How to fine-tune the quality of network services on IBM AIX 7.1 and Linux in a multi-tenant cloud environment

Brief introduction

With the rapid development of small and medium-sized enterprises, scalability, low initial investment, simple deployment, reduce footprint, pay by use, security, etc. become hot issues, these can be solved through the cloud model. In all cloud components, the multi-tenant model helps to share software and hardware layers. In layman's terms, a tenant uses all the assets of other people and pays for the use. Similarly, multi-tenant in cloud terminology means that multiple clients leverage the shared resources provided by the cloud infrastructure owner.

The concept of Multi-Tenant forms the foundation of two mature cloud implementation models: Software as a service (SaaS), where multiple clients share software as a resource, infrastructure as a service (IaaS), where multiple clients share applications, platforms, and network resources. The main function of multi-tenant support cloud vendors is to maintain separation between clients and manage the quality of services provided to clients.

This article helps cloud administrators build different network QoS policies for each IP address using Linux instead of dedicated hardware (such as a network allocator) through selected price schemes, providing fine-grained QoS tuning to tenants.

Network patterns in a multi-tenant environment

Let's consider a scenario where a cloud vendor has multiple tenants, and the vendor wants to provide a different quality of service across the network based on the solution chosen by the customer. The following illustration shows a sample QoS implementation diagram where some of the machines are partitioned into different QoS pools, and cloud administrators need to divide the network bandwidth.

Figure 1. Typical network configurations in the Enterprise cloud

To drive the context, we assume that a cloud vendor has initiated multiple billing levels (slab) and presented the following scenario details:

Customers under the Platinum scheme will get 60% of the total speed.

Customers under the Gold scheme will get 25% of the total speed.

Customers under the Silver scheme will get 10% of the total speed.

Customers under the bronze scheme will get 5% of the total speed.

Now consider a scenario where customer A and client B leverage the same data server infrastructure, where customer a belongs to the Platinum scenario and customer B belongs to the bronze scenario. If customer B's workload requires a large amount of network resources on a shared linux-based cloud server, it is clear that if network isolation or restrictions are not applied, client B will occupy the network resources reserved for customer A, which indirectly affects the workload of client A. and violates service level agreements (SLAs) made for advanced customers. This problem is highlighted in a multi-tenant environment, in order to meet the quality of service and SLAs of Multi-Tenant customers, cloud providers should use manageable fine-grained network shaping (network shaping) for network interface cards that meet client requests.

To demonstrate how the above problem is resolved, we will demonstrate the configuration of network QoS with a simple setup that involves a Linux server (as a central general-purpose cloud server hosted cloud application) and an AIX 7.1 client (which forms the tenant system to access cloud applications), as shown in the following illustration. In hypothetical cases, network-related QoS and the SLAs associated with AIX clients require that the client's data download speed be limited to 1MBps. In this demo, we will adjust the network bandwidth on the Linux server and limit the speed of the AIX client to obtaining or downloading data to 1MBps. This helps to comply with the network QoS associated with the AIX client.

Figure 2. Example settings between a Linux server and an AIX 7.1 client