Performance counters

(Source: sentry.com, msdn.com)

Object		Counter	Rule of Thumb	Notes
Paging File	% Usage		< 70%		The amount of page file currently in use. High page file usage indicates memory shortages that are so severe they must be spilled to the I/O subsystem, resulting in a serious performance bottleneck.
Processor	%Processor Time		<= 80%		The amount of total CPU in use across all system activities. The higher this metric is, the more likely users will be delayed.
System	Processor Queue Length		< 2		This is the single best indicator of CPU pressure on the Windows Server. It shows how many threads are currently waiting on an answer by a CPU on the server. Less than 12 per CPU is good/fair, less than 8 is better, and less than 2 is best
Physical Disk	Avg ms/read		< 8		This is metric indicates the time, in milliseconds, spent waiting to complete a read operation on average. It is also known as read latency. The most common physical bottleneck is usually the I/O subsystem. Microsoft’s guidance for HDD-based systems is that read latency of greater than 20 is poor, less than 20 is good/fair, less than 12 is better, and less than 8 is best.
Physical Disk	Avg ms/write		<1		This is metric indicates the time, in milliseconds, spent waiting to complete a write operation on average. It is also known as write latency. As with the previous metric, Microsoft’s guidance for HDD-based systems is that write latencies of greater than 4 is poor, less than 4 is fair, less than 2 is better, and 1 or less is best – on systems with a write-cache. On I/O subsystems without a write-cache, a write latency greater than 20 is poor, less than is 20 fair, less than 12 is better, and 8 or less is best.
Memory	Available Mbytes		=>300		An indicator showing the amount of physical memory, in megabytes, available to run processes on the machine. You generally want a few hundred MB available to handle the random file copy or similar activity needed for system administration.
SQL Server: Memory Manager	Memory Grants Pending		~0		This counter shows the current number of processes waiting for a workspace memory grant within SQL Server. You want this value to be close to zero most of the time. If it is frequently greater than zero, then SQL Server needs more memory or is running T-SQL code that is wasteful of memory.
SQL Server: SQL Statistics	Batch Requests/Sec		Proportionate to workload		This shows the overall busyness of a SQL Server, that is, the number of batch requests the database engine receives per second. High-end servers running modern versions of SQL Server can sustain 10’s or even 100’s of thousands of batch requests per second on a well-tuned workload. (Note - this counter does not track all Hekaton transactions).
SQL Server: SQL Statistics	SQL Re-Compilations/sec		< 10% of SQL Compil-ations/sec		This indicator shows the number of times, per second, that T-SQL compiled code attempted to execute but had to be recompiled before completion. This number is preferably at or near zero, since recompiles can cause deadlocks and exclusive compile locks. For most workloads, it should remain in proportion to Batch Requests/sec and SQL Compilations/ sec.
SQLServer:Access Methods	Full Scans/sec				The number of unrestricted full scans. These can either be base table or full index scans.
SQLServer:Buffer Manager	Buffer Cache Hit Ratio				The percentage of pages that were found in the buffer pool without having to incur a read from disk.
SQLServer:Latches	Average Latch Wait Time				The average latch wait time, in milliseconds, for latch requests that had to wait. If this number is high, your server might have resource limitations.