Table performance metrics

Table metrics allow drilling down and locating specific areas of application workloads that are the source of performance issues. If you notice a performance trend at the OS or cluster level, viewing table metrics can provide a more granular level of detail.

The metrics for KeyCache Hits, RowCache Hits, and SSTable Size can be viewed only on a single table at a time. Otherwise, all table metrics are available for specific tables as well as for all tables on a node. In addition to monitoring read latency, write latency and load on a table, monitor the hit rates on the key and row caches for tables that rely on caching for performance. The more requests that are served from the cache, the faster the response times. Viewing SSTable Size and SSTable Count for a specific table (or counts for all tables) can help with compaction tuning.

OpsCenter has been optimized to efficiently handle thousands of tables. If a table experiences a dramatic dip in performance, check the Pending Tasks metrics for a backup in queued operations.

Table metrics are prefaced with TBL.

  • TBL: Local Writes [cf-write-ops]

    Local write requests per second. Local writes update the table’s memtable and appends to a commitlog.

  • TBL: Local Write Latency (percentiles) [cf-local-write-latency]

    The min, median, max, 90th, and 99th percentile of the response times to write data to a table’s memtable. The elapsed time from when the replica receives the request from a coordinator and returns a response.

  • TBL: Local Reads [cf-read-ops]

    Local read requests per second. Local reads retrieve data from a table’s memtable and any necessary SSTables on disk.

  • TBL: Local Read Latency (percentiles) [cf-local-read-latency]

    The min, median, max, 90th, and 99th percentile of the response time to read data from the memtable and sstables for a specific table. The elapsed time from when the replica receives the request from a coordinator and returns a response.

  • Read Request Latency (percentiles) [read-histogram]

    The min, median, max, 90th, and 99th percentiles of a client reads. The time period starts when a node receives a client read request, and ends when the node responds back to the client. Depending on consistency level and replication factor, this may include the network latency from requesting the data’s replicas.

  • TBL: Live Disk Used [cf-live-disk-used]

    Disk space used by live SSTables. There might be obsolete SSTables not included.

  • TBL: Total Disk Used [cf-total-disk-used]

    Disk space used by a table by SSTables, including obsolete ones waiting to be garbage collected.

  • TBL: SSTables per Read (percentiles) [cf-sstables-per-read]

    The min, median, max, 90th, and 99th percentile of how many SSTables are accessed during a read. Includes sstables that undergo bloom-filter checks, even if no data is read from the sstable.

  • TBL: Cell Count (percentiles) [cf-column-count]

    The min, median, max, 90th, and 99th percentile of how many cells exist in partitions for this table.

  • TBL: Partition Size (percentiles) [cf-partition-size]

    The min, median, max, 90th, and 99th percentile of the size (in bytes) of partitions of this table.

  • TBL: Pending Reads/Writes

    The number of pending reads and writes on a table. Pending operations are an indication that Cassandra is not keeping up with the workload. A value of zero indicates healthy throughput. If out-of-memory events become an issue in your Cassandra cluster, it might help to check cluster-wide pending tasks for operations that could be clogging throughput.

  • TBL: Bloom Filter Space Used [cf-bf-space-used]

    The total size of all the SSTables' bloom filters for this table.

  • TBL: Bloom Filter False Positives [cf-bf-false-positives]

    Number of bloom filter false positives per second.

  • TBL: Bloom Filter False Positive Ratio [cf-bf-false-ratio]

    Percentage of bloom filter lookups that resulted in a false positive.

  • TBL: Bloom Filter Off Heap [cf-bf-offheap]

    Total off heap memory used by bloom filters from all live SSTables in a table.

  • TBL: Index Summary Off Heap [cf-index-summary-offheap]

    Total off heap memory used by the index summary of all live SSTables in a table.

  • TBL: Compression Metadata Off Heap [cf-compression-data-offheap]

    Total off heap memory used by the compression metadata of all live SSTables in a table.

  • TBL: Memtable Off Heap [cf-memtable-offheap]

    Off heap memory used by a table’s current memtable.

  • KeyCache Requests [key-cache-requests]

    The number of key cache requests per second. This metric only applies to SSTables created by DSE versions earlier than 6.0.

  • KeyCache Hits [key-cache-hits]

    The number of key cache hits per second. This will avoid possible disk seeks when finding a partition in an SSTable. This metric only applies to SSTables created by DSE versions earlier than 6.0.

  • KeyCache Hit Rate [key-cache-hit-rate]

    The percentage of key cache lookups that resulted in a hit. This metric only applies to SSTables created by DSE versions earlier than 6.0.

  • RowCache Requests [row-cache-requests]

    The number of row cache requests per second.

  • RowCache Hits [row-cache-hits]

    The number of row cache hits per second.

  • RowCache Hit Rate [row-cache-hit-rate]

    The percentage of row cache lookups that resulted in a hit.