Skip to main content

Performance Metrics

Various behaviors of the EigenDA networks can be described along the following performance metrics.

Blob Confirmation Latency​

The end-to-end latency of EigenDA is measured as the time between the following two events:

  • When a client disperses a blob via the DisperseBlob() endpoint on the EigenDA Disperser API.
  • When the blob has been dispersed to operators within a batch, such that a DA batch certificate containing aggregated BLS signatures has been posted to the EigenDAServiceManager contract.

Blob confirmation latency is configurable within the EigenDA Disperser by controlling how often the disperser aggregates blobs into a batch. The exact setting depends on the network and can be found on the network's description.

Encoded System Throughput​

The total volume of data per unit time written to EigenDA operator set is known as the Encoded System Throughput (ECS). The "encoded" part denotes that this data is not the raw input data that was sent to the disperser by the client. Rather the encoded data is an extended version of the input data which is constructed using Reed-Solomon erasure coding techniques. Erasure coding the input data introduces redundancy to the data such that some of it can be lost to offline or dishonest operators without the raw input data becoming unavailable.

When a client dispersers a blob, two configuration parameters in the disperser determine how the data is encoded and how much of the ECS the blob will consume.

  1. Quorum Threshold is the minimum percentage of stake that must attest in order to consider the blob dispersal successful. As such, clients can use this setting to customize liveness requirements. For example, a lower quorum threshold means that a smaller set of operators are required to meet a dispersal request, whereas a high quorum threshold requires more operators to be available to provide liveness.

  2. Adversary Threshold is the maximum percentage of the stake which can be held by adversarial nodes before the availability of a blob is affected.

Based on chosen Adversary Threshold and Quorum Threshold, the Coding Ratio is derived as follows:

Coding Ratio = 100 / (Quorum Threshold - Adversary Threshold)

More operational figures for the disperser:

  • The maximum blob size per DisperseBlob invocation is 4 MiB.
  • The maximum throughput supported by the Disperser is approximately 10 MiB/s.
  • Blobs are stored for 14 days.

Rate Limits​

The EigenDA system will rate limit any client can post data. In the future, these rates will be determined via a system of reservation and on-demand payment. However, at the current testnet stage, the system enforces the following uniform rate limits for all users, as identified by the IP address of their request:

  • The maximum throughput for an individual Rollup user is approximately 3 KiB/s encoded.
  • For a constant amount of data, the Disperser's KZG proving system more efficiently processes a small number of large blobs compared to a large number of small blobs. Thus, we also enforce a limit on the rate at which each rollup posts blobs to the system. Currently, rollups are limited to approximately one blob per batching interval (~100 seconds).

Both of these rate limits are enforced over an interval of about 10 minutes. The user's request sent to the Disperser will be rate limited if their total throughput exceeds the above rate limits, and the user's request to DisperseBlob will return a rate limit error.