SDN NIL blog
28.9.2015

The idea of software defined network (SDN) stems from the desire to increase practicality in managing IT resources, since IT specialists wanted to separate the management, control and data aspects of the network. Traditional providers fell short of achieving this goal in time; but SDN was right around the corner—as a punishment of some sort. You see, networks are the backbone of modern business operations, without them you can’t have connectivity, transactions, cooperation …

Virtualization, which is one of key components of software-defined networks in data centers, has also found itself in a duopoly situation. Gartner’s segmentation model of virtualization solutions, places Microsoft and VMware are at the forefront of this field. Because of their hypervisor technologies, of course. But when it comes to networks in general, we cannot ignore Cisco, a company that wants to be (the main) part also in SDN. As experts in networking and connectivity solution, here at NIL we constantly monitor the progress of all three leading providers.

So I took the time to conduct a short comparison of these industry specialists and their SDN solutions. And I’ll do it in alphabetical order, comparing mostly technology-related aspects rather than their business implications. You have been warned.

The key advantage of SDN solutions it that they free networks from being static, but enable application-controlled network management. This means that the application (or rather its metadata) tells the network what kind of conditions the network should provide of it to properly function. At the same time, these conditions can dynamically  alter (because the level of activity in applications also alters).

Cisco ACI

Cisco rolled out the Cisco ACI solution more than a year and a half ago. Cisco likes to point out that they already have more than 900 business users worldwide, even though nobody (except for Cisco, of course) knows the exact number of implementations. But never mind, at this point it’s not even that important. Cisco’s approach to SDN differs from its competition’s—this networking giant is well aware that SDN can be applied to switches as well, so they carry a level of importance. By combining its switches with the ACI solution, Cisco is now offering a complete solution—both hardware and software included—which is a firm guarantee that this pair will work, and work truly well. But looking from the client’s point-of-view, this might seem like a drawback. Cisco clients have to purchase their (specific) more expensive switches, which greatly affect the scale of their investment. But still, we have to hand it to Cisco, they sort out many issues through their hardware which otherwise would call for a separate software solution. For example, they offer a firewall that function at the connectivity speed. Latency is minimal, and a faster connectivity is deeply appreciated. The network topology has also seen a shift—it now entails backbones and leafs (each leaf being attached to two backbones). This way only two hops are needed to reach each point in the network. The same distance separates each network component, which simplifies access. But even SDN solutions aren’t perfect; traffic that goes through a network is often sub-optimally dispersed, because it goes through the same links. Well, the software part handles this issue.

Of course the software part of the solution still manages the higher-level functionalities of SDN—for example, distributed firewalls, load-sharing, enabling applications and services, etc. What isn’t managed by the software and the controller, SDN manages through the use of APIs (Application Programming Interface), which contact other solutions. This means that Cisco ACI can manage also firewalls and load-balancers of other vendors and can be extended to hypervisors. With this granularity network administrators see into individual virtual machines and their network resources.

So what can I say about Cisco—you get what you pay for (which is a lot). If only you have such deep pockets.

Microsoft Windows Server 2016 and System Center 2016

An even most thankless task is describing Microsoft’s SDN solution. I have to say, this one is interesting, since it promises a lot. You can check its functionality in the Technical Preview of Windows Server 2016. The solution has a Network Controller that enables a centralized, programmable automation of management, configuration and control of both virtual and physical network infrastructure in data centers.

Judging from its predecessor, Windows Server 2012 R2, Microsoft wouldn’t even stand a chance. In this system, Microsoft has in a way forced the SDN solution through—you could see they were in a hurry to release it. This meant that network administrators have had to manually perform a lot of the tasks.  Somehow, the point of virtualization and resource automation was in this way promptly lost.

But if there’s one thing Microsoft can do is catch on quickly. From the first glimpse at the solution you can see that they copied their competition’s (VMware’s) approach to SDN. All they have to do is bring it to a satisfactory functioning level. They have until next spring, roughly speaking, when Windows Server 2016 and System Center 2016 become actually available. At this moment, Microsoft’s SDN solution looks promising, but because it’s available only in its technical preview version, it can be applied only in testing environments, and not production (it’s not even close to the final version).

Microsoft’s SDN solution has one obvious advantage. It’s going to be a part of a new operating system, making it less expensive compared to competitive solutions. They also have a large user base playing for them­—they comprising mostly those who already have a virtualized data center running Microsoft technologies. These business users will wait for Microsoft’s next spring’s release, and they will not only receive Windows Server 2016 but also software-defined storage of data and IT resources (Hyper-V), which is practically a complete software-defined data center. This time there’s no fear that Microsoft will fail to meet our expectations, since the foundation of many incorporated technologies is Microsoft Azure, which assures you that this technology has been are tested out and can perform in even the largest environments.

The drawback in this case stands out. The solution is not market-ready and won’t be available for months to come. Also, you should consider how this will bind (lock) you to the entire ecosystem of Microsoft software solutions. It will sure be interesting to see how Microsoft will cooperate with different vendors of switches and other network equipment. Since SDN solutions are software solution in fact, they assume that the network stands for reliability and speed. But is this the case?

VMware NSX

By purchasing Nicira—a company made up of former Cisco engineers—VMware entered the SDN market. Cisco and VMware fought a great fight trying to purchase Nicira, and luckily for the industry, VMware triumphed. Otherwise, SDN solutions wouldn’t be developing as fast and as successfully as they are at this moment.

With its NSX solution, VMware is expanding its model of virtualized machines onto the data center network, placing its bets on both the cost aspect and security of the solution. NSX turns a physical network into a pool of network resources, to which you can relatively simply add virtual devices and security solutions. It is being used by more than 700 customers, and its main downside is actually only the price—VMware charges its licenses per processor (core), so the price can quickly reach high numbers. To make up for it, VMware offers an easy-to-use but highly-capable control system and the ability of micro-segmentation.

We have to grant VMware the recognition of being the industry-leading force in the field of virtualization (of practically anything), with Cisco and Microsoft merely following its solutions. Development is always an expensive aspect, you have to invest significantly.

NIL’s contribution

So who of the three players will in the long-term end up winning the game? NIL places its bets on all three players. We are independent and we use all three listed solutions. We have a variety of different implementations, including setting up the VMware NSX solution, planning the Cisco ACI environment, which we are setting up these days, and at the same time we are conducting detailed laboratory testing of Microsoft’s SDN solution. In short, we are ready for the future.

NIL is well aware of the limitations of these solutions. Even though SDN aims at freeing companies from being locked to a single vendor/solution, we all know that sooner or later they find themselves in a lock-in situation, whether regarding its hardware of software solutions. And let’s not forget that once you say “I do”, a divorce due to “irreconcilable differences” is difficult and mostly expensive.

NIL’s practical experiences with SDN are priceless. Separating networks from hardware (i.e. switches) brings a lot of network flexibility but only if the software-defines solutions are based on quality infrastructure. Often even IT specialists refuse to understand that white-label switches in itself have limitations—most support only a single-vendor software solutions. And that’s not what we want when we’re trying to avoid lock-in, do we? A lower price of these switches also comes with a greater risk—namely, who’s going to take responsibility when things don’t work? Whose team will take time to look for a solution? The hardware or software team? Established vendors have established partners with a stable ecosystem of (guaranteed) support. If we paraphrase this—even the best sports car can only go so fast if you put on a low-quality tires.