How to add high availability to the network infrastructure of a multitenant cloud environment using the DLMP aggregation technology introduced in Oracle Solaris 11.1.
This article is Part 1 of a two-part series. In Part 1, we will cover how to implement network HA using datalink multipathing (DLMP) aggregation technology, which was introduced in Oracle Solaris 11.1.
In Part 2 of this series, we will explore how to secure the network and perform typical network management operations for an environment that uses DLMP aggregations.
Once we virtualize a network cloud infrastructure using Oracle Solaris 11 network virtualization technologies—such as virtual network interface cards (VNICs), virtual switches, load balancers, firewalls, and routers—the network itself becomes an increasingly critical component of the cloud infrastructure.
In order to add resiliency to the network infrastructure layer, we need to implement an HA solution at this layer, such as we would do for any other mission-critical component of the data center.
A DLMP aggregation allows us to deliver resiliency to the network infrastructure by providing transparent failover and increasing throughput.
The objects that are involved in the process are VNICs, irrespective of whether they are configured inside Oracle Solaris Zones or in logical domains under Oracle VM Server for SPARC.
Using this technology, you can add HA to your current network infrastructure without the cross-organizational complexity that might often be associated with this kind of solution.
The benefits of this technology are clear and they take into account the limitations of existing technologies:
Since the IEEE 802.3ad trunking standard does not cover the case for building a trunk across multiple network switches, the network switch becomes a single point of failure (SPOF). Some vendors have added this capability to their product, but these implementations are vendor-specific and, therefore, prevent combining switches from multiple vendors when building a multi-switch trunk. Because Oracle Solaris provides resilience, DLMP aggregation can be implemented across two different network switches, thus eliminating the network switch as a SPOF. As an additional benefit, because the aggregation is implemented at the operating system layer, there is no need to set anything up on the switch.
Building a network HA solution that is based on previously available IP network multipathing (IPMP) can be a complex task. With IPMP, HA is implemented at Layer 3 (the IP layer), which needs to be configured in the global zones and within each zone, and requires multiple VNICs to be assigned to each zone or virtual machine (VM). This involves more configuration steps, requires spare IP addresses out of the address pool, and generally can be an error-prone process. In contrast, the DLMP aggregation setup is much simpler since all the configuration takes place at Layer 2 in the global zone; therefore, every non-global zone can directly benefit from the underlying technology without the need for additional configuration. In addition, every new Oracle Solaris Zone that is provisioned automatically benefits from this capability. Moreover, we can create an aggregation over four 10 Gb/sec network interfaces; combining all the interfaces together, we can achieve up to 40 Gb/sec of network bandwidth.
DLMP can provide additional benefits when employed together with other network virtualization technologies that are implemented in the Oracle Solaris 11 operating system, such as link protection and the ability to configure a bandwidth limit on a VNIC or a traffic flow to meet service-level agreements (SLAs). Combining these technologies provides for a uniquely compelling network solution in terms of HA, security, and performance in a cloud environment.