Querying memecoin volatility on Dune using SQL and Spark

Memecoin volatility follows market liquidations

The cryptocurrency market experienced a massive phase where liquidations exceeded $19 billion in a single session, causing Bitcoin to fall to $104,782 from a high of $122,574 while Ether slipped to $3,436. The memecoin market plunged 4% with a total capitalization of $53.4 billion, although large gainers were recorded as users flocked around specific tokens. Dogecoin slumped 0.5% with $3.1 billion in trading volume. Shiba Inu recorded 12.8% exits this week. PEPE, Dogwifhat, and BONK saw outflows of 5.5%, 2.5%, and 1.8%. Solana memecoins plunged 11% with $2.1 billion in trading volume. Slerf declined 14%. Potato notched a 273% rise while BobaCat and Poncho saw upticks of 130% and 66%. WOJAK surged 87% in 24 hours. The total memecoin market capitalization surged close to $69 billion. Dogecoin showed slight momentum in the last two days but still trades at weekly losses of 9.2%. A report from JP Morgan shows that bull run expectations tanked compared to levels at the start of the year. You should understand how quickly these assets move between fear and froth.

Scaling SQL with columnar data

Dune transitioned from a PostgreSQL database to an instance of Apache Spark hosted on Databricks. This system uses a column-oriented approach instead of a row-oriented one. The database reads data from storage into memory to return query results. In Postgres, indexes force the database to look at a structured subset, but this approach does not scale for massive datasets. In DuneV2, the platform uses Parquet, a hybrid approach where the database is roughly stored in a row oriented format, but individual values are stored on pages in column orientation. The transition to Databricks SQL replaces traditional indexes with column chunk level min/max values stored in parquet file footers to skip unnecessary data during a scan. A search for a transaction hash alone forces the engine to read all pages because hash strings lack useful min/max metadata for the database to skip files or chunks efficiently. If a user includes block_time or block_number in a WHERE clause, the runtime drops to 13 seconds. The engine aggregates data across millions of rows much faster because it reads only the specific column data required. The Spellbook populates decoded tables like dex.trades. These tables use a naming convention like {protocol}_{chain}.{Contract}_evt_{EventName}. A Uniswap v3 Swap event appears as uniswap_v3_ethereum.Pair_evt_Swap. Raw tables like ethereum.transactions contain the full unprocessed history.

Subscription tiers and AI tools

Dune provides different access levels based on monthly credit allocations.

The Free tier allows unlimited queries with up to a 10-minute execution delay and 2,500 credits. Additional credits cost $5.00 per 100 credits for Free users and $1.396 per 100 for Plus users. The Analyst plan costs $780 per year while the Plus plan costs $4,188 per year, providing users with significantly increased credit amounts and faster execution lanes for their custom blockchain queries. Storage caps exist at 100 MB, 1 GB, or 15 GB depending on the plan. Paid account holders access Dune Agent in beta. This chatbot uses a Modular Contextual Prompting approach to help users find tables through natural language. Plume and Noble integrations exist on the platform. The system covers 100+ chains with full transaction and event data, which includes data from Ethereum, Polygon, Optimism, Binance Smart Chain, and xDAI. Users query in the Data Hub or stream data via API, Datashare, dbt, or BI connectors. The platform’s monthly credit wallets allow for flexible usage. Can users bypass the credit system entirely?