1. On the client machine (ie. the one you are SSH’ing from) you will need to create an SSH RSA key. So run the following command – ensure you don’t supply a password:

[root@node01 ~]# ssh-keygen
Generating public/private rsa key pair.
Enter file in which to save the key (/root/.ssh/id_rsa):
Created directory '/root/.ssh'.
Enter passphrase (empty for no passphrase):
Enter same passphrase again:
Your identification has been saved in /root/.ssh/id_rsa.
Your public key has been saved in /root/.ssh/id_rsa.pub.
The key fingerprint is:
c6:66:93:16:73:0b:bf:46:46:28:7d:a5:38:a3:4d:6d root@node01
The key's randomart image is:
+--[ RSA 2048]----+
|            .    |
|       . + o     |
|      . @ E      |
|       * & .     |
|      . S =      |
|       = + .     |
|          o      |
|         .       |
|                 |
+-----------------+

This will generate the following files:

[root@node01 ~]# cd ~/.ssh
[root@node02 .ssh]# ls -l
total 8
-rw-------. 1 root root 1675 Jul 27 15:01 id_rsa
-rw-r--r--. 1 root root  406 Jul 27 15:01 id_rsa.pub

3. On the client machine tighten up file system permissions thus:

[root@node01 ~]# chmod 700 ~/.ssh
[root@node01 ~]# chmod 600 ~/.ssh/*
[root@node01 ~]# ls -ld ~/.ssh & ls -l ~/.ssh
drwx------. 2 root root 4096 Jul 27 15:01 /root/.ssh
-rw-------. 1 root root 1675 Jul 27 15:01 id_rsa
-rw-------. 1 root root  406 Jul 27 15:01 id_rsa.pub

4. Now copy the public key to the machine you want to SSH and fix permissions (you will be prompted for the root password):

[root@node01 ~]# ssh root@node02 'mkdir -p /root/.ssh'
[root@node01 ~]# scp /root/.ssh/id_rsa.pub root@node02:/root/.ssh/authorized_keys
[root@node01 ~]# ssh root@node02 'chmod  700 /root/.ssh'
[root@node01 ~]# ssh root@node02 'chmod  600 /root/.ssh/*'

You can also use the utility ssh-copy-id to do the above steps. If you don't have scp on the remote machine you will need to install it:

[root@node01 ~]# ssh root@node02 'yum install openssh-clients'

You should now be able to ssh directory from node01 to node02 without providing a password:

[root@node01 ~]# ssh node02
Last login: Wed Jul 27 15:41:56 2011 from 10.255.5.57
[root@node ~]#

IMPORTANT

There is a bug in CentOS 6 / SELinux that results in all client presented certificates to be ignored when SELinux is set to Enforcing. To fix this simply:

[root@node01 ~]# ssh root@node02 'restorecon -R -v /root/.ssh'
restorecon reset /root/.ssh context system_u:object_r:ssh_home_t:s0->system_u:object_r:home_ssh_t:s0
restorecon reset /root/.ssh/authorized_keys context unconfined_u:object_r:ssh_home_t:s0->system_u:object_r:home_ssh_t:s0

1. cd /etc/sysconfig/network-scripts

2. vi ifcfg-eth0 so it looks like (change your MAC address accordingly):

DEVICE=eth0
BOOTPROTO=dhcp
ONBOOT=yes
HWADDR=f4:ce:46:82:55:f4

3. Then create your VLAN interface configuration. So vi ifcfg-eth0.1051:

DEVICE=eth0.1051
BOOTPROTO=dhcp
VLAN=yes
BRIDGE=virbr0
ONBOOT=yes

4. Then create your bridge interface configuration: So vi ifcfg-virbr0:

DEVICE=virbr0
TYPE=Bridge
ONBOOT=yes
DELAY=0
BOOTPROTO=dhcp

Note that TYPE must be Bridge with a capital B – otherwise it won’t work. And there you have it – when the box boots it gets a DHCP lease on eth0 and on virbr0 which is on VLAN 1051.

