[{"id":"cassandraintermediate-001","courseId":"cassandra-intermediate","domain":"data-modeling","topic":"Partition size counted in cells","prompt":"A team is checking whether a table's partitions are too wide. Under Cassandra's sizing guidance, what is partition size measured in?","hint":"Think about the unit the hard limit is written in.","explain":"A partition can be too wide long before it is too heavy in bytes, so sizing starts by choosing a counting unit. The sizing formula is built from rows and per-row entries, and the hard ceiling uses that unit.","source":"https://cassandra.apache.org/doc/latest/cassandra/developing/data-modeling/data-modeling_refining.html","interaction":{"type":"single"},"options":[{"id":"a","text":"The number of bytes the partition takes on disk","whyNot":"Disk bytes are estimated separately as a size-on-disk figure, not as the partition size measure."},{"id":"b","text":"The number of rows, with each row counted once","whyNot":"Rows are the main driver of size, but the measure is cells; one row can hold several cells."},{"id":"c","text":"The number of cells (values) stored in the partition","why":"The guidance measures partition size by the cells, or values, stored in the partition."},{"id":"d","text":"The number of partitions stored on each node","whyNot":"Partitions per node is a capacity question; it does not measure how wide one partition is."}],"answer":"c","explanation":"Partition size is measured by the number of cells (values) stored in the partition, and the 2 billion hard limit is counted the same way.","concept":"Partition size is measured by the number of cells (values) stored in the partition, and the 2 billion hard limit is counted the same way.","rev":1,"updatedAt":"2026-10-26","kind":"scenario","takeaways":["Size a partition in cells first; bytes on disk are a separate estimate."]},{"id":"cassandraintermediate-002","courseId":"cassandra-intermediate","domain":"data-modeling","topic":"Per-partition cell ceiling","prompt":"Because Cassandra's hard limit is 2 billion cells per partition, a partition can keep growing to that limit with no performance cost.","hint":"Compare a hard ceiling with the point where problems begin.","explain":"A hard limit only says how large a partition the database will accept. The work a node does on a partition grows with the cells it holds, so sizing asks how that work scales, not only what is accepted.","source":"https://cassandra.apache.org/doc/latest/cassandra/developing/data-modeling/data-modeling_refining.html","interaction":{"type":"true-false"},"options":[{"id":"a","text":"True","whyNot":"The guidance warns you will likely hit performance issues before reaching 2 billion cells, so the ceiling is not a safe target."},{"id":"b","text":"False","why":"Performance issues usually appear well before the 2 billion cell limit, so the limit is not the target for a healthy design."}],"answer":"b","explanation":"The 2 billion cell figure is a hard limit, but performance issues usually appear before a partition reaches it.","concept":"The 2 billion cell figure is a hard limit, but performance issues usually appear before a partition reaches it.","rev":1,"updatedAt":"2026-10-26","kind":"scenario","takeaways":["Treat the 2 billion cell limit as a ceiling you never design toward."]},{"id":"cassandraintermediate-003","courseId":"cassandra-intermediate","domain":"data-modeling","topic":"Adding a column to the partition key","prompt":"One hotel's partition holds every room record for two years and is growing too large. Which change does the sizing guidance recommend to split it?","hint":"Think about what decides which partition a row lands in.","explain":"A partition lives as one unit on a single node, so one overgrown group becomes a bottleneck. A split changes the identifying part of each row so rows fan out across more groups.","source":"https://cassandra.apache.org/doc/latest/cassandra/developing/data-modeling/data-modeling_refining.html","interaction":{"type":"single"},"options":[{"id":"a","text":"Raise the cell ceiling for the table","whyNot":"The 2 billion cell figure is a hard limit, so changing the ceiling does not split any data."},{"id":"b","text":"Add a column to the clustering columns only","whyNot":"Clustering columns order rows inside one partition, so adding one leaves the partition itself unsplit."},{"id":"c","text":"Add a column to the partition key","why":"The guidance's technique for splitting a large partition is to add an additional column to the partition key."},{"id":"d","text":"Remove the static columns from the table","whyNot":"Static columns hold one shared value per partition; removing them does not split the rows into more partitions."