How Bitwyre cut latency by 43% for lightning-fast crypto trades

Challenge

Time is money for financial exchanges

Bitwyre is a centralized trading platform that offers spot and derivative crypto trading services, where investors can trade Bitcoin, Ethereum, Zcash, Monero, Tezos, Cosmos, and other cryptocurrencies and stablecoins. 

In the volatile cryptocurrency market, where prices move in the blink of an eye, a single second could mean the difference between a user buying the last Bitcoin for a bargain or paying double the price half a second later. If you multiply this example by thousands of users, an inefficient crypto exchange quickly becomes an expensive nightmare.

With billions of dollars of daily trading volume, Bitwyre needed a real-time data platform that could support vast volumes of data generated every second — without sacrificing performance. The company also needed to swiftly track and deliver data without attenuation or loss.

Bitwyre initially looked to Apache Kafka^®, but performance tests showed latency spikes that would undercut the company’s stringent SLAs. Kafka also proved to be labor-intensive to maintain. Its reliance on Apache ZooKeeper™ and the Java Virtual Machine (JVM) called for significant in-house expertise, as well as constant tuning and re-tuning of system resources. 

“Time is of the essence, and systems cannot afford a single loss or low predictability — one could spell disaster.” - Khairil Akbar, NOC Infrastructure Engineer, Bitwyre

Why Redpanda

Built from the ground up for speed, simplicity, and cost savings

Bitwyre conducted a thorough evaluation and extensive testing when choosing its streaming data platform. Performance considerations included consistent low tail latencies, the ability to scale to massive volumes, and optimizing hardware resources to limit capital expenditures. Ultimately, they chose Redpanda due to the following: 

• Operational simplicity — less in-house expertise needed
• Low operational and capital expenses compared to Kafka
• Significantly lower latencies 
• Full compatibility with Kafka APIs

The team at Bitwyre had enough experience with Kafka to know the high administrative overhead it required, so Redpanda proved to be a massive sigh of relief as it was vastly simpler to deploy and operate. To seal the deal, internal performance tests comparing Kafka vs. Redpanda revealed 43% lower P99 latencies and a significantly smaller hardware footprint. 

Results

Streamlined data pipelines for a lightning-fast trading

Switching from Kafka to Redpanda instantly created a 19.5% boost in the average time data is transferred between Bitwyre’s internal backend system — a cut of almost 0.1 seconds.

“For a crypto and derivatives exchange like us, it’s the difference of billions of dollars of daily trading volume.”

With Redpanda’s highly performant design, Bitwyre slashed its P99 latencies by 43%, significantly improving the platform’s ability to execute timely transactions while also reducing the company’s maintenance and infrastructure costs.

Redpanda’s developer experience was another major benefit, according to the Bitwyre team. “Redpanda follows the same Kafka APIs and tools, so the move to Redpanda was as easy as changing a few lines of code,” Khariril says.

Moreover, due to Redpanda's all-in-one nature, which skips the need for external dependencies, Bitwyre’s developers could rest easy knowing their program would run as efficiently as possible on their machine without any further modifications.

Today, Bitwyre continues to rely on Redpanda for its simplicity, the quality of its support, and its ability to reliably deliver latency-sensitive real-time messages across its system.