1. Determine the management IP addresses of the nodes in your Mercury cluster . To do this log into the SMU (eg. https://smu) as admin. Then go to Home -> Server Settings -> Cluster Configuration and record the IP addresses of the cluster nodes (eg. 192.0.2.200 and 192.0.2.201).
2. Then ssh into the SMU (eg. ssh manager@smu). Hit q to get to the command prompt.
3. At the command prompt of the SMU ssh to each cluster node individually (eg. ssh 192.0.2.200).
4. Run the following command: ver -h | grep Tachyon
5. The output of that command will determine which hardware revision you are running:
   1. Rev A hardware is identified as follows: Tachyon: QX4 1.2 (HPFC-6440C)
   2. Rev B hardware is identified as follows: Tachyon: QE4+ 1.1 (PM8036 (Rev B))

1. Work out the EVS ID:

evs list

2. Select the EVS:

evs-select <id>

3. Disable domain authentication for CIFS:

cifs-auth off

1. You will need to configure your storage system. I was using a BlueArc Mercury cluster. The configuration of the BlueArc is probably beyond the scope of this post but in essence you need to create a File System of your required size. Then assign that File System to an EVS (BlueArc terminology for a virtual storage system) with an assigned cluster node and IP address on the storage VLAN. You then need to create iSCSI Logical Units within the File System. One LUN will be required for each host. Lastly create iSCSI targets within the EVS iSCSI domain with access configuration only permitted from the host that will use it along with the LUN ID and LUN name. You will end up with is a series of Globally Unique Names that are of a finite size (eg. 500GB)  that are only accessible from a single host: iqn.2011-04.spotcloud:sc-evs-iscsi01.sc-target01.

2. Back to the CentOS side of things – make sure your interfaces are configured correctly and you can ping the storage system. I have two Virtual Connect modules in the HP C7000 enclosure – hence two interface were available. Static IPs were used on the storage network. I edited:

/etc/sysconfig/network-scripts/ifcfg-eth0
/etc/sysconfig/network-scripts/ifcfg-eth1
/etc/sysconfig/network

3. Make sure the iSCSI daemons are installed. You can do this via yum or from the original source media.

Via yum:

yum install iscsi

Via virtual media:

mount /dev/cdrom /mnt
cd /mnt/CentOS
rpm -ivh iscsi*
cd /
umount /mnt

Don’t forget to eject the virtual media.

4. Make sure iSCSI starts on boot and start the daemon:

chkconfig iscsi on
service iscsi start

5. Discover your iSCSI targets:

iscsiadm -m discovery -t sendtargets -p 10.255.4.10

The IP address is that of the storage system.

6. Delete any unnecessary iSCSI nodes:

service iscsi stop
iscsiadm -m node <nodename> -o delete
service iscsi start

The <nodename> is the UIN mentioned earlier. Sometimes you will always discover multiple nodes – so you need to configure the storage system to filter available LUNs by client source IP address.

7. Work out which device is the iSCSI node:

fdisk -l

8. Create a partition then format it:

fdisk /dev/sdb
mkfs.ext4 /dev/sdb1

7. Label the device:

e2label /dev/sdb1 /sc-node01

8. Configure the mount in /etc/fstab (note the _netdev mount option to ensure the iSCSI LUN is mounted after networking has been brought up):

LABEL=/sc-node01 /var/lib/xen/images ext3 defaults,_netdev,noatime 0 0

And that’s it – you are in business.

Lastly, if you are interested here is the Virtual Connect configuration used to configure the blades. This configures blade 1A and 1B interfaces 1 and 2. Interface 1 is assigned untagged VLAN of 1050 (eth0) and tagged VLAN 1051 (eth0.1051). Interface 2 is assigned untagged VLAN 1052 (eth1) – which is the storage network.

add profile D4-C2-B01 -NoDefaultEnetConn -NoDefaultFcConn -NoDefaultFcoeConn
add enet-connection D4-C2-B01
add enet-connection D4-C2-B01
add server-port-map D4-C2-B01:1 SC-Management VlanID=1050 Untagged=True
add server-port-map D4-C2-B01:1 SC-VM VlanID=1051
add server-port-map D4-C2-B01:2 SC-iSCSI VlanID=1052 Untagged=True
assign profile D4-C2-B01 enc0:1A

add profile D4-C2-B02 -NoDefaultEnetConn -NoDefaultFcConn -NoDefaultFcoeConn
add enet-connection D4-C2-B02
add enet-connection D4-C2-B02
add server-port-map D4-C2-B02:1 SC-Management VlanID=1050 Untagged=True
add server-port-map D4-C2-B02:1 SC-VM VlanID=1051
add server-port-map D4-C2-B02:2 SC-iSCSI VlanID=1052 Untagged=True
assign profile D4-C2-B02 enc0:1B

Windows by default uses the pagefile as additional memory when your RAM runs out.  Because it uses the hard drive, it’s a lot slower than your RAM.

