Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

Introduction

...

For large enterprises, it may make sense to use custom-configured server hardware to host the 3VR VisionPoint systems. While we do not intend this guide to replace or override the recommendations of IT professionals, we can offer some general guide lines guidelines of best practices for managing storage that will give our customers the best possible experience with our software solutions. Note that this document will make extensive reference to different types of RAID (Redundant Array of Inexpensive Disks) configurations and partitions.

Table 1 - Minimum Recommended Storage Summary

...

The bare minimum HDD storage configuration must be four enterprise class HDD's. In this scenario, the four drives must be divided into two independent RAID 1 pairs. The first RAID 1 pair can be small enough to hold just the Windows operating system along with at least 50% additional free space. The second RAID 1 pair must be as large as possible to hold both the video information and the database file. Please see the table below:

RoleDisk DrivesConfigurationSettingsPartition(s)
Boot VolumeTwo Enterprise Class HDD's any RPM/Cache

RAID 1

(Mirror)

N/ARecommend single partition with NTFS or ReFS to hold Windows Operating System
Video/DB Volume(s)Two Enterprise Class HDD's, 7200 RPM/256MB Cache

RAID 1

(Mirror)

N/ARecommend single partition for both video and database


Table 2: Ideal Storage Summary

...

In the ideal scenario, there are separate RAID arrays backing the boot volume, video application volume and the SQL storage volume. Again, only enterprise class HDD's should be considered. This scenario presumes that the database will be stored locally instead of a dedicated Database Server. If you are using an external Database Server, you can ignore the Database Volume row.

RoleDisk DrivesConfigurationSettingsPartition(s)
Boot VolumeTwo Enterprise Class HDD's any RPM/Cache

RAID 1 (Mirror)

N/ARecommend single partition with NTFS or ReFS to hold Windows Operating System
Video VolumeTwo Enterprise Class HDD's any RPM/Cache

RAID 1 or

RAID 5

If RAID 5, default larger stripe block sizesizes (i.e. 64K) tend to give best performance. Recommend single partition

Database

Volume

Two Enterprise Class HDD's, 7200 RPM/256MB Cache

RAID 1 or

RAID 10

if RAID 10, consider minimal stripe block sizes to improve IOPSRecommend single partition


RAID Options Detail

...

It is beyond the scope of this document to try and illustrate every possible hardware/software RAID configuration. Instead, Identiv offers the following general recommendations that the above tables are based upon.

...

Some additional Windows settings can have some significant additional performance improvement potential.

Data Compression

...

Internal Identiv testing has not shown any performance gains by choosing a non-default file system allocation block size. Your experience may differ, but at the current time we do not recommend any changes to the default Windows file system allocation block size settings.

Volume Cache Flushing Option

...

Once the RAID volumes have been created and a file system created on a RAID volume, there is one last setting which can dramatically improve performance on the volume: disabling the file system write buffer flushing. This setting controls whether Windows aggressively tries to keep file system related information synchronized between what is stored in the relatively fast RAM and what is stored in the relatively slower permanent storage system. To keep the data between the two in sync, the data in RAM needs to be written out to the RAID system as often as possible in case there is a sudden and unexpected loss of power. In the case of sudden power loss, any information that was kept in RAM but not committed to the permanent storage system (i.e. the RAID drives) will be lost forever. However, this strategy to protect against potential data loss comes at the price of performance when the system is under heavy loads. But sometimes it is the cache buffer flushing which is actually contributing to the heavy system workloads. This means that one of the of the performance improving options for systems with heavy workloads is to disable this aggressive cache flushing behavior.

...