}],"answer":"c","explanation":"The technique for splitting a large partition is to add an additional column to the partition key.","concept":"The technique for splitting a large partition is to add an additional column to the partition key.","rev":1,"updatedAt":"2026-10-26","kind":"scenario","takeaways":["Split a too-wide partition by widening its partition key, not its clustering columns."]},{"id":"cassandraintermediate-004","courseId":"cassandra-intermediate","domain":"data-modeling","topic":"Time bucketing of partitions","prompt":"A table stores room availability per hotel per night, and each hotel's partition is growing too large. The application never searches one room's availability across dates. Which TWO partition key changes does the guidance name as most suitable?","hint":"Weigh the split each change makes against the queries the application actually runs.","explain":"Bucketing keeps a group near a target size by choosing a grouping value of the right grain. A finer grain makes many tiny groups and a coarser one leaves large groups, so how many rows each value collects decides it.","source":"https://cassandra.apache.org/doc/latest/cassandra/developing/data-modeling/data-modeling_refining.html","interaction":{"type":"multiple","tellCount":true},"options":[{"id":"a","text":"Add the date column to the partition key so each partition covers one night","why":"The guidance names this first design approach as most suitable, though it gives significantly smaller partitions, perhaps too small."},{"id":"b","text":"Add a month column, as an integer, to the partition key so each partition covers one month","why":"The guidance buckets by month to keep partitions moderate, and the month column still groups related dates."},{"id":"c","text":"Raise the replication factor so each partition is stored on more nodes","whyNot":"Replicas add copies of the data; they do not make any single partition smaller."},{"id":"d","text":"Add room_id to the partition key so each partition covers one room across all dates","whyNot":"The guidance offers room_id only to keep a wide design, and it is less suitable because no query searches one room's availability."},{"id":"e","text":"Add the month column to the clustering columns instead of the partition key","whyNot":"Clustering columns only order rows within a partition, so they do not split the partition."},{"id":"f","text":"Store each room's history in a collection inside the partition","whyNot":"A collection keeps its items inside the same partition, adding to its size rather than splitting it."}],"answer":"ab","explanation":"The guidance ranks adding the date, or a month column, to the partition key as the most suitable splits; room_id only preserves a wide design and no query needs one room.","concept":"The guidance ranks adding the date, or a month column, to the partition key as the most suitable splits; room_id only preserves a wide design and no query needs one room.","rev":1,"updatedAt":"2026-10-26","kind":"scenario","takeaways":["Bucketing a time column into the partition key keeps partitions moderate-sized."]},{"id":"cassandraintermediate-005","courseId":"cassandra-intermediate","domain":"data-modeling","topic":"Row count driving partition size","prompt":"The guidance estimates a table's size on disk by adding up four terms of a formula. Put the four terms in the order the guidance works through them.","hint":"Follow the ordinal words the guidance uses to count its terms.","explain":"Estimating disk space adds up pieces that scale differently. Some are fixed, some are multiplied by a count, and some are a flat overhead on each unit stored. Notice which pieces get multiplied.","source":"https://cassandra.apache.org/doc/latest/cassandra/developing/data-modeling/data-modeling_refining.html","interaction":{"type":"sort","items":[{"text":"Partition key bytes, counted once","why":"The first term sums the sizes of the partition key columns; it is a fixed amount, not multiplied by rows."},{"text":"Static column bytes, counted once","why":"The second term sums the sizes of the static columns, which are not multiplied by the number of rows."},{"text":"Clustering and regular cell bytes, times rows","why":"The third term multiplies the clustering and regular column sizes by the number of rows; the guidance calls it the most involved."},{"text":"Per-cell metadata, times the value count","why":"The fourth term multiplies a metadata estimate, typically 8 bytes, by the number of values in the partition."}]},"options":[],"answer":"o:abcd","explanation":"The guidance's on-disk formula sums partition key sizes, static sizes, row-multiplied cell sizes, and per-cell metadata, in that order.","concept":"The guidance's on-disk formula sums partition key sizes, static sizes, row-multiplied cell sizes, and per-cell metadata, in that order.","rev":1,"updatedAt":"2026-10-26","kind":"scenario","takeaways":["Disk size estimates separate fixed terms from terms that scale with row count."]}]