TinyTricia – A Space-Optimized Patricia Trie For Transparent Access to Edge Computing Services

Abstract

Multi-access Edge Computing (MEC) is an essential piece of 5G telecommunication systems to satisfy the challenging low-latency demands of future applications. Our previous publications argue that edge computing should be transparent to clients. We introduced an efficient solution to implement such a transparent approach, leveraging Software-Defined Networking and virtual IP+port addresses for registered edge services. A core component of our architecture is a Patricia Trie, which stores all our virtual IP+port addresses. Unfortunately, most implementations of Patricia Tries are not geared toward use cases with millions of keys where a low memory footprint becomes essential. In this paper, we present TinyTric$\dot{w}$, a space-efficient open-source implementation of a Patricia Trie for keys up to 256 bits. TinyTricia can keep track of up to half a billion (229) keys $\leq$57 bits and up to a quarter of a billion (228-1) keys $\leq$256 bits with tiny memory requirements. In the latter case, each key can have a data value of any type. Keys $\leq$57 bits require only 8 to 16 bytes per key (i.e., only 0 to 8 bytes overhead per 57-bit key); for keys $\geq$58 bits, add the key size (in whole bytes) to these values. Thus, for 16.78 million (224)57-bit keys, our solution requires between 128 and 256 MiB of memory. Unlike some other spaceefficient implementations, TinyTriia allows adding and removing keys at runtime.