background image

8

 

 

Server

 

Redundancy

 

It

 

is

 

risky

 

to

 

have

 

a

 

single

 

physical

 

server

 

for

 

your

 

enterprise,

 

even

 

if

 

you

 

take

 

regular

 

backups.

  

We

 

recommend

 

you

 

to

 

take

 

steps

 

to

 

expedite

 

recovery

 

from

 

an

 

outage

 

in

 

accordance

 

with

 

an

 

established

 

Business

 

Continuity

 

and

 

Disaster

 

Recovery

 

(BCDR)

 

plan.

  

 

Hot

 

Backup

 

Databases

 

Increase

 

the

 

redundancy

 

of

 

the

 

system

 

by

 

replicating

 

the

 

Endpoint

 

Encryption

 

Object

 

Directory

 

to

 

a

 

second

 

physical

 

server.

  

A

 

dedicated

 

replication

 

tool

 

“Object

 

Directory

 

Backup”

 

which

 

is

 

optimized

 

to

 

follow

 

the

 

change

 

log

 

of

 

an

 

Endpoint

 

Encryption

 

v5

 

Object

 

Directory

 

is

 

supplied

 

with

 

the

 

product

 

suite.

  

 

In

 

this

 

case

 

set

 

up

 

a

 

resilient

 

system

 

using

 

two

 

physical

 

boxes,

 

both

 

hosting

 

Endpoint

 

Encryption

 

Servers

 

 

one

 

hosting

 

the

 

master

 

ODB

 

and

 

the

 

other

 

having

 

a

 

hot

 

backup.

  

In

 

case

 

the

 

master

 

server

 

fails,

 

the

 

Endpoint

 

Encryption

 

Server

 

on

 

the

 

second

 

backup

 

box

 

can

 

be

 

restarted

 

in

 

“master”

 

mode.

  

Then

 

rebuild

 

or

 

replace

 

the

 

affected

 

machine

 

and

 

create

 

a

 

new

 

master.

  

 

The

 

ODB

 

Backup

 

utility

 

can

 

also

 

be

 

used

 

to

 

make

 

regular

 

backups

 

of

 

the

 

ODB,

 

giving

 

further

 

recovery

 

options

 

in

 

case

 

of

 

a

 

disaster.

  

This

 

method

 

however,

 

requires

 

manual

 

interaction

 

to

 

start

 

the

 

failover.

  

 

A

 

HotBackup

 

document

 

discussing

 

this

 

scenario

 

is

 

available.

 

 

Clustering

 

Fully

 

automated

 

failovers

 

for

 

applications

 

usually

 

employ

 

a

 

cluster

 

server

 

environment.

  

Although

 

the

 

McAfee

 

Endpoint

 

Encryption

 

Object

 

Directory

 

and

 

Manager

 

can

 

run

 

on

 

a

 

cluster,

 

we

 

recommend

 

against

 

using

 

‘shared’

 

resources

 

where

 

possible.

  

As

 

per

 

McAfee

 

KB53698,

 

Windows

 

Cluster

 

environment

 

has

 

not

 

been

 

fully

 

tested

 

at

 

this

 

time

 

in

 

engineering.

  

 

Load

 

Balancing

 

Given

 

the

 

best

 

configuration

 

is

 

usually

 

a

 

single

 

high

 

performance

 

server

 

with

 

DAS

 

then

 

the

 

least

 

optimal

 

way

 

to

 

perform

 

clustering

 

is

 

to

 

put

 

the

 

Object

 

Directory

 

on

 

a

 

network

 

share

 

(NAS)

 

and

 

then

 

install

 

the

 

Management

 

Center

 

on

 

two

 

servers

 

which

 

access

 

the

 

share

 

simultaneously.

  

 

NOTE

:

 

The

 

latter

 

will

 

function,

 

but

 

it

 

will

 

be

 

significantly

 

detrimental

 

to

 

server

 

performance.

 

 

 

You

 

should

 

note

 

that

 

if

 

you

 

use

 

special

 

load

 

balancing

 

switches

 

to

 

split

 

network

 

load,

 

you

 

should

 

set

 

them

 

to

 

allow

 

each

 

client

 

active

 

connection

 

to

 

occur

 

with

 

the

 

same

 

switch

 

throughout

 

the

 

sync

 

event

 

(and

 

not

 

split/distribute

 

each

 

packet

 

during

 

a

 

single

 

sync).

 

 

Making

 

remote

 

connections

 

to

 

the

 

database

 

is

 

slower

 

than

 

local

 

connections,

 

so

 

this

 

design

 

is

 

often

 

too

 

slow

 

to

 

work

 

effectively.

  

 

If

 

DAS

 

is

 

not

 

used

 

and

 

there

 

are

 

issues

 

such

 

as

 

performance,

 

object

 

corruption

 

(especially

 

as

 

object

 

numbers

 

in

 

the

 

McAfee

 

Endpoint

 

Encryption

 

Object

 

Directory

 

increase)

 

McAfee

 

support

 

will

 

recommend

 

moving

 

to

 

DAS

 

and

 

high

 

performance

 

dedicated

 

server.

 

 

If

 

a

 

SAN

 

is

 

the

 

only

 

option

 

available,

 

please

 

note

 

SAN

 

arrays

 

can

 

prioritize

 

the

 

connections

 

to

 

the

 

physical

 

box

 

in

 

what

 

is

 

known

 

as

 

Tier

 

levels.

  

Tier

 

1

 

is

 

the

 

highest

 

priority,

 

Tier

 

3

 

is

 

the

 

lowest.

