- Sandook software coordinates many SSDs to avoid slowdowns due to garbage collection
- Two-tier control system redirects workloads between clustered units in real time
- Performance gains are close to theoretical limits, but depend on large clustered storage environments.
Researchers at MIT and Tufts University have created a storage management system called Sandook that brings clustered SSDs closer to their theoretical limits. The project targets a long-standing problem within large storage clusters where identical drives rarely perform identically.
Solid state drives slow down for several reasons, including internal garbage collection cycles and the slower nature of write operations compared to reads. Such slowdowns can impact workloads when multiple applications share the same storage pool.
Instead of letting each SSD handle performance issues on its own, the system divides control tasks into two coordinated layers that manage activity across the entire group of drives.
Article continues below.
Unlocking the potential of SSDs for data centers
As Blocks and files A central controller reportedly collects performance telemetry from each large SSD and reviews scheduling options approximately 5 times per second.
Local agents within storage servers transmit performance signals and congestion warnings as workloads change.
When a unit begins cleaning tasks, such as garbage collection, the system reduces its priority and transfers traffic to healthier units in the group. That redirection occurs without requiring changes to the applications that access the storage.
The method builds on techniques already used in enterprise storage, including block replication for structured reads and writes to logs that can land on any available device.
Testing included database processing, neural network training, large-scale image compression, and latency-critical storage services, and the system was reported to deliver 30 to 82 percent higher raw input and output performance compared to previous approaches that targeted single bottlenecks.
In clustered workloads, application performance gains ranged from 12 to 94 percent, with latency reductions reaching up to 88 percent. In some cases, storage performance reached approximately 1.7 times previous levels.
The profits come entirely from software, meaning that the SSDs available on the market remain unchanged. The CPU and memory overhead for monitoring dozens of units per server was described as minimal.
The research paper, titled “Unlocking the Potential of Data Center SSDs by Controlling Performance Variability,” is available for viewing here.
Despite the headline numbers, this is not something most consumers can do at home. The design relies on large groups of SSDs working together, along with Linux-based infrastructure and common enterprise networking configurations in data centers.
That clustering effect is where most of the performance improvement comes from. Without spare drives to transfer workloads, a single-drive system would see little benefit.
Blocks and files notes that the work will be discussed at the USENIX NSDI 2026 event in May, where the researchers plan to show how coordinated scheduling helps solve the unpredictable behavior of SSDs in large clusters.
Follow TechRadar on Google News and add us as a preferred source to receive news, reviews and opinions from our experts in your feeds.
