By Stephen J. Bigelow, Features Writer
Not all network upgrades require major hardware additions or renovation to support high-availability applications. In many cases, application performance and availability can be improved by changing the network's existing architecture to alleviate traffic bottlenecks or supplement single points of failure between a user and application server. Server virtualization has also emerged as a means of enhancing availability by reducing server hardware in the network -- as long as network designers take the appropriate precautions.
The first installment of this Hot Spot Tutorial introduced the basic concepts of high-availability network applications. The second section addressed specific network hardware upgrades and the implications for solution providers. This final installment focuses on non-hardware network upgrades that solution providers can deliver to their clients.
Software and network design upgrades
While new switches or faster Ethernet infrastructures can do a lot to improve the availability of applications, substantial hardware upgrades aren't always necessary or cost-effective for businesses. The actual network design may cause bottlenecks that impair availability. A common problem is "segment sprawl," which occurs when more users are added to a network segment, thereby increasing traffic on that segment and slowing performance for all applications. A review of the client's network documentation can often reveal overtaxed network segments.
There are several solutions for these design issues. Solution providers can consider a redesign of the overtaxed segments, breaking a single busy segment into multiple smaller segments that connect to different switch ports. For example, if 20 users on a segment are all routinely using the same bandwidth-intensive application, you might ease the traffic congestion by creating two separate segments with 10 users each. This aids all network tasks, not just high-availability applications.
Network redesign may be used in conjunction with simple upgrades for faster Ethernet segments, such as replacing older congested 10/100 segments with 1 Gigabit Ethernet (GigE) or even 10 GigE segments. Availability can also be improved by updating or adding redundant network links at key locations (e.g., between the application server and switch, or between switches and storage systems). This kind of trunking allows traffic aggregation for faster data transfers, and can be implemented to support load balancing so that traffic is roughly equal on each element of the trunk for best performance. Redundant links can also support failover, allowing traffic to continue across remaining links when one link fails.
Software tools can also play an important role in ensuring network performance for high-availability applications, according to Dave Sobel, CEO of Evolve Technologies, a solution provider in Fairfax, Va. "You can look at how the network is implemented and its failover and what you're able to accomplish from a raw uptime perspective," Sobel said. "But additionally you can build in the monitoring and management pieces to achieve that [availability] on an ongoing basis."
Still, software won't prevent or fix every network performance problem. For example, switch firmware updates are a normal maintenance task, but the effect on overall performance and availability is incidental at best -- particularly in smaller or older environments.
Virtualization has emerged as a crucial technology for network resiliency and availability. For example, server virtualization allows one physical server to be divided into multiple "virtual machines" (VMs), and each VM draws from the server's common pool of processing power to run its own independent workload. Today's servers provide the fast multi-core processors, RAM and I/O needed to run multiple workloads on the same hardware platform. This reduces the number of physical servers in the network, reduces cost, eases management and maintenance complexity, and forestalls the need for additional servers in the future.
Since fewer servers are needed, IT shops find it more practical to invest resources in other resilient network technologies, such as clustering. Virtualization can also control the allocation of processing resources, allowing administrators to increase or decrease the amount of CPU and RAM available to each virtual machine. This can affect application performance in the corresponding VM.
Limitations of non-hardware network upgrades
Generally speaking, the limitations and disadvantages of non-hardware upgrades are identical to those of major, hardware-focused network upgrades. Network redesign or segmentation can add some technical complexity to the client's network -- often resulting in added and ongoing maintenance overhead.
The introduction of virtualization has had an even greater impact on network resiliency. While server virtualization can ease costs by reducing the sheer volume of server hardware needed, the cost benefit of a lower server count can actually introduce serious vulnerabilities. Remember that virtualization allows a single physical server to run multiple virtual machines. If the server hardware fails, all of the virtual machines on that physical server (and the applications those VMs run) can be disrupted. Where a single server may have presented a single point of failure in the past, virtualization means that a single physical server may now result in multiple points of failure.
Solution providers suggest leveraging the cost benefit of virtualization to invest in multiple redundant servers through server clustering or other technologies. "In our system designs, we're being very careful to build the resiliency necessary to justify that migration," said Scott Gorcester, president of Moose Logic, a network solution provider located in Bothell, Wash. "I don't think I have a single installation of virtualization today that isn't on at least two virtual servers. … Don't put all those eggs in one basket, because now we can afford two or more baskets."
Revenue opportunities with non-hardware network upgrades
Capable solution providers can often find a standard suite of revenue opportunities that go beyond simple one-time project costs, including recurring network evaluation, network testing and network management/configuration. But some solution providers may identify opportunities to train growing clients that wish to service and manage their own networks -- especially once virtualization technology is deployed.
Solution providers should also consider potential storage sales opportunities. "ESG research indicates 86% of those companies using server virtualization support that environment with networked storage through some combination of iSCSI, NAS and SAN," said Bob Laliberte, analyst at the Enterprise Strategy Group, an industry analyst firm in Milford, Mass. This may include adding storage to existing platforms, adding new storage platforms, replacing aging storage platforms, and possibly even storage provisioning and other management tasks.