Disabling the pagefile will speed up everything on your server.

Microsoft recommends to not do this, but they fail to mention one important thing; the pagefile isn’t actually disabled – paging instead goes to your RAM.  You must have enough RAM to pull this off or else your server will slow to a crawl.  At least 2GB should be good (preferably 4GB). There is one caveat to this tweak though.

Everytime your server starts up it might take a little longer since everything gets paged to your memory.  But once its running, you’ll really notice a difference.  It’s especially noticeable when running applications that read/write to the hard drive a lot (like game servers).  I’ve been using this tweak for the last 5 years on a Windows 2003 Standard Server with no ill effects.

To disable page file do the following:

1. Right click my computer on desktop.
2. Select Properties.
3. Click “Advanced” tab.
4. In the Performance section, click the “Settings” button.
5. In the “Performance Options” window, click the “Advanced” tab.
6. In the “Virtual Memory” section, click the “Change” button.
7. Click the C drive from the drive list and click “No Paging File”.
8. Click the “Set” button next and click Yes on the next window.
9. Click Ok button and restart the computer.

Windows Pagefile - will it increase my Window Server Performance

disable pagefile - will it increase performance on my Windows Server

The following steps will help you to increase your game performance:

1) Check your process list before starting your game.  Often anti-virus software and other applications that read or write to the hard drive will slow down your game performance.  Especially applications that defrag it, they really slow it down.

2) Defragging. Not only does it increase the life span of your hard drive, but it also makes games run faster.  Why?  It because it puts all the files that the game needs to run in one continuous space instead of spread out all over the hard drive.  You should defrag at least once a month, but better yet is daily (preferably during off hours like at night).

The defragger built into windows cannot be scheduled so you’ll either have to remember to run it periodically or use My Defrag.  It’s a freeware defrag tool that can be scheduled.

3) If you did the two things above and your games are still performing poorly.  There’s a few things you can do in-game to improve your game performance:

• Screen Resolution: extremely high screen resolutions will really hurt game performance. : try reducing it to 1024×768.
• Shadows: sure they add realism but they really put alot of strain on your video card.  Try putting them at low or even off.
• Aniscopic Filtering: this makes objects ahead of you more clear.  It’s a real game performance hog though so you should leave it disabled.
• Anti Aliasing: this makes the edges of objects slightly blurry.  It also puts a lot of strain on your video card and unless you’re really paying attention, you wont notice it most of the time so just leave it disabled for that extra performance boost.
• Texture Quality: the last thing you can try is reducing texture quality.  Higher textures take up more video card memory.

If you tried all these things and your game is still performing badly than your only course of action is to upgrade your computer to increase your game performance.

First look at your video card.  How much memory does it have?  If its less than 256MB, you really should upgrade it as most games today require at least 512MB to run smoothly with higher settings.  Also look at how much ram you have.  You should have at least 2GB on 32-bit windows.  Don’t get higher than that on a 32-bit operating system because 32-bit windows is unable to enable more than 2.2 GB.  The best thing you can do is get 4GB but only on 64-bit windows (all the memory will be usable).

If you want to run your game server as a Windows Service you can do so with FireDaemon Pro and Manage via the web/iPhone/iPad with Fusion (free Beta).

If you have any other tips on game performance, feel free to share them in the comments below.

1. Create your service as normal
2. Set Settings -> Upon Program Exit -> Disabled (or Report the Termination)
3. Set Scheduling -> Schedule a Restart -> At Specific Time

This will cause FireDaemon to run your app initially to completion. When the app quits FireDaemon doesn’t restart it. At the “Specific Time” the app is run again and the cycle continues.