--
There are many aspects to validator performance on Solana, and different metrics are important to different people. For users of the Solana network, throughput (transactions per second) and latency (how quickly a transaction lands) are key metrics. In this article we’ll dive into two factors that affect those: skip rate and block size. We’ll explain how Chorus One is able to outperform both network average and the superminority on these metrics. If all validators performed as well as Chorus One on these metrics, Solana would be able to process 11.4% more transactions per second.
As a Solana user, when you submit a transaction, you want it to be included in the chain as quickly as possible, as cheaply as possible. When the chain can process only a limited amount of transactions per second, that means that only users who are willing to pay high priority fees can get their transaction included. When the chain can process more transactions per second, transaction processing capacity becomes less scarce, and transaction fees go down. Solana’s throughput is determined by the validators that make up the network, so for good network performance, it is important to delegate to a validator that performs well.
For this article we look at the month of July 2024. All metrics are reported over the period from midnight July 1st until midnight August 1st in the UTC time zone. (Slot 274965076 until 280826904, for those who want to reproduce our findings.)
In this article we contrast Chorus One against two groups of validators: the entire network (including Chorus One), and the superminority. The superminority is the smallest set of validators that together control more than one third of the stake. We use the superminority from epoch 650, the final epoch in July. It consists of the top 19 validators by stake.
In the Solana network, validators periodically have a duty to produce blocks. Before the start of the epoch, the protocol sets the leader schedule, which determines when every validator has to produce a block. Validators with more stake get assigned more blocks to produce.
If all goes well, when a validator’s turn comes to be the leader, the validator produces a block. The chain grows by one block, and users’ transactions get included. When things don’t go well, the leader fails to produce a block, or the block may not be accepted by the other validators. When the leader fails to extend the chain, this is called a skip, and the fraction of blocks skipped out of blocks assigned in some period of time is called the skip rate. Skips are bad for users of the network, because during a skip, no transactions get processed. Skips lower the throughput of the chain, and delay when transactions get processed. A lower skip rate is therefore better.
A validator can skip for multiple reasons. Of course a validator that is offline will be unable to produce a block, but even when it is online and produces a block, that can still result in a skip. For example, the validator could have been slightly late, and the network has already moved on, assuming the validator skipped its duty. Many of the factors that affect skip rate are directly or indirectly under the validator’s control, but some amount of skipping is inevitable in a decentralized network. During times of high activity, skip rate is generally higher network-wide than during quiet periods. Therefore, the skip rate is not meaningful in isolation, but comparing skip rate between validators is one way to judge their performance.
Over July 2024, Chorus One achieved a skip rate of 2.03%, while the network-wide skip rate was 5.19%. This means that average Solana validators fail to produce their blocks more than 2.5 times as often as Chorus One.
Maybe network average is not a fair comparison though? It may be the case that a few bad validators are pulling up the average. So let’s look at the superminority, the top validators by stake. This relatively small set of validators has the responsibility to produce one third of the blocks, so its influence on the chain’s throughput is large. Over July 2024, the superminority together achieved a skip rate of 5.68%, which is even worse than network average. Superminority validators fail to produce their blocks almost 3× as often as Chorus One.
The Solana network is effectively leaving 3.3% of its blocks on the table by keeping stake delegated to validators with high skip rates.
Aside from skip rate, a major factor for throughput is the number of transactions that every block contains. When blocks can fit more transactions, the throughput of the chain goes up. When validators are able to build larger blocks, fewer user transactions have to be postponed to the next block, so latency goes down. Furthermore, more capacity means lower transaction costs.
Over July 2024, blocks produced by Chorus One contained on average 1696.2 transactions. (This includes vote transactions that contribute to Solana’s consensus mechanism.) The network-wide average over this period was a mere 1573.3 per block. This means that Chorus One includes 7.8% more transactions per block than average validators.
Again, let’s compare this to the validators with the greatest responsibility and disproportionate impact on chain-wide throughput: the superminority. Here we see that with 1640.6 transactions per block, the superminority does outperform the network average, but nonetheless Chorus One outperforms the superminority by 3.4%.
This means that the Solana network is effectively leaving a 7.8% throughput boost on the table, by keeping stake delegated to low-performing validators. This number is only for produced blocks, we don’t count skips as zero transactions per block. This means that the 7.8% boost would come on top of the 3.3% skip rate boost. Combined, this means that Chorus One achieves 11.4% more transactions per second than average validators.
Why is Chorus One able to process 11.4% more transactions per second than other validators? As is often the case with performance optimization, there is no single trick, but if you stack enough small optimizations, the combined result can be substantial. A few of the techniques we use:
In this article we highlighted two key Solana performance metrics that matter for users of the network: skip rate and block size. Lower skip rates and larger block sizes mean that users can get their transactions included faster and for a lower fee. These two metrics contribute to how many transactions per second Solana can process. Through multiple optimizations and operational practices, Chorus One achieves 11.4% more transactions per second than the network average. If all delegators would delegate to validators who perform as well as Chorus One, Solana would be able to process 11.4% more transactions per second.
About Chorus One
Chorus One is a leading institutional staking provider, securing over $3 billion in assets across 60+ Proof-of-Stake networks. Since 2018, Chorus One has been a trusted partner for institutions, offering enterprise-grade solutions, industry-leading research, and investments in cutting-edge protocols.
