The short answer is quite simple — for the people and technology.
Let me elaborate on both:
Finding a group of like-minded people (who also happen to lead the company) who excel in their core competencies but are open to learn and listen to people who are stronger in theirs, sounds like a “mission impossible,” but, rephrasing an old saying – “those who seek, find.”
Talking about distributed systems with David Cheriton, discussing graph partitioning strategies with Sasha Ratkovic and future industry directions with Mansour Karam, amount and quality of knowledge they have got and willingness to share. But also – focus on execution, building the best product.
There’s very little else I could wish for a “perfect job.”
I got very excited about programmable networking in early 2010s, when the fact that we (networking industry) got stuck had become rather obvious.
A Little History
That initial excitement led me to focus on technologies that facilitated better knowledge about the network by extracting and then modeling structured data from and about the network. This, as one might expect, involved development of data distribution over existing channels such as link-state IGPs, BGP-LS, and new ones, such as Netconf (Restconf/gRPC) in addition to data modeling. This drew me into work on YANG efforts in IETF, Openconfig, and data modeling for networking with other organizations.
Very early in that phase I realized and began to promote the underlying principle that “we must decouple reachability from policies” which would allow us to scale both independently. The premise of this idea was to leave reachability distributed in the network while logically centralizing policies, data stores reflecting the various states of the network, and telemetry validating those states. Russ White and I wrote the book Navigating Network Complexity which details the logic behind these principles.
The ultimate “distributed” reachability ideas manifested themselves in the IETF RIFT (Routing in Fat Tries) working group initiative.
Let’s Jump to Today
Looking back, we, as the networking industry did a great job building the data modeling layer with near real-time performance, at scale telemetry (gRPC and YANG Push), and using a standardized representation that facilitated programmable networking. So we now have enough data to actually do something tangible about it!
The missing part is obviously how to make the data actionable and how does one apply this, rather abstracted term to networking. Another missing element has been transforming business logic to networking with a continuous validation layer which Intent-Based Networking addresses.
Doing even simple correlation between the intended (intent) and operational (applied/derived) states is anything but trivial. Doing this at scale and including continuous validation makes it an extremely complicated problem to address. This is made even more complicated as networks continue to grow in size, scale and complexity (think 5G, Edge Computing and similar). A massively scalable platform with a high performing backend are “must have” attributes of a solution that can scale and be able to address these challenges. The need to combine this with a graph that will allow the system to query and reason about the data, rationalize its meaning and relevance, and incorporate highly flexible analytic probe pipelines, gives you just a few hints of the underlying sophistication of the solution required to solve this problem.
Apstra presented at NFD16 (Network Field Day) where I had the opportunity to be a panelist. I was immediately impressed by the massively scalable and distributed design, the sophisticated in memory processing, and a pluggable architecture operating over sub-pub and API-driven logic. This all combined seamlessly with a graph infrastructure that represented and correlated states and events to make Apstra Intent-Based Data Center Automation uniquely capable of addressing and leading Intent-Based Networking and empowering the Intent-Based Data Center.
That is why I’m here, excited, building the future with the team one could only dream of!