  

McAfee

 

Endpoint

 

Encryption

 

needs

 

optimal

 

disk

 

access

 

so

 

would

 

need

 

Tier

 

1

 

priority

 

with

 

dedicated

 

LUNS

 

to

 

provide

 

the

 

highest

 

speed

 

connection.

  

This

 

is

 

necessary

 

for

 

full

 

and

 

prompt

 

service

 

synchronization

 

requests

 

and

 

administration.

  

This

 

avoids

 

corrupted

 

databases,

 

objects,

 

clients

 

and

 

slow

 

administration

 

performance.

 

Running

 

on

 

SAN

 

is

 

not

 

recommended,

 

but

 

if

 

it

 

must

 

be

 

done,

 

then

 

the

 

connection

 

must

 

be

 

Tier

 

1.

  

Summary of Contents for ENDPOINT ENCRYPTION ENTERPRISE - BEST PRACTICES GUIDE

Page 1: ...1 McAfee Endpoint Encryption Enterprise Best Practices Guide November 2009...

Page 2: ...d No part of this publication may be reproduced transmitted transcribed stored in a retrieval system or translated into any language in any form or by any means without the written permission of McAfe...

Page 3: ...OF 5000 USER MACHINE OBJECT DIRECTORY 10 VIRTUAL SERVERS 10 GLOBAL DEPLOYMENTS 11 OPTIMISATION ACTIONS 11 OPTIMISATION ACTIONS OVERVIEW 12 NAME INDEXING DBCFG INI 13 WARNINGS 13 DBCFG INI 13 GROUP SI...

Page 4: ...19 CLEARING THE AUDIT 19 DELETED ITEMS CLEANUP 20 CHECKING FOR DATABASE CORRUPTION 20 WHY DOES THE DATABASE GET CORRUPTED 20 ORPHANED OBJECTS 21 RESTORE COMMANDS 21 CLEANUP COMMANDS 21 DUMP MACHINE D...

Page 5: ...e product and the environment in which it is being used before arriving at any decision on implementation strategy Calculations and figures in this guide are based on field evidence and not theoretica...

Page 6: ...a distributed way For example the Web Helpdesk component can be installed on a dedicated web server while the rest of the components are on a separate Endpoint Encryption Server However the majority o...

Page 7: ...hared Server can be used for low numbers Please see Virtual Server section in this guide Virtual hardware has to be of higher specification if resources are shared See Page 11 2000 5000 users systems...

Page 8: ...nvironment has not been fully tested at this time in engineering Load Balancing Given the best configuration is usually a single high performance server with DAS then the least optimal way to perform...

Page 9: ...s The default settings of the Communication Server limit the queue to 200 entries a balance between taking connections and processing connections After that point the connections are refused This is a...

Page 10: ...use of Virtual Servers is a result of Lack of resources dedicated to the virtual server Dynamically assigned resources to the virtual server which starves it of the necessary performance during peak p...

Page 11: ...y is necessary it is better to include endpoints from all regions in the pilot phase Optimisation Actions NOTE These are generic recommendations based on experience but not always be suitable for your...

Page 12: ...me to five minutes Disable NTFS Last Access Update with a registry change Increase the size of the NTFS Master File Table MFT with a registry change Optimize backups Exclude the Object Directory and t...

Page 13: ...smaller than 5000 systems otherwise you find the number by multiplying the number of users or systems in the database by 0 6 Example If the number of users in the database is 10 000 the Locktimeout s...

Page 14: ...lable from your McAfee representative Attribs SingleFile No If this is set to Yes the attributes for objects will be placed into a single file instead of each one having their own file Not generally u...

Page 15: ...ours NOTE A similar setting KeepAliveInterval has a default 1000 1 second this setting is correct so do not change this Last Access Time Stamp NtfsDisableLastAccessUpdate With large databases it is po...

Page 16: ...3 or 4 instead of the default value of 1 Object Directory Backup Tool Setup If you set up your Object Directory backup tool make sure it is not running too many times a day because the in between tim...

Page 17: ...ed you schedule EEPC command line tool SBADMCL to cleanup machine audit and the user audit See Endpoint Encryption Object Directory Maintenance section below File Cache on Raid Hard Drive Controller L...

Page 18: ...containing old deleted users systems and other objects and are found through the System tab in the Endpoint Encryption Manager These objects can slow searches down If these objects are needed for aud...

Page 19: ...ds please see the Endpoint Encryption Scripting Tool User Guide which is found in most normal installations of the Endpoint Encryption Manager Extracting and Clearing Audit from the Database The audit...

Page 20: ...within the Object Directory is renamed The extension of the folder is renamed from RMV to WPE With a very large database these empty removed folders can sometimes slow down searches In a test lab try...

Page 21: ...have a full backup of SBDATA before doing this Restore Commands To restore orphaned user objects back into a group use this command SBADMCL Command RestoreUsers Adminuser Admin Adminpwd mypassword Gr...

Page 22: ...ects in the DumpMaDesc log can be deleted from the database If the normal deletion doesn t work use Windows Explorer to browse to the actual location in the database and delete the physical folder In...

Page 23: ...utes servicing each client EEPC has excellent password synchronization across all the endpoint clients a user is assigned to It is therefore logical that adding thousands of users to each machine will...

Page 24: ...nate Requests This option stops the machine from entering hibernation mode Note this option is not supported in Vista With later versions of EEPC v5 x this should normally be left disabled to allow no...

Page 25: ...ppear as a fixed drive and therefore swaps with the fixed disk after booting from it It can cause recovery problems with Remove or Emergency Boot for example Alternatively use with a floppy disk drive...

Page 26: ...tion to encrypted data Using one autoboot user for too many machines Instead use more autoboot users to reduce the multiple connections and load on the autoboot user object in the database Autoboot us...

Reviews: