subscribe: Daily Newsletter

 

Don’t forget the digital plumbing

0 comments

Kathy Gibson at Gartner Symposium, Cape Town — To enable the digital business, digital plumbing has to be in place, and new networks are going to be needed to run the new systems.
Neil Rickard, networking analyst at Gartner, explains that the new scenario is seeing a new type of network, based on new technologies and delivered by new vendors.
“The existing technologies and relationships are not going to go away,” he says. “But, as network professionals, we need to step across the chasm and start thinking about the other networks that we have to use.
“If you are bimodal, by plan or not, you are using another network. So you need to change your mindset to embrace these new services and capabilities.”
Gartner estimates that, by 2020, 30% of an enterprise’s strategic network suppliers will change. And the new suppliers may be organisations that most of us don’t think of as being network suppliers at all, Rickard says.
The biggest trend set to impact the network is digital business. “The blurring of the physical and virtual worlds is significant because it places enormous burdens on the network as the glue that holds it all together, and makes it possible at all,” Rickard says.
As an example, Google had to redesign its load balancer to make Pokemon Go work. “There are implications in implementing things like augmented reality, and you will have to rethink your network to make it work.”
The Internet of Things (IoT) is another trend that is going to impact the network — and it could be just about any part of the network depending on the type of IoT being used: at the edge, the connectivity, the gateway, the cloud or the enterprise.
With 30-billion IoT endpoint expected to be connected by 2020, the implications for networks could be massive, and network professionals are urged to start getting their networks ready.
“You may not even be doing IoT yourselves, but you will likely have to connected to systems that do have IoT,” Rickard says.
In the past, emphasis has been placed on the data centre and the data centre network, Rickard points out. “What we ae seeing now is that the enterprise data centre is shrinking. More and more workloads are being put into co-location or cloud data centres.
“So there are fewer workloads inhouse, but they tend to be the most critical workloads. So we are able to focus our attention on a smaller footprint inside the data centre. So we should be reducing our focus on the data centre network and start thinking about the networks that are not our own, but that we have to work with.
“So the data centre has to stop being the centre of the universe, and we need to see it as one source of information in the network.”
New networking technologies are quickly coming to market, Rickard points out.
“For many organisations, cloud is quickly getting adopted. And the network is often the thing that is holding the cloud back from being implemented. The network is often perceived as the problem in the cloud roadmap.”
Network professional should re-architect their data centre networks to make them more cloud-ready; and they should look at re-architecting the wide area network (WAN) so it can seamlessly connect to the cloud, with good performance.
“You need to think about re-architecting the network to support cloud. So when the organisation decides to implement Office 365 you can accommodate it immediately. The goal is to be able to rapidly assimilate new cloud services.”
Gartner’s enterprise hype cycle indicates that software-defined network (SDN) is currently in the trough of disillusionment, having failed to live up to some of the hype. But Ethernet switching fabrics are now on the slope of productivity.
A number of new WAN technologies are at the peak of inflated expectations now, indicating the amount of activity and interest in the WAN.
“What happened to SDN?” Rickard asks. “It was going to solve all our LAN and then WAN problems.”
The problem, he says, is that SDN is an architecture for making the network more agile, giving administrators the ability to automatically provision networks. There are, however, several alternatives to solving those problems.
“There are about 3 000 companies that have implemented SDN, which is a relatively small number,” Rickard points out. “We think SDN is a perfectly viable solution for an enterprise data centre network, but it’s not the only solution.”
There is a lot of activity taking place in the WAN environment. “We are seeing genuine generational change taking place in the WAN,” Rickard says. For instance, routing is no longer the only way of building WANs.
SD-WAN is slightly different to SDN, although it shares similar structures, bringing centralised policies and automation to the WAN.
“SD-WAN is a new class of edge device that allows us to build WANs where the network is a system, rather than every device being a system. This truly is a better way to build WANs and we are seeing enormous interest in that.”
Technologies that will impact the data centre network include containers and hyperconverged infrastructure (HCI).
Container networking is still in its infancy, so Garner recommends investing tactically and opting where possible for open systems. “Avoid over-committing, that landscape is changing very fast and you don’t want ot be locked into a vendor that could become proprietary.”
HCI is a big deal for networking professionals, and they must ensure they drive HCI decisions, Rickard says. “If the networking professional is not involved in these conversations they could end up with huge problems in the data centre network.”
Because HCI has networking infrastructure built in, this could conflict with the existing network and make running and managing the data centre network cumbersome and problematic.
“You need to make sure the standards and protocols of HCI are compatible with the network. You can’t let people go out and buy it, and plug it into the network.”
Networking has traditionally been done on the basis of incremental change, taking a long time to build up sedimentary layers of things. “This incremental change becomes a problem because you could have painted yourself into a corner,” Rickard says. “This approach could stifle agility and it could be a real problem to accommodate the new demands that will be placed on the network.”
Radical change could be the better solution, he says. Trying to change existing infrastructure to act in new ways can be difficult and counter-productive.
“Sometimes you have to make the big leaps, bite the architectural bullet and go for the big change.”
Network professionals have to find new ways to deal with downtime as well. The traditional way, of buying two of everything, may not work in the future. In the future, networks need to be anti-fragile, with equipment integrated in a way that avoids fragility.
“We also need to change our practices, to ensure blameless post-mortems that hunt for the problem and solutions to it rather than find a culprit. The issue needs to be about finding and fixing the processes.”
Related to this is the need to use more automation in the network. “Our networks are still far too manual,” Rickard says. SDN helps with this.
Networks also need to be instrumented and tooled so administrators can see what is going on. “We don’t control the performance of the Internet — in lieu of control we need to ensure visibility,” Rickard says. “There is a lot of interest now in application visibility on the network.”
In the journey to network agility, it’s important to get away from manual processes, Rickard says. Currently, 82% of changes on the network today are manual. The solutions include automation, orchestration, new policies and intent.
Intent-based networking, still a relatively new area, will being business logic and policy into the network. So network performance will be automatically and consistently balanced to ensure the best outcomes for the business.
Rickard points out that networking people still enjoy using the command line interface (CLI) to micro-manage the network. ‘It’s where a lot of network professionals see their core value.”
Bu 2020, though, only 30% of network operations team will use the CLI as their primary interface — down from 85% today. This change may be difficult, but it has to happen as more automation and anti-fragility become central features of the network.