Citrix Systems 5.6 manual NIC bonds, Using VLANs with dedicated storage NICs

Using VLANs with dedicated storage NICs

Dedicated storage NICs can be configured to use native VLAN / access mode ports as described above for management interfaces, or with trunk ports and XenServer VLANs as described above for virtual machines. To configure dedicated storage NICs, see the section called “Configuring a dedicated storage NIC”.

Combining management interfaces and guest VLANs on a single host NIC

A single switch port can be configured with both trunk and native VLANs, allowing one host NIC to be used for a management interface (on the native VLAN) and for connecting guest VIFs to specific VLAN IDs.

NIC bonds

NIC bonds can improve XenServer host resiliency by using two physical NICs as if they were one. If one NIC within the bond fails the host's network traffic will automatically be routed over the second NIC. NIC bonds work in an active/active mode, with traffic balanced between the bonded NICs.

XenServer NIC bonds completely subsume the underlying physical devices (PIFs). In order to activate a bond the underlying PIFs must not be in use, either as the management interface for the host or by running VMs with VIFs attached to the networks associated with the PIFs.

XenServer NIC bonds are represented by additional PIFs. The bond PIF can then be connected to a XenServer network to allow VM traffic and host management functions to occur over the bonded NIC. The exact steps to use to create a NIC bond depend on the number of NICs in your host, and whether the management interface of the host is assigned to a PIF to be used in the bond.

XenServer supports Source Level Balancing (SLB) NIC bonding. SLB bonding:

•is an active/active mode, but only supports load-balancing of VM traffic across the physical NICs

•provides fail-over support for all other traffic types

•does not require switch support for Etherchannel or 802.3ad (LACP)

•load balances traffic between multiple interfaces at VM granularity by sending traffic through different interfaces based on the source MAC address of the packet

•is derived from the open source ALB mode and reuses the ALB capability to dynamically re-balance load across interfaces

Any given VIF will only use one of the links in the bond at a time. At startup no guarantees are made about the affinity of a given VIF to a link in the bond. However, for VIFs with high throughput, periodic rebalancing ensures that the load on the links is approximately equal.

API Management traffic can be assigned to a XenServer bond interface and will be automatically load- balanced across the physical NICs.

XenServer bonded PIFs do not require IP configuration for the bond when used for guest traffic. This is because the bond operates at Layer 2 of the OSI, the data link layer, and no IP addressing is used at this layer. When used for non-guest traffic (to connect to it with XenCenter for management, or to connect to shared network storage), one IP configuration is required per bond. (Incidentally, this is true of unbonded PIFs as well, and is unchanged from XenServer 4.1.0.)

Gratuitous ARP packets are sent when assignment of traffic changes from one interface to another as a result of fail-over.

Re-balancing is provided by the existing ALB re-balance capabilities: the number of bytes going over each slave (interface) is tracked over a given period. When a packet is to be sent that contains a new source MAC address it is assigned to the slave interface with the lowest utilization. Traffic is re-balanced every 10 seconds.

71