Redis replaced: Six case studies in superior performance

In the ever-evolving data management landscape, businesses are continually seeking ways to enhance efficiency, reduce costs, and improve performance. Redis has been a popular choice for many, but several organizations have found even greater success by transitioning to alternative solutions. This post explores six real-world case studies where companies achieved remarkable improvements in efficiency and performance by replacing Redis.

1. AppsFlyer: Scaling with precision

The challenge: AppsFlyer, a mobile attribution and marketing analytics platform, had to scale its operations to handle an exponential increase in data volume.

AppsFlyer migrated from Redis to Aerospike to establish the world's largest real-time mobile device database, drawn in by the latter's superior performance and scalability. While Redis is well known in the NoSQL database space, it struggled to cope with the increasing volume and the mission-critical speed AppFlyer’s business requirements demanded. Aerospike delivered high throughput, minimal latency, and enhanced horizontal scalability, enabling AppsFlyer to handle billions of daily transactions effectively. Not to mention, Aerospike's strong consistency and durability were critical solutions for upholding AppsFlyer's data integrity, ensuring reliable and accurate real-time analytics.

"Before we had Aerospike," says AppFlyer software architect Ido Barkin, “we had a very big installed database that was very unstable, [that] was composed out of almost 50 instances. The cluster was very unstable. Daily backups took more than a day to accomplish, and we were not happy with support – this is an understatement. We couldn’t get any proper help with the instability issues. We were in constant fear that something would happen to this database.

The first step that made us move to Aerospike was stability. Our first move with Aerospike was to replace our most mission-critical database. It was the user installs database. We migrated around two billion records to a new Aerospike cluster, which was more stable and smaller. After we finished the migration, we immediately moved to add new features that depend on it.”

The solution: By switching from Redis to a more advanced data solution, AppsFlyer scaled from 1.4 billion devices to 5 billion in the first year. This transition also significantly reduced false positives, cutting them from 0.1% to 0%.

The outcome: 

• Improved scalability

• Enhanced data accuracy

• Reduced false positives

For more on AppsFlyer’s successful transition, read "How We Built the World’s Biggest Real-time DB of Mobile Devices."

2. Adjust: Optimizing fraud prevention

The challenge: Adjust’s core business involves validating billions of advertising impressions and effectively combating fraudulent activities on a large scale. Adjust needed to optimize its server usage and improve data accuracy to strengthen its fraud prevention measures. 

Adjust shifted from using Redis to Aerospike because of the serious scalability and performance obstacles encountered throughout the company's expansion. Initially, Redis seemed a good fit - serving fast response times and delivering high throughput. Yet, as Adjust's infrastructure expanded from a few servers to well over a thousand, with requests per second jumping to an eye-opening 750,000, Redis's constraints suddenly stood out in sharp relief. Redis’ single-threaded design created significant bottlenecks, inevitably leading to latency spikes and costly manual interventions for failover and data management. The challenges were only further compounded by rising expenses derived from vertical scaling, largely thanks to pricey hardware requirements for RAM, server nodes, and storage. For these reasons, Adjust decided to explore a more effective solution.

"When it comes to our use case,” says Adjust’s VP of Engineering Robert Abraham, "Aerospike acts very much like we wanted Redis to behave. It has clustering built in, and you can seamlessly add and remove servers from your cluster. So, only the lookup index, which in our case is the hash of device identifiers, is kept in memory, and the data itself is stored on SSDs. This means that the majority of data can be kept on storage that costs one-tenth [of Redis’ costs], and as described earlier, we needed fast storage anyway. We were able to reduce our cluster node count from 40 to 6. This means we were able to cut our infrastructure costs by almost 85 percent and we also needed to maintain it a lot less.”

The solution: Adjust replaced Redis with Aerospike, enabling it to reduce its server count from forty (40) to six (6). This change improved latency, failover capabilities, and data accuracy, helping to prevent overall ineffective campaign spending.

The outcome:

• Significant reduction in server infrastructure costs

• Improved latency and failover

• Enhanced data accuracy

Watch the presentation, "User attribution at scale: All-flash, ½ trillion objects” to hear more about Adjust's journey.

3. Wayfair: Streamlining e-commerce operations

The challenge: Wayfair, the global e-commerce giant known for its furniture and homewares, aimed to enhance operational efficiency by minimizing server load and improving customers' checkout experiences.

Wayfair transitioned from Redis and Memcached to Aerospike to meet the growing scale of its AdTech infrastructure. As expansion continued, its legacy Redis setup was incapable of processing the sheer volume of data that followed and the low-latency parameters it demanded. After considering Cassandra, among other options, Wayfair chose Aerospike for its superior performance, easy data management, and ability to scale up and scale out. The transition drastically reduced overall costs and reduced its server count from 60 to 7 while improving data processing speeds and system reliability.

Wayfair's AdTech team manages millions in ad spend yearly and relies on real-time data to optimize advertising strategies. Aerospike's high performance and scalability were crucial for Wayfair's recommendations, event tracking, and various other applications, allowing growth to remain on track while advancing technological innovation. Wayfair’s transition over to Aerospike exemplifies its commitment to enhancing the overall customer experience and operational efficiency.

