10.0

© 2011 ABBYY. All rights reserved.
80
7. Pending Tasks – the number of tasks taken into processing by the Processing Server, but not yet assigned to a Process-
ing Station. Such tasks are displayed in the Processing Server Monitor in a Pending state. Note that this number is not
the number of tasks queued in the Processing Server.
This number may be proportional to the total number of cores in the system, but it should not be increased indefinitely.
Admissible value: up to 2 tasks per 1 core.
8. Running Tasks – the total number of tasks being executed in the Processing Server at the moment.
9. Export Count, Import Count, Recognition Count, Other Tasks Count – the number of corresponding tasks (export,
import, recognition and others) over the past 5 minutes.
10. Export Time, Import Time, Recognition Time, Other Tasks Time – an average execution time for the corresponding
tasks (export, import, recognition and others) over the past 5 minutes.
11. Modification Server Locks Count – the number of documents that are blocked for executors. The counter applies for
the Processing Station only.
System counters
S
ometimes, insufficient performance can be caused by the used hardware. In order to determine whether the hardware complies
with the load and whether it has bottlenecks, system performance counters must be used. Counter used for different system com-
ponents are described below.
RAM
1. Memory: Available Mbytes – the amount of physical memory (RAM), in bytes, available to processes running on the
computer. RAM consists of the physical memory and a swap file. If the RAM in the system is not enough, the paging
mechanism is used which can lead to slowdown.
2. Paging File: Usage – the use of paging. Data and code in the memory are divided into pages. On Intel, one page is
equal to 4096 byte. Paging is the process of moving pages between the physical to virtual memory. Excessive move-
ment of pages from disk into memory and vice versa can lead to severe CPU load. Such a situation may look like a
problem with the processor or disk.
3. Memory: Committed Bytes – the committed memory. The memory reserved in the file pagefile.sys in case you need to
dump the contents of physical memory to disk. The amount of allocated memory of the process characterized the
amount of memory actually consumed by it. The amount of allocated memory is limited to the size of the paging file.
The limit of the amount of allocated memory in the system (Memory: Commit Limit) is determined by how much
memory can be allocated to processes without increasing the size of the paging file. The counter shows the total amount
of allocated memory for all processes that is the actual amount of memory used by the system.
4. Memory: Page Faults/sec shows how often the data are outside the working set. The working set is the physical mem-
ory (RAM) visible to a process or a program. Page faults occur when the program requests a code or data page which is
not in the working set and must be found elsewhere. Includes soft page faults and hard page faults. A soft page fault is a
situation when the program requests a page which is memory but out of working set. In this case, restoring data from
disk is not required. A hard page fault is a situation when the program requests a page which is not in the physical
memory (RAM) and must be restored from disk. Faults of this type are the best to show the presence of bottlenecks in
the memory configuration. More than 5 faults per second show that RAM should be increased.
5. Memory: Page Input/sec total number of pages read from disk to resolve hard page faults. Comparing this counter with Page
Faults/sec shows the number of soft page faults.
6. Memory: Pages/sec – total number of pages read from disk and written to disk. This is the sum of Page Outputs/sec
(the number of pages that had to be written to the disk to make room in RAM for other pages as a result of page fault)
and Page Inputs/sec. The admissible average value is 0 – 20. The Pages/sec value of more than 5 per second indicates a
bottleneck in memory configuration. Comparing this counter with Page Faults/sec gives an idea about the number of
soft page faults and hard page faults.
It is also recommended to monitor the memory occupied by the following processes:
• FlexiBrSvc.exe – the process of the Processing Server and the Processing Station. If both services are installed on
the computer (not recommended), they can be distinguished by PID specified for the corresponding services.
• w3wp.exe – IIS working processes.
• sqlserv.exe – SQL Server process
Processor
It is recommended to exclude memory and other bottlenecks that load the processor before diagnosing a bottleneck caused by
a process.