Anda di halaman 1dari 6

Typical VIOS Network Configuration in Production Environment

v1.0
Bejoy C Alias
IBM India Software Lab

Typical VIOS Network Configuration in Production Environment

Revision History
Date of this revision: 29-Sep-2010 Date of next revision : TBD

Revision Number V1.0

Revision Date 29-Sep-2010

Summary of Changes 1. Initial creation

Changes marked N

Preface
This document explains a typical network configuration scenario using VIOS in a production environment to provide better network performance and redundancy. This document is created based on the recent experience in setting up a consolidated production environment using dual VIOS with redundant network configuration.

IBM Corporation, 2010. All rights reserved.

Page 2 of 6

Typical VIOS Network Configuration in Production Environment

Table of Contents
1. 2. 3. 4. Introduction .......................................................................................................................................... 4 NIB for VIOS Network Access Redundancy .......................................................................................... 4 Ether Channel for Network Performance and Redundancy .................................................................5 SEA Failover for VIOS Redundancy ....................................................................................................... 6

IBM Corporation, 2010. All rights reserved.

Page 3 of 6

Typical VIOS Network Configuration in Production Environment

1. Introduction
In a production environment, one of the main considerations everyone will think of is to avoid any single point of failure, be it Network, SAN, Disk etc. This document explains a typical setup one can use in the production environment to provide maximum network redundancy and performance when using IBM Power Systems with dual VIOS configuration. This example environment has two VLANs for the production servers, one for the public network and another for the private network. The configuration explained in this document avoids multiple points of failure in the network path and provides good network bandwidth for the LPARs hosted in the IBM Power6 System. To avail the redundancy and performance, this configuration uses Ether Channel, Network Interface Backup (NIB) and SEA failover. This setup also uses two separate network switches for connecting the two VIOS and hence the production LPARs to avoid single point of failure in the network switch side, also the switches are being shared for both private and public VLAN. The below given diagram (Figure 1.1) shows the logical representation of the configuration explained in this document.

Figure 1.1

2. NIB for VIOS Network Access Redundancy


VIOS LPAR itself must be connected to the network to do any dynamic LPAR operations and for normal operations like backup/restore or simple file transfer. Typically the VIOS network need not be configured for higher performance since the operations happening on VIOS is not so network intensive (e.g.; dynamic LPAR operations). Since this example configuration is hosted on IBM Power6 hardware and since VIOS need not require a high bandwidth network path, the network configuration for the VIOS is done using the IVE (Integrated Virtual Ethernet) feature. There are 2 LHEA ports assigned for the VIOS and each port is coming from two separate HEA
IBM Corporation, 2010. All rights reserved. Page 4 of 6

Typical VIOS Network Configuration in Production Environment

(Host Ethernet Adapter) card to avail redundancy, and this ports are configured with NIB (network interface backup) in VIOS so that if the network cable connected to one HEA port or either of the two HEA card fails the VIOS network will be still up and running. Also, these HEA ports are connected to separate network switches as shown in figure 1.1 to provide redundancy in network switch level. The first LHEA interface in both the VIOS is coming from the same HEA adapter and the second LHEA interface is coming from another HEA adapter itself. NIB configuration is done as shown in the below figure 2.1. Run smitty and navigate to Devices > Communication > EtherChannel / IEEE 802.3ad Link Aggregation > Add An EtherChannel / Link Aggregation, input the values as shown, based on your environment change the device names and other parameters. In this example, ent0 is the device for LHEA port 1 and ent1 is the LHEA port 2.

Figure 2.1 This will create a new interface with the next free device name, say ent16. You need to configure the VIOS IP address on the newly created interface ent16. Once the IP is configured, you can test the configuration by removing one of the network cables connected to the HEA port being used for the VIOS. Similar configuration is done for the redundant VIOS as well.

3. Ether Channel for Network Performance and Redundancy


There may be many production LPARs hosted on the single IBM Power system and sharing a single network interface with all the LPARs will be a performance disaster. To provide maximum network performance we need to configure port aggregation with multiple network ports and share the aggregated port to production LPARs. In the example scenario we have 16 production LPARs hosted in a single p6 570 box and from the historic monitoring reports for those production servers, the combined bandwidth of all those servers never gone above 3Gbps at a given time. So we have decided to do link aggregation with 4 network ports of 1Gbps which would give a combined bandwidth of 4Gbps and redundancy up to 3 NIC (Network Interface Card) failures. We have assigned 5 no. of dual port Network cards to each VIOS, 1 port from 4 NICs used for creating the 4Gbps pipe, 1 port dedicated for the private VLAN, remaining 5 ports are kept free for future expansion based on the new production server requirements. We have created an LACP based Ether Channel in the network switch and configured the same in the VIOS. You can run the below mentioned commands in your network switch (Cisco) to configure it for LACP port-aggregation. These commands may not work on other network switches.
interface Port-channel1

IBM Corporation, 2010. All rights reserved.

Page 5 of 6

Typical VIOS Network Configuration in Production Environment


switchport switchport access vlan 210 switchport mode access interface GigabitEthernet4/42 switchport access vlan 210 switchport mode access channel-group 1 mode active

Once the port aggregation is done from the switch side, ether channel has been created in the VIOS as shown below in figure 3.1. Please note that for creating an LACP Ether Channel the Mode has to be selected as 8023ad. In this example configuration, ent2, ent4, ent6 and ent8 were the devices for the 1st network ports of each NIC.

Figure 3.1 When the Ether Channel is created, it will create a new network device ent17 and the SEA has to be created using this newly created Ether Channel device. Similar way, another Ether Channel has been created in the redundant VIOS.

4. SEA Failover for VIOS Redundancy


Since we are using dual VIOS for the production environment, the SEA in each VIOS has to be configured in such a way that it can failover to the other VIOS in case of a VIOS failure or if the VIOS is brought down for any planned maintenances like applying fix pack. In this example environment, there are two SEAs created, one for the public VLAN and another for the private VLAN. When you create the SEA, make sure you are defining the control channel interface as well, to allow faster switch over to the redundant VIOS SEA in case of a VIOS failure. The command used to create the SEA using the Ether Channel is given below.
mkvdev -sea ent17 -vadapter ent12 -default ent12 -defaultid 1 attr ha_mode=auto ctl_chan=ent13

The SEA for the private VLAN is created by the command:


mkvdev -sea ent10 -vadapter ent14 -default ent14 -defaultid 2 attr ha_mode=auto ctl_chan=ent15

Same commands were run in the redundant VIOS as well.

IBM Corporation, 2010. All rights reserved.

Page 6 of 6