The solution: By migrating from Redis and Memcached, Wayfair reduced its server footprint to just 1/8th of its previous size. This transition also improved the average cart checkout size by 6%, which is huge in the e-commerce world.

The outcome:

• Significant decrease in server overhead

• Customer experience enhanced by improved checkout process

• Boosted average basket value

Watch "Moving sofas in millisecond time" from Aerospike Summit for full details.

4. ironSource: Enhancing ad serving reliability challenge

The challenge: ironSource, a prominent AdTech firm, sought a dependable data solution to curb inefficient ad delivery and enable the introduction of innovative business functionalities.

ironSource transitioned to Aerospike from Redis to improve its ad delivery system. They had been facing challenges in scaling and managing high I/O demands. Aerospike was chosen for its performance, scalability, and ease of integration. The migration led to significant improvements: reduced server count, enhanced performance with up to 130k reads and 75k writes per second, and near 100% uptime with sub-millisecond latency. This enabled ironSource to enhance ad campaigns' precision, refine user interactions, bolster business results, and streamline DevOps maintenance.

The solution: ironSource chose a data solution with improved reliability and lower latency compared to Redis. This transition prevented inefficient ad serving and enabled the implementation of new business functionalities.

The outcome:

• Improved ad serving reliability

• Enhanced latency performance

• Introduced new business features

Watch "Edge databases and systems of record" here. 

5. TomTom: Reducing operational costs and carbon footprint

The challenge: TomTom, a global leader in navigation and mapping solutions worldwide, set out to slash operational expenses and shrink its environmental impact.

TomTom's move from Redis to Aerospike showcases a strategic shift driven by the need for enhanced efficiency and sustainability. In a recent panel discussion, Sustainability in Data Management, companies emphasized reducing environmental impact, with data management playing a pivotal role. The transition to Aerospike is part of TomTom's broader effort to optimize data infrastructure, minimize server footprints, and reduce CO2 emissions.

Efficiency in data storage and processing plays a crucial role in driving this migration. TomTom's traffic service is heavily dependent on seamless data ingestion and indexing. By harnessing Aerospike's exceptional performance and minimal latency, TomTom can manage substantial data volumes more effectively. This not only enhances performance but also minimizes its environmental impact.

The sustainability focus of this migration is critical. Data centers are major contributors to CO2 emissions, underscoring the critical need for operational optimization to bolster environmental sustainability. TomTom aims to halve its server footprint by migrating to Aerospike, aligning with global trends and regulatory pressures like the European Union's Corporate Sustainability Reporting Directive (CSRD).

The solution: Replacing Redis allowed TomTom to handle a 1 TB use case with significant hardware reductions, saving 93% in costs. This change also prevented outages, saving €100,000 per avoided outage, and reduced its carbon footprint by 86%.

The outcome:

• Significant cost savings

• Improved reliability and uptime

• Significant reduction in carbon footprint

Watch the "Sustainability in Data Management” webinar now.

6. Wix: Improving personalization and cost efficiency

The challenge: Wix, a popular web development platform, needed to enhance personalization features while cutting costs and reducing latency.

At an AWS re:Invent 2022 presentation, Tal Gabay, Engineering Manager at Wix, relayed the issues that the company was facing with its user personalization service. “Our user personalization service, aka profile service, is a real-time fact store,” he says. “It is integrated with most of Wix's internal services, and actually, recent mapping that we've done showed that over 70 development teams are integrated with that service in one way or another.” This service leverages complex insights from real-time user behavior and business intelligence data to drive intelligent, dynamic decisions across various use cases, such as identifying specific user groups and running machine learning models. However, the service needed to handle enormous amounts of data. “This is a self-service computing over 3,500 facts,” he shares. “It serves 50,000 reads per minute, peeking at 5 million writes per minute and we calculate everything for over 1.9 billion different keys.”

Initially, Wix used a combo of both Redis and HBase for its data management. When users logged into Wix, its data was loaded from HBase into Redis for fast access and updates. Initially, this setup seemed to be successful, until scalability issues arose, leaving Redis incapable of handling all the data in real-time, especially for users who hadn’t been recently active. They had to drop inactive users' data from Redis after three days and introduce heavy snapshot replication to manage data consistency. With data for inactive users not readily available, it impacted the ability to customize user experiences in real time.

To solve these issues, Wix replaced its two-layer infrastructure of HBase and Redis with Aerospike. This change streamlined its system from two layers to one, eliminated the need for the heavy daily snapshot job, and ensured all data was consistently available in real time. The move to Aerospike significantly reduced the amount of code to manage, removed legacy issues, and ensured that all user data was accessible in real time, enhancing both internal operations and user experiences.

The solution: By transitioning from Redis and HBase, Wix saved 45% in costs and reduced latency from 18ms to just 2-3ms. This improvement enabled improved personalization and bolstered the responsiveness of the user experience.

The outcome:

• Substantial cost savings

• Drastic reduction in latency

• Enhanced personalization capabilities

To hear Wix’s Tal Gabay tell the story of the migration from Redis to Aerospike himself, watch the full presentation.

Ready for a replacement?

These six case studies showcase Redis' strengths, yet reveal that alternative data solutions may deliver higher performance, increased efficiency, and substantial cost savings. By thoughtfully selecting optimal data management tools, businesses can drive remarkable transformations and maintain a competitive edge in the market.