STATISTICS(SSM)

Shore Programmer's Manual - 18 September 95

NAME

statistics \- Shore Storage Manager performance information

SYNOPSIS

class sm_stats_info_t;

static rc_t
ss_m::gather_stats(sm_stats_info_t& stats, bool reset=false);

DESCRIPTION

The Shore Storage Manager keeps myriad statistics. The statistics are available to higher layers of software through the method gather_stats. Each statistic kept has a name and a manifest constant of that name for use with the generic statistics-gathering programming interface described in statistics(fc) . The names and descriptions of the statistics are listed in sections below.

All the statistics are unsigned long integers except where noted below. Below each statistic is identified by the manifest constant for the statistic and also by its member name in the structure sm_stats_info_t.

After gathering the storage manager statistics into a sm_stats_info_t, the sm_stats_info_t can be used with the w_statistics_t package for display:

sm_stats_info_t s;

rc =  gather_stats(s); 
if(!rc) {
	w_statistics_t stats << s;
	cout << stats;
}

Set the argument reset of gather_stats to true if you want the statistics to be cleared (set to zero) after they are copied.

BUFFER POOL STATISTICS

SM_rec_pin_cnt: sm_stats_info_t::rec_pin_cnt. Records pinned in the buffer pool. (This counts the pinning events, not the distinct records pinned.)
SM_rec_unpin_cnt: sm_stats_info_t::rec_unpin_cnt. Records unpinned.
SM_bt_find_cnt: sm_stats_info_t::bt_find_cnt. B-tree lookups (calls to ss_m::find_assoc());
SM_page_fix_cnt: sm_stats_info_t::page_fix_cnt. Times pages were fixed in the buffer pool.
SM_page_unfix_cnt: sm_stats_info_t::page_unfix_cnt. Times pages were unfixed.

B-TREE STATISTICS

SM_bt_insert_cnt: sm_stats_info_t::bt_insert_cnt. B-tree inserts (calls to ss_m::create_assoc());
SM_bt_remove_cnt: sm_stats_info_t::bt_remove_cnt. B-tree removes (calls to ss_m::destroy_assoc());
SM_bt_scan_cnt: sm_stats_info_t::bt_scan_cnt. Number of B-tree scans started.

LOGICAL-ID STATISTICS

These will be zero if you do not use logical-IDs. Logical IDs are stored in a B-tree index, and recently used IDs are cached in a transient cache.

SM_lid_lookups: sm_stats_info_t::lid_lookups. Times the logical-ID index was searched for an ID.
SM_lid_remote_lookups: sm_stats_info_t::lid_remote_lookups. Times the index was searched for a second time for indirect references.
SM_lid_inserts: sm_stats_info_t::lid_inserts. Logical-IDs added to the index.
SM_lid_removes: sm_stats_info_t::lid_removes. Logical-IDs removed from the index.
SM_lid_cache_hits: sm_stats_info_t::lid_cache_hits. Times a logical-ID look-up request was satisfied by the cache of recent requests, avoiding a B-tree look-up.

PAGE OPERATIONS

SM_page_alloc_cnt: sm_stats_info_t::page_alloc_cnt. Pages allocated from free pages in allocated extents.
SM_page_dealloc_cnt: sm_stats_info_t::page_dealloc_cnt. Pages deallocated. These pages are free for re-allocation in the same store.

TRANSACTION STATISTICS

SM_begin_xct_cnt: sm_stats_info_t::begin_xct_cnt. The number of transactions that were begun. This includes the number that resulted from chaining transactions.
SM_commit_xct_cnt: sm_stats_info_t::commit_xct_cnt. The number of transactions that were committed. This includes the number that resulted from chaining transactions.
SM_abort_xct_cnt: sm_stats_info_t::abort_xct_cnt. The number of transactions that were aborted.
SM_rollback_savept_cnt: sm_stats_info_t::rollback_savept_cnt. The number of requests to roll back to a save point, without rolling back the entire transaction.

LOCK STATISTICS

The following statistics measure activity in the lock table. Many of the statistics are of interest only the the developers of Shore, and are likely to be removed in future releases of Shore.

SM_unlock_request_cnt: sm_stats_info_t::unlock_request_cnt. High-level unlock requests.
SM_lock_request_cnt: sm_stats_info_t::lock_request_cnt. High-level lock requests (could have been satisfied with the lock cache or with the lock table.)
SM_lock_cache_hit_cnt: sm_stats_info_t::lock_cache_hit_cnt. The number of cache hits (avoiding request to acquire locks through the lock table).
SM_lock_acquire_cnt: sm_stats_info_t::lock_acquire_cnt. The number of times a request was made to acquire a lock through the lock table (as opposed to the lock cache).
SM_lock_head_t_cnt: sm_stats_info_t::bt_scan_cnt. The number of lock_head_t structures put in the table.
SM_lock_request_t_cnt: sm_stats_info_t::lock_request_t_cnt. The number of lock_request_t structures chained from a lock_head_t.
SM_lock_query_cnt: sm_stats_info_t::lock_query_cnt. The number of times the lock table was queried about a given lock.

The following statistics measure collisions from the lock table hash function. They are generally not of interest to the user, but they might be of interest to someone who is building a value-added server and might consider using a different hash function. Only buckets of non-zero length are counted when these statistics are computed. They are all zero when there is no transaction active, since the lock table is empty at that time. These statistics are computed when gather_stats is called; it traverses the entire lock table, so it should not be used habitually when transactions are active.

SM_lock_bucket_cnt: sm_stats_info_t::lock_bucket_cnt. The total number of buckets used. This is some fraction of the total number of buckets available, which is determined by the value of the option sm_locktablesize.
SM_lock_max_bucket_len: sm_stats_info_t::lock_max_bucket_len. The largest bucket in use.
SM_lock_min_bucket_len: sm_stats_info_t::lock_min_bucket_len. The smallest bucket in use.
SM_lock_mode_bucket_len: sm_stats_info_t::lock_mode_bucket_len. The mode of the lengths of the buckets used.
SM_lock_mean_bucket_len: float sm_stats_info_t::lock_mean_bucket_len. The mean bucket length (of the buckets used).
SM_lock_var_bucket_len: float sm_stats_info_t::lock_var_bucket_len. The variance of the bucket length (of the buckets used).
SM_lock_std_bucket_len: float sm_stats_info_t::lock_std_bucket_len. The Standard deviation of the bucket length (of the buckets used).

OPERATIONS ON LOCAL DATA VOLUMES

SM_vol_reads: sm_stats_info_t::vol_reads. Lowest-level read requests made to the `diskrw' process, which effects non-blocking disk I/O for data volumes.
SM_vol_writes: sm_stats_info_t::vol_writes. Lowest-level write requests made to the `diskrw' process, which effects non-blocking disk I/O for data volumes.

OPERATIONS ON THE LOG

SM_log_records_generated: sm_stats_info_t::log_records_generated. The number of log records written.
SM_log_bytes_generated: sm_stats_info_t::log_bytes_generated. The number of bytes written to the log.
SM_log_sync_cnt: sm_stats_info_t::log_sync_cnt. The number of times the log was `sync'-ed, meaning flushed to disk (done frequently when the log is a Unix file). (Later releases will permit the log to be a raw device.)

VERSION

This manual page applies to Version 0.9.3 (Beta) of theShore software.

SPONSORSHIP

The Shore project is sponsored by the Advanced Research Project Agency, ARPA order number 018 (formerly 8230), monitored by the U.S. Army Research Laboratory under contract DAAB07-91-C-Q518.