“Simplicity is the ultimate sophistication.”
Clare Boothe Luce
According to a recently published survey by Cisco Systems, Inc., IT organizations spend 67% of their budgets on operational expense, and consume 80% of their time in the process. That is a staggering amount of time and money spent on simply managing the technology infrastructure of an enterprise. It comes as no surprise, then, that so many people are focused on reducing those numbers, and on increasing time-to value of new technology in the process.
Glue Networks is one of the companies trying to solve this problem by bringing automation to the network. Jeff Gray, CEO of Glue Networks, claims that they have built the “first model driven, multi-vendor, software defined network, orchestration platform that allows organizations to control their networks in a new way.” That’s a lot of words, and a bold claim, to be certain, but there’s many a slip twixt the cup and the lip, and claims are easy to make. Proving them out is more difficult.
One of the challenges in today’s networks, particularly when speaking about automation, is the wide variety of products, platforms, hardware, circuits, etc., that exist and need to be controlled. If I want to make a change in my QoS policy across a specific path in my network, for instance, I may have to touch several brands of gear, as well as several different models within one vendor’s product line. Automating any kind of task, let alone a complex change, has historically been extremely difficult. Even with great scripting, it’s difficult to make changes everywhere at once.
Glue has taken a unique approach to this problem by creating data-driven models based on intent. In other words, their Glueware Control platform looks at what the intent of the operator is in requesting a change. You’re operating at a higher level of abstraction from the raw gear to be changed, and letting the control platform figure out how to execute the changes needed in order to fulfil your intent. You tell the system what hardware you have, and it uses its knowledge of that hardware to execute changes.
Whenever you abstract another layer above a hardware change, you rely more and more on the accuracy of your models and formulas to tell the hardware what needs to happen. If your automation engine can only talk to three models of switches, it is not going to be spectacularly useful. Currently the Glueware Control platform comes with models for 13 different multi-vendor packages and operating systems, with many more on the way. These are the recipes for how the system talks to your gear, so more is always better. The current goal, according to Gray, is to release a new vendor package every three weeks.
Glue also announced the release of the Glueware Community, which is what they’re calling their user-driven, online, ecosystem for collaborating with fellow users. Here is where a robust community of users exchanges recipes and formulas that they have written, which may not be something that Glue has released themselves. In other words, maybe you have a somewhat rare device that few people have, and no one has written a model for it yet. You wrote a model (there are plenty of examples and instructions on the community site) to support your unique device, put it into the community repository, and now other users can benefit from your solution. This is quite a good way to both encourage community participation, and to rapidly increase adoption by growing the models and formulas repository in leaps and bounds.
Responding to customers, Glue has also taken their traditionally cloud-based platform and extended it into an on-premises solution, saying that many customers needed a “behind the firewall” solution. They have also expanded from a purely WAN-based solution, into both LAN and datacenter environments, extending their usefulness across the whole of the enterprise, rather than simply running as a point solution.
Another feature that is extremely promising is the ability for the platform to dynamically create models, based on your gear, in a brownfield environment. You can install this product, and based on what it knows, it will model your network devices and pull them into the system. This shortens the time to value equation by allowing users to immediately derive value from the product, something which helps to prevent an expensive purchase from becoming shelf-ware.
All in all, I’d say that Glue has made great strides in this release, and definitely is at the forefront of vendors providing solutions to one of the most pressing issues of the day. While many other products and solutions purport to solve the automation problem, reducing operational expenses and staff utilization, far too many require large investments in what are ultimately non-vendor agnostic ecosystems. These latter systems tend to move the problem from OPEX to CAPEX, while introducing tremendous amounts of complexity. Glue offers a solution that is both simple and powerful, and should definitely be something you take a look at implementing.
For more information on the platform, how it works, and what problems it solves, take a look at this presentation by Olivier Huynh Van, CTO and Co-founder of Glue Networks: