Configure jumbo frame for RAC optimization and jumboframe for rac Optimization

Configure jumbo frame for RAC optimization and jumboframe for rac Optimization

Configure jumbo frame for RAC Optimization)

First, let's talk about the concept of MTU: in network communication, there is a concept of MTU (Max Transmission Unit), that is, the maximum frame size in network Transmission, the default value is 1500 bytes (the Ethernet Variable frame size is 46-1500Byte ).

[Root @ node1 ~] # Ifconfig bond0

Bond0 Link encap: Ethernet HWaddr C8: 1F: 66: FB: 6F: CD

Inet addr: 10.10.10.105 Bcast: 10.10.10.255 Mask: 255.255.255.0

Inet6 addr: fe80: ca1f: 66ff: fefb: 6fcd/64 Scope: Link

Up broadcast running master multicast mtu: 1500 Metric: 1

RX packets: 353 errors: 29 dropped: 0 overruns: 0 frame: 29

TX packets: 254 errors: 0 dropped: 0 overruns: 0 carrier: 0

Collisions: 0 txqueuelen: 0

RX bytes: 250669 (244.7 KiB) TX bytes: 160443 (156.6 KiB)

Therefore, in the default configuration, if the size of the data transmitted exceeds bytes, the system splits the frame into several frames for transmission:

[Root @ node1 ~] # Traceroute-F node2-priv 1500

Trace to node2-priv (10.10.10.106), 30 hops max, 1500 byte packets

1 node2-priv.localdomain (10.10.10.106) 0.234 MS 0.217 MS 0.204 MS

[Root @ node1 ~] # Traceroute-F node2-priv 1501

Trace to node2-priv (10.10.10.106), 30 hops max, 1501 byte packets

1 node1-priv.localdomain (10.10.105) 0.024 MS! F-1500 0.005 MS! F-1500 0.004 MS! F-1500

[Root @ node1 ~] #

In the RAC environment, we need to pay attention to one place.

RAC private networks are mainly used for network heartbeat communication between nodes, but in addition, nodes often need to transmit data blocks through private networks. In Oracle, the size of a database is 8192 bytes. Therefore, according to the default configuration, a data block has to be divided into several frames for transmission, which increases the load on the private network, therefore, we recommend that you set the MTU of the private Nic to 9000 in Oracle.

Let's take a look at my modification steps:

1) view the private NIC (executed on two nodes)

[Root @ node1 ~] # Oifcfg getif

Em1 192.168.10.0 global public

Bond0 10.10.10.0 global cluster_interconnect

2) set the private Nic mtu (executed on two nodes)

[Root @ node1 ~] # Ifconfig-s bond0 mtu 9000

The settings have been set successfully. We can test the settings through traceroute or ping:

1) traceroute

[Root @ node1 ~] # Traceroute-F node2-priv 9000

Trace to node2-priv (10.10.10.106), 30 hops max, 9000 byte packets

1 node2-priv.localdomain (10.10.10.106) 0.346 MS 0.364 MS 0.413 MS

[Root @ node1 ~] # Traceroute-F node2-priv 9001

Trace to node2-priv (10.10.10.106), 30 hops max, 9001 byte packets

1 node1-priv.localdomain (10.10.105) 0.043 MS! F-9000 0.010 MS! F-9000 0.010 MS! F-9000

[Root @ node1 ~] #

2) Ping

[Root @ node1 ~] # Ping-c 2-M do-s 8972 node2-priv

PING node2-priv.localdomain (10.10.10.106) 8972 (9000) bytes of data.

8980 bytes from node2-priv.localdomain (10.10.10.106): icmp_seq = 1 ttl = 64 time = 0.552 MS

8980 bytes from node2-priv.localdomain (10.10.10.106): icmp_seq = 2 ttl = 64 time = 0.551 MS

 

--- Node2-priv.localdomain ping statistics ---

2 packets transmitted, 2 bytes ed, 0% packet loss, time 1001 ms

Rtt min/avg/max/mdev = 0.551/0.551/0.552/0.023 MS

[Root @ node1 ~] # Ping-c 2-M do-s 8973 node2-priv

PING node2-priv.localdomain (10.10.10.106) 8973 (9001) bytes of data.

From node1-priv.localdomain (10.10.10.105) icmp_seq = 1 Frag needed and DF set (mtu = 9000)

From node1-priv.localdomain (10.10.10.105) icmp_seq = 1 Frag needed and DF set (mtu = 9000)

 

--- Node2-priv.localdomain ping statistics ---

0 packets transmitted, 0 encapsulated ed, + 2 errors

 

[Root @ node1 ~] #

How to use Jumbo Frame in vSphere 40)

Configure the vSwitch to use a giant frame by changing the MTU size of the vSwitch. follow these steps:
Step 0: Use the vicfg-vswitch-m <MTU> <vSwitch> command in VMware vSphere CLI to set the MTU size for the vSwitch. You can use this command to set MTU for the uplink of the vSwitch. Set the MTU size to be the largest among all virtual network adapters connected to the vSwitch. For example, vicfg-vswitch-m 9000 vswitch0 sets the MTU of vSwitch vswitch0 to 9000.
Step 1: Use the vicfg-vswitch-l command to display the vSwitch list on the host and list the vSwitch attributes. Use this command to check whether your settings are correct.
To enable a mega frame on the vmkernel port, follow these steps:
Step 0: directly log on to the ESX host Console
Step 1: Use esxcfg-vmknic-a-l <ip address>-n <netmask>-m <MTU> <port group name> to create a VMkernel connection that supports giant frames.
Step 2: run the esxcfg-vmknic-l command to display the VMkernel interface list and check whether the interface with mega frame enabled is correctly configured.
Step 3: configure all physical switches and any physical machines or virtual machines connected to the VMkernel interface to support giant frames.
Enable giant frames on a physical switch
Enabling a giant frame on a physical switch depends on the structure and type of the switch. To enable a mega frame on the customer's operating system, you can use an enhanced VMXNET or VMXNET3 virtual Nic and enable mega frame through the customer's operating system.

Jumbo frame

It is actually a giant frame, which is larger than the current Ethernet frame. You can leave a contact information for details ~