Best Practices for Application Delivery in Virtualized Networks – Part I
Earlier this year the Webtorials Analyst Division, co-founded by Dr. Jim Metzler, surveyed their subscriber base of IT professionals. Not surprisingly, 75% admitted that when a core business application degrades in performance, the end user notices before IT does. Therefore, 85% also believe that it is important, very important and even critical to senior managers that they take a more proactive approach to managing acceptable application delivery (See Figure 1).
Click here for the 2011 Application Service Delivery Handbook – Cisco
Contributing to the challenges of ensuring good application performance are the very innovations that are meant to simplify business and IT operations. These include data center consolidation, virtualization and the wide variety of applications that IT must support– all of which creates operational issues for IT. Not to worry – there are best practices that IT organizations can implement as application delivery challenges continue to evolve. In Part I of this blog post on application and services delivery, I’ll share what I consider to be key learnings from Dr. Metzler’s comprehensive 129 page guide. We’ll start with some core challenges:
Key Application Delivery Challenges
Proliferation of different types of applications: Today, companies utilize a wider variety of applications than ever. Some applications are business-critical. Others enable other business functions. And still more applications support communication and collaboration. Not only do they vary in criticality, but they also vary in their demands on the network. For instance, video streaming, which causes a lot of strain on the network may be key on some occasions (think company-wide all hands meetings a la Apple’s tribute to Steve Jobs), but recreational during other times. IT managers must audit company-wide application use, pinpoint a select group of business critical applications and formulate and execute a plan for optimization.
Data center consolidation: Consolidating data centers is meant to optimize resource use and simplify management. However, as data centers are consolidated, the distance between remote users increase. According to Webtorials, 80% of a company’s employees access applications over the WAN today and the number of mobile workers accessing networks and applications via personal mobile devices are growing. Even small delays in the network can contribute to significant application latencies. Today, only about a quarter of IT managers take into account how well an application will perform over the WAN into their planning process.
Virtualization: In 2011, 32% of survey responders had adopted virtualization across 50% or more of their environment compared to just 20% in 2010. These organizations typically run on data centers that comprise of physical servers, switches, WAN optimization controllers, Application Delivery Controllers (ADCs), firewalls and oftentimes their virtual counterparts. The environment is no longer static, but dynamic as virtual machines are instantiated and constantly mobile. Without the proper management tools that work in both physical and virtual environments, IT organizations may lack transparency into VM-to-VM traffic which causes security, performance monitoring and troubleshooting issues.
Application Delivery Best Practices
When it comes to best practices to ensure acceptable application delivery, Dr. Metzler cites wisdom from Lewis Carroll’s Alice in Wonderland, but I like to call upon Leroy Jethro Gibbs, the granddaddy of all NCIS agents and Rule #38 of his 50+ rules – “Your case, your lead.” The key here is that it is that when it comes to application delivery, it is on IT to determine the right course of action and establish the key players.
- ID and categorize applications on the network
- Establish SLAs to ensure acceptable performance of business critical applications
- Determine management and planning disciplines to implement
a. Application Performance Management – a holistic approach based on integrating application management with end-to-end IT infrastructure. This is like application performance utopia, and difficult to achieve.
b. Application Performance Engineering – a lifecycle approach to planning and managing application performance. The key is to:
i. Set performance objectives
ii. Map out network topology
iii. Do performance modeling
iv. Do performance testing and analysis
v. Implement network & application optimization tools. We will explore this in more detail in blog post Part II.
c. Explore solutions to meet challenges of virtualization
i. Implement an intergrated DNS, DHCP and IPAM solution to simplify data gathering
ii. Use a virtual performance management appliance or a virtual server management system
iii. Use Distributed Virtual Switching (DVS) which decouples the control and data plane and allows integration of a virtual switch’s control plane with the control planes of other switches
iv. Look into Edge Virtual Bridges (EVBs) which enables switching at the physical access level as opposed to within the hypervisor.
In Part II of this blog, out later this month, we will focus on network and application optimization and the fact that the majority of organizations either implement very little functionality or only on a case-by-case basis. We’ll explore opportunities here for businesses to improve network performance for application delivery.