|
The Monitoring & Optimisation Section
What to Revise : Monitoring Resources, Optimising Server Performance, Optimising Network
Performance, Using Performance Optimizer.
Watch For:
- Monitoring Resources
There are three basic resources of a server to
monitor for performance – processor, disk and memory. Of these, processor is the
least likely to be the problem and as such you should only consider upgrading the
processor if there are no other bottlenecks caused by disk or memory. To monitor server
performance use Performance Monitor from the Administrative Tools menu. Some of the common
objects and counters to monitor for server performance are outlined below:
Processor
% Processor Time is used to determine how "busy" your processor is. A
value consistently over 80% may indicate a processing bottleneck.
Interrupts/Sec is used to determine how much time the processor spends dealing with
hardware requests. A high value may indicate hardware driver problems.
Memory
Pages/Sec is used to determine the total amount of paging to and from RAM. A high
value may indicate that more RAM is required.
Page Reads/Sec and Page Writes/Sec are used to breakdown the total paging into
what is going into RAM and what is coming out. If your system always has to page out prior
to paging in this indicates a bottleneck.
Physical or Logical Disk
The hard disk is divided into Physical Disk and Logical Disk. Use logical disk to
determine problems with specific volumes. Use physical disk to determine problems with
disk drives. The counters must be "turned on" before they are available, due to
the extra system overhead involved. To turn the disk counters on type DISKPERF –Y
from the command prompt (or –YE for RAID arrays) and restart your computer. To turn
the counters off type DISKPERF –N and again restart your computer.
%Disk Time is the amount of time that the hard disk or volume is busy. A high
percentage may indicate that your hard disk is a bottleneck.
When Exchange Server 5.5 is installed, a new, extended set of objects and counters is
added to the Performance Monitor application. These Exchange-specific objects allow you to
constantly monitor the individual components of an Exchange Server – MTA, Private
Information Store, Public Information Store, Directory Replication Connectors and
Messaging Connectors.
- Optimising Server Performance
The recommended configuration for optimising the
performance of an Exchange Server is to place the transaction log files on the fastest
physical disk and have the disk dedicated just to this purpose. This will ensure the
fastest write operation. To include fault tolerance, create a mirror of the
partition on another disk (RAID 1). Then, create a stripe set for the database files. This
will ensure the fastest read operation. Again, you can build-in fault tolerance,
this time by making it a stripe set with parity (RAID 5).
To manually move files to optimum locations you should take the following steps:
- Stop the Exchange Server services (from Control Panel, Services).
- Take a backup of all files in case of accidents.
- Physically move the files to their new locations (using Windows
Explorer).
- In Exchange Administrator select the File Locations tab of the
properties of the server object and update the paths to the transaction logs and database
files as necessary.
- Start the Exchange Server services (from Control Panel, Services).
Performance Optimiser can be used to assist with these procedures
and is covered in a separate section.
- Optimising Network Performance
Network performance is especially important when
there are multiple servers within a site and/or when there are multiple sites within an
organisation. To optimise performance of Exchange Server data transfers over a poor
quality network, adjust the messaging defaults values that are found on the MTA
Site Configuration properties sheet from Exchange Administrator. The parameters and an
explanation for each are given below:
RTS (Reliable Transfer Service)
| Name |
Explanation |
| Checkpoint Size |
The amount of data
(in k) transferred before inserting a checkpoint. This amount of data will be
re-transferred in the event of an error. For reliable networks increase the checkpoint
size to reduce overhead. For unreliable networks increase the checkpoint size to reduce
overall transfer sizes. |
| Recovery Timeout |
The amount of time
(in seconds) that is waited after an error for a reconnection before starting the
transmission again from the beginning. |
| Window Size |
The amount of time
(in seconds) that is waited after an error for a reconnection before starting the
transmission again from the beginning. |
| Window Size |
The number of
checkpoints to be sent before acknowledgements are received for the previously transmitted
data. If you have a slow network, increase the number to prevent early suspension of data
transfer. |
Connection Retry Values
| Name
|
Explanation |
| Max Open Retries |
The total number of
times a transmission is attempted. Decrease on a reliable network where the problem is
likely to be other things. |
| Max Transfer Retries |
The maximum number
