June 23, 2026
Route drama enters the shard era
Scaling Akvorado BMP RIB with Sharding
After years of headaches, users say Akvorado may finally stop choking on giant routing tables
TLDR: Akvorado says it has a real fix for a long-running scaling problem by splitting its giant route database into smaller pieces for faster updates. The community’s mood is mostly relieved but cautious, with praise for the persistence and jokes about whether "close enough" routing guesses will come back to haunt people.
A very niche infrastructure problem just got the kind of glow-up that makes server people cheer in all caps. Akvorado — a tool companies use to match internet traffic with routing info — has been struggling with massive route lists for ages. We’re talking over a million routes, with some users needing it to handle tens of millions without falling over. The big fix? Split the giant routing database into smaller chunks so updates can happen at the same time. In plain English: instead of one overloaded checkout line, Akvorado opens more registers.
And the community reaction has the full spectrum of internet emotion. The loudest camp is basically saying, "finally". After years of attempts, patches, and issue threads, some users sound relieved that this long-running pain point may actually be under control. Others are a little more side-eye about the compromise: if one device doesn’t have a route, Akvorado can borrow one from another device as an "good enough" estimate. Supporters call that practical; skeptics call it the networking equivalent of guessing with confidence.
The funniest reactions are classic engineer humor: jokes about "sharding" sounding like a fantasy weapon, and comments treating the routing table like a monster that keeps growing no matter how many fixes you throw at it. There’s also admiration for the sheer persistence here — multiple failed or partial fixes, months of work, and yet another redesign. The vibe is equal parts battle-scarred relief, cautious optimism, and nerdy applause for finally making the beast behave.
Key Points
- •Akvorado uses BMP to import routing data so it can attach AS paths and BGP communities to observed flows.
- •The article says Akvorado must scale from Internet routing tables with more than 1 million routes to tens of millions of routes, and introduces RIB sharding as the fix.
- •Akvorado can fall back to routes exported by another device when a route is missing from a specific device, leaving the operator to judge whether that approximation is acceptable.
- •The previous RIB design combined a prefix tree based on the bart package with route storage in a Go map keyed by combined prefix and route indexes.
- •To reduce garbage-collector pressure, Akvorado avoids storing per-prefix route arrays directly in the tree and instead uses prefix IDs, route maps, and interned structures for NLRIs, next hops, and route attributes.