Wide area network (WAN) bandwidth and network performance can suffer in a cloud computing environment. To learn more about the effects cloud computing has on WAN bandwidth, SearchEnterpriseWAN.com spoke with longtime IT network expert Zeus Kerravala, who manages Yankee Group's infrastructure research and consulting. In this 10-minute podcast interview, we discuss what cloud computing is, how cloud computing affects WAN workloads and therefore WAN bandwidth, and what a WAN manager can do to improve cloud performance and network speeds. You can use the table of contents to skip to questions you want answered or scroll below to read an adapted transcript of the podcast audio.
- TABLE OF CONTENTS
- Cloud computing defined [:48 seconds into podcast]
- How cloud computing affects WANs [2:25 minutes into podcast]
- How cloud computing affects WAN bandwidth [4:35 minutes into podcast]
- How to improve WAN bandwidth/cloud performance [5: 42 minutes into podcast]
Download for later:
- Internet Explorer: Right Click > Save Target As
- Firefox: Right Click > Save Link As
Zeus Kerravala: The buzz around cloud computing has certainly permeated through almost every segment of IT and really every part of technology. Exactly what cloud computing is -- and isn't -- is up for debate. It tends to vary from vendor to vendor and even from country to country because of regulatory issues. From Yankee Group's definition, cloud computing has seven essential features. It needs to be:
- SLA supported
- Usage priced
From our perspective, that would be a service that a company buys, can pay for by the month, and can access from anywhere. In many ways, that excludes things like internal private cloud. Although that is a form of cloud, I think the long-term definition of cloud computing is to have it offered as a service versus a company that builds it itself. I know there are arguments to be made on both sides, but from our definition, those seven key attributes -- scalable, virtualized, on-demand, Internet-based, multi-tenant, SLAs and usage priced -- are the key criteria for defining cloud computing.
Kerravala: Cloud computing has a very profound impact on the wide area network. We're moving into an era where this isn't your father's WAN anymore. The WAN at one time was used really just to distribute Internet traffic, a lot of things that most workers viewed as nice-to-have but not need-to-have. When we move to cloud computing, the WAN actually becomes in many ways the backplane of these virtual, cloud-based data centers. The data center, for one, is becoming much more distributed. IT assets in the data center were known as either hardware or software -- so you had server storage devices, apps and OSes. These are becoming virtualized, and the concept of managing these moves away from the hardware and software aspects themselves and becomes virtual workloads, such as storage, app development, or data analysis. The workload itself is what becomes virtualized and housed in the cloud. Once those workloads become virtualized, you can put them on the move. You are able to take that workload and move it from my data center to a cloud provider to another cloud provider. You might want to do that based on business policy or time of day. So, for instance, you might want to have your billing application, which would be used in the last few days of a month to send bills out to people, move to a very high-performance cloud-based server for two days in a month -- and then move away from that because then it goes unused. You can see that once we start moving these virtual workloads all around our network, it's going to put a lot of extra demands on the WAN that weren't there before. Companies need to understand where these demands are coming from, how to manage them and how to adapt them for cloud computing.
Kerravala: Cloud computing affects WAN bandwidth from the development of these workloads and the movement of these virtual workloads -- or VMDKs, if you want to use VMware-speak. They can be very large. In fact, in some cases, 10, 12 or ideally 20 gig VMDKs are being moved around from one cloud to another or from a private cloud on my data center to a cloud provider. It's very large chunks of data that need to be moved in real time. WANs are used to much smaller amounts of bandwidth that are very versed in things like Internet traffic and email; WANs never had to move very large workloads that contained mission-critical applications. So you're going to need more bandwidth in your WAN. You're also going to need lower-latency networks. We're trying to emulate almost a local area network (LAN) with our WAN status for cloud computing.
Kerravala: There's a number of steps a WAN manager can take to lessen the impact cloud computing has on WAN bandwidth. Some steps are more short term; some are more long term. Thinking about how to get that WAN to perform more like a LAN puts the network manager in the right frame of mind. Over the years, service providers have been rolling out carrier Ethernet and IP VPN services that have been pushing the limits of the WAN to achieve higher performance. Offerings such as VPL services or GigaWAN are designed to make the wide area network behave like a local area network. As we start moving towards cloud and virtualization, we're going to be able to emulate that even more because we're going to be carrying Fibre Channel and storage fabric, etc. A very short-term enhancement that WAN managers can do to fix WAN bandwidth is when they look at who their cloud computing provider is, take a look at what network they run on -- if it's offered from a telco, obviously, that would be the network itself – and then choose their cloud computing provider and their network provider together. If you can keep the traffic on-net, you minimize the number of hops that you have and you minimize the latency and the movement of these virtual workloads. If you can't keep everything on one network, minimize the number of hops by choosing operators that have a peering relationship with one another. Thinking more strategically about your WAN and who you choose as a network operator in conjunction with your cloud provider can have some very good short-term benefits, because you remove a lot of the unknowns. Many people think that the Internet is just a big cloud out there, but there actually is no such thing as the Internet. It's a collection of networks together, and if you understand how traffic moves across the Internet, you can actually make a better decision. That's what a WAN manager can do in the short term to improve how cloud computing affects WAN bandwidth.
The second thing you can do is a little more long term: Look into using WAN bandwidth optimization tools that are used to support data centers today, like application delivery controllers, or in the case of wide area traffic, something like a WAN optimizer. Look at products that were built specifically to improve WAN performance. By using a WAN-optimized cloud service, a company can cut down the amount of traffic that moves across the WAN by as much as 90% in some cases. By using a WAN optimization controller to compress and/or accelerate cloud traffic, the organization can get much better performance, even though these products were designed for other WAN traffic. There's no reason why these products can't be used in the cloud, but one caveat is that you may have to talk to your cloud provider to understand which WAN optimization solution they can support and choose your WAN optimization vendor that way. In many cases, you may even be able to deploy it yourself if you buy a virtual appliance. Those are just a couple of things that a WAN manager can do to improve the performance of cloud computing over the WAN: understanding how traffic flows around the Internet; choosing your service provider in conjunction with your cloud provider, and if you enter a private cloud environment, using some sort of WAN optimization solution to improve cloud performance over the WAN as you would with poor-performing applications.
For more in-depth information, read Kerravala's tip on using WAN performance optimization for cloud computing environments.
This was first published in July 2010