of times a re-transmission over any one open connection is attempted. Increase for an
unreliable network with intermittent problems |
| Open Interval |
The time (in
seconds) waited before re-opening a failed connection. Set this appropriate to the pattern
of errors on your network. |
| Transfer Interval |
The time (in
seconds) waited before re-transmitting on a connection. Set this appropriate to the
pattern of errors on your network. |
Association
Parameters
| Name |
Explanation |
| Lifetime |
Time (in seconds)
that a connection is left open for re-use. Reduce on a very busy network or if you
commonly do not transmit many messages between the same servers. |
| Disconnect |
Time (in seconds)
that a connection is left open for re-use even after a disconnect message is received from
the other server. Reduce on a very busy network or if you commonly do not transmit many
messages between the same servers. |
| Threshold |
The number of
messages queued for any one association. For maximum speed of transfer reduce the
threshold. To prevent overburdening of a busy network increase the threshold. |
Transfer Timeouts
| Name |
Explanation |
| Urgent |
The time for an
urgent message (in seconds per k of message size) that the message is given to reach its
destination before the attempt is given up and an NDR message is sent. |
| Normal |
The time for a
normal message (in seconds per k of message size) that the message is given to reach its
destination before the attempt is given up and an NDR message is sent. |
| Non-Urgent |
The time for a
non-urgent message (in seconds per k of message size) that the message is given to reach
its destination before the attempt is given up and an NDR message is sent. |
4. Using Performance Optimizer
Performance Optimizer is the utility which is installed automatically in a default
installation of Exchange Server to help gain the maximum performance from the server. The
administrator is automatically prompted to run Performance Optimizer after an installation
and should also run Performance Optimizer after a change in hardware (processor, disks or
memory) or usage (change in numbers of mailboxes, public folders, site connectivity or
replication). Performance Optimizer can be run from the Exchange Server menu or
alternatively from a command line. The location and file name is
c:\exchsrvr\bin\perfwiz.exe and the /v option can be used for verbose mode, allowing the
administrator to override adjustments otherwise made automatically by the utility.
Performance Optimizer will automatically adjust placement of essentials files and
configuration of memory. In order to do this you will be asked to supply the number of
mailboxes on this server, the number of mailboxes in total for the
organisation and each role of this server within the organisation – whether it
contains mailboxes, public folders, replication connectors, etc.
The Troubleshooting Section
What to Revise
MSMail Synchronisation,
DNS Configuration, Preventing
Denial of Service Attacks, POP3 Mailbox Access,
Internet Message Formats, RPC
Binding Order.
Watch For:
- MSMail Synchronisation
If you are implementing Exchange Server 5.5 alongside an
existing MSMail post office and you wish to synchronise directories, remember that an
MSMail group will not be able to span users from more than one post office (and your
Exchange Server will also be seen as a separate post office). Exchange DLs (Distribution
Lists) do not have this limitation.
- DNS Configuration
The most likely cause of problems in sending messages using SMTP
over the Internet or an intranet is incorrect configuration of DNS. DNS must be configured
with at least one MX record for each domain. The MX record specifies the name of the
server that contains the incoming Internet Mail Service. This record in turn must
have a corresponding HOST record to be able to resolve the host’s IP address.
- Preventing Denial of Service Attacks
The following are the most common ways of preventing DOS (Denial
of Service) attacks on your mail service from the Internet: Installing and configuring a
firewall, disabling automatic replies on the Internet, disabling out of office responses
on the Internet and setting message size limits on the Internet Mail Service itself.
- POP3 Mailbox Access
A user who accesses their messages stored on the Exchange Server
both from the office using MAPI and from home using POP3 over the Internet must remember
that any messages that he or she saves into private folders will not be visible from home,
since POP3 is limited to only viewing the contents of the mailbox itself and not private
(or public) folders. To overcome this limitation you could allow the user to use IMAP4
instead of POP3 to retrieve their mail from home.
- Internet Message Formats
If a user wishes to send mail via the Internet or
intranet to Unix users via the Internet Mail Service then you must configure the Internet
Mail Service to format messages destined to the domains where the Unix users are using
UUEncode instead of MIME. You can do this from the Internet Mail page of the properties
sheet for the Internet Mail Service object in Exchange Administrator.
 |
