The general concept of sharding in computing is to break up the storage of data into smaller, more manageable chunks. Sharding is useful when an unsharded system gets too large and starts to see degradation in performance. Through sharding, the storage and processing load is split over multiple computers, and so the system as a whole is more scalable.
So what does sharding mean in the context of DLT or blockchain systems?
For DLTs, including blockchains, it’s a similar concept. These systems process transactions. As the purpose of a DLT is to come to consensus on the state of a distributed ledger stored across multiple nodes (computers), those computers by definition have to have the same view on the current state of the ledger to come to consensus on transactions to update that state. In a DLT or blockchain without sharding, this means that all nodes must store and process the entire state of the ledger; the system as a whole, even though it is distributed over multiple computers, can thus only process transactions as quickly as a single computer can.
This is already causing major bottlenecks. Imagine the entire world’s traffic running through a single lane! The Bitcoin blockchain, for example, can only process around 7 transactions per second (TPS) in total. Even the fastest single lane (or single shard) blockchains today can “only” process ~65,000 TPS. This is still too low to fulfill all the world’s transaction needs by several orders of magnitude.
Sharding is thus needed if DLTs are to ever truly scale. For blockchains, this means breaking the ledger into multiple parts that run in parallel. Some implementations even use entirely separate blockchains.
But this introduces a new problem. A very simple approach of sharding the ledger into separate independent parts means that transactions are no longer composable. Composability is critical for DeFi, as DeFi dApps are most useful when they can be composed together. So many proposed sharding approaches, while offering theoretical throughput gains, fall apart when considering the real-world intended DeFi use case for these networks
Radix will solve this with its Xi’an network release with a unique approach to sharding. Radix Xi’an will “pre-shard” its ledger into 2^256 shards, and all transactions will be atomically composable across-shard by default through the use of Radix’s unique consensus “braiding” approach introduced in the sharded form of Cerberus. Radix will thus be the only decentralized network that can scale without friction.
Don’t worry. We’ve explained all of this with pictures as part of the Cerberus Infographic Series. Check out:
Cerberus Infographic Series - Episode 3 - Why Blockchains Can’t Scale - for why blockchains can’t scale without breaking atomic composability
Cerberus Infographic Series - Episode 5 - The Shardspace and Validator Sets - for an explanation of Radix’s approach to sharding