Designing a cost-effective network is nothing short of an art form requiring foresight and technical savvy. In this interview with SearchNetworkingChannel.com, network design expert Tom Lancaster explains how networking consultants and value-added resellers (VARs) can design cost-effective networks that ensure performance, scalability and redundancy.
By submitting your personal information, you agree that TechTarget and its partners may contact you regarding relevant content, products and special offers.
What does a VAR need to know from a client in order to improve a network's cost-efficiency?
Lancaster: You need to explore [the client's] tolerance for outages. How much does it cost them in revenue or customers? What really are their expectations for network service levels? If their network is down for 30 minutes, is it the end of the world or not a big deal? If the network is down for an hour, do they send their employees home? A lot of companies have a hard time attempting to put a dollar sign next to their outages. It's hard to deal with them unless they have more formal metrics.
Find out what their plans are for the advanced features. Do they think they'll need Power over Ethernet over the next five years? Do they think they'll want to go to Gigabit Ethernet? Do they need security features in their network? Frankly, a lot of companies just don't need any of those things, and others do. It makes a big difference in terms of cost-effective networks for a number of reasons. First, if you look at the pricing for almost any vendor, you'll see a big jump in price going to Gigabit or Power over Ethernet. You're going to spend a lot of capital there, so make sure you need those features. Some of those features are also going to be more difficult to support, especially Power over Ethernet. It can reduce the network's reliability because it might be more prone to failure than one without. As a VAR, you would want to walk your customer through a cost-benefit analysis and explain to them which one of these network types most closely describes them.
A lot of disaster recovery planning is about transaction volumes and times. For instance, if your customer is a medium-sized architectural firm, it may drag extraordinarily large files across its networks on a daily basis. But if the network's down for 30 minutes, they can wait and drag the file later. If, on the other hand, they're doing a lot of e-business overseas and outsourcing to India and South America or something like that, or if they have customers waiting, then their revenue will be impacted. I might recommend that one company spend their dollars on Gigabit Ethernet and not spend their money to make their network redundant.
What are some best practices for designing a cost-effective network that delivers greater performance, scalability and redundancy?
Lancaster: First, it's important to approach the problem holistically and to include the capital and labor costs. When trying to determine a good cost-effective network, it's important to understand that performance, scalability and redundancy are almost always a tradeoff for cost. More performance is going to cost you more money. More scalability is going to cost you more money. More redundancy is always going to cost you more money. It's a question of how much more it will cost and how to optimize your network so you're getting appropriate performance and scalability rates. As for the design, you should focus more on the design and less on the manufacturer of the hardware. A good network design has a greater impact on performance, scalability and redundancy than who made the switch or the router. You can certainly configure a Cisco router where it doesn't provide redundancy or availability.
When you get down to the design of these things, it's an art, like writing a book. For example, don't just think about redundant switches that take twice as much capital, but also think about how much labor it is going to take to support these two switches. These other support costs include maintenance, monitoring and management. Most of the companies that I deal with understand that these things are going to be expensive. It has been a question of, "OK, so I know I need two switches for redundancy. What's the most effective way of configuring those together?" It depends on where the switches are and what they're doing, if they're at the edge or the distribution or the core, what's plugged into these switches and so on.
How can redundancy make the network more cost-effective?
Lancaster: I'm not entirely sold on the idea that redundancy makes every customer's network more cost-effective. Give your customer this choice of two outage scenarios: Would you rather have a chance of a couple of brief outages with very high odds of getting back online rapidly, or do you want the higher availability, but if something's wrong, it could be a while to get back online? That's how I look at the redundancy component. In terms of being cost-efficient, it all boils down to how much revenue the business will lose when the network goes down.
If you have one simple switch and your configuration is simple and the switch fails, it's easy to fix the problem. The troubleshooting process is quick because it's obvious what's wrong. If you have a more complex network where you have an advanced configuration that fails over different switches, then, for starters, you're less likely to have an outage. If a router breaks, another redundant one picks it up.
But if something does go wrong, and there's an outage, it's a lot harder to troubleshoot because it's not obvious where the problem is. You get into this area where, on the one hand, something breaks and you want it back in service quickly, so you rip it out and stick it back in and you've suffered a brief outage. You hardly ever suffer an outage, but when you do, it lasts for several hours.
Can virtualization make a network more cost-effective, and when should a consultant recommend it?
Lancaster: Virtualization is interesting because virtualization of the network has been around a lot longer than virtualization on the server. We've been doing VLANs for decades now, so when you talk about virtualization on the network, specifically routing or switching, I definitely think it can make a network more cost-efficient. As an example, if you have a firewall and maybe a couple of DMZs on the firewall and your internal network, you could actually host those DMZs and different VLANs on the same Cisco switch. That becomes cost-effective because you don't have to support more switches and you save on hardware. You have the more complex configuration where you have different subnets riding on the switch, and it takes some security and controls to make sure that traffic from one network doesn't leap across the switch and bypass the firewall. There are all kinds of security features that protect VLANs from attacks now.
You can do similar things in the WAN with lots of newer technologies.
About the author
Tom Lancaster, CCIE# 8829 CNX# 1105, is a consultant with 15 years of experience in the networking industry. He is co-author of several books on networking, most recently, CCSP: Secure PIX and Secure VPN Study Guide, published by Sybex.