

Sampled Retrieval IPNI Verification

Sampled Retrieval IPNI Verification

Daily p50, p90 and p95 of how long IPNI verification took, in seconds, during sampled retrieval checks for the selected provider and probe location.

The dashed line shows the daily sample count. Tail percentiles (p90/p95) capture the slow verifications that median latency hides.

For more information about sampled retrieval checks, see Sampled Retrieval.

Loading ...

Throughput

Upload Throughput

Upload Throughput

Daily p50, p90 and p95 of upload throughput (MiB/s) for the selected provider and probe location. Computed as the file size divided by the upload duration, over successful storage deals only.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slowest uploads that median throughput hides.

For more information about the data storage checks, see Data Storage.

Upload throughput from Europe for a 10MiB file

Download Throughput

Download Throughput

Daily p50, p90 and p95 of retrieval throughput (MiB/s) for the selected provider and probe location. Computed as the bytes retrieved divided by the total transfer time, over successful retrieval checks only.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slowest retrievals that median throughput hides.

For more information about the retrieval checks, see Retrievals.

Download throughput from Europe for a 10MiB file

Pull Throughput

Pull Throughput

Daily p50, p90 and p95 of pull throughput (MiB/s) for the selected provider and probe location, over successful pull checks only. Measures how fast the provider transferred the synthetic test piece while pulling it from dealbot.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slowest pulls that median throughput hides.

For more information about the pull checks, see Pull Check.

Loading ...

Sampled Retrieval Throughput

Sampled Retrieval Throughput

Daily p50, p90 and p95 of sampled retrieval throughput (MiB/s) for the selected provider and probe location. Computed as bytes retrieved divided by the total transfer time, over successful sampled retrieval checks only.

The dashed line shows the number of samples each day, plotted against the right axis. Tail percentiles (p90/p95) capture the slowest retrievals that median throughput hides.

For more information about sampled retrieval checks, see Sampled Retrieval.

Sampled retrieval throughput from Europe

Pull Status

The pull success rate and related metrics are no approval criteria.

Daily Pull Success

Daily Pull Success

The share of pull checks that succeeded each day, for the selected provider and probe location. In a pull check dealbot publishes a synthetic test piece at a public address and asks the provider to fetch ("pull") it. The check succeeds when the provider pulls the piece, reports that it finished, and dealbot then re-downloads the piece from the provider and confirms the contents match, all within the allotted time. A timeout, a rejected request, or a failed content verification counts as a failure.

The bars show the daily success rate; the overlaid line shows how many pull checks were run that day (the samples used to compute the rate), plotted against the right axis.

For more information about the pull checks, see Pull Check.

Loading ...

Sampled Retrieval

Sampled retrieval results are no approval criteria.

Daily Sampled Retrieval Success

Daily Sampled Retrieval Success

The share of sampled retrieval checks that succeeded each day, for the selected provider and probe location. Unlike the regular retrieval check, a sampled retrieval check fetches real tenant data — not dealbot's own test corpus.

The bars show the daily success rate; the overlaid line shows how many sampled retrieval checks were run that day (the samples used to compute the rate), plotted against the right axis.

For more information about sampled retrieval checks, see Sampled Retrieval.

Loading ...