DIRECT
discontinuedby University of Wisconsin-Madison (David DeWitt et al.) · since 1978 · also known as DIRECT database machine
Wisconsin's late-1970s multiprocessor database machine, and the project that produced the Wisconsin Benchmark.
discontinued. Research prototype (c. 1978-1984) at UW-Madison; ideas carried into Gamma and later shared-nothing parallel databases.
Overview
DIRECT was a relational database machine built at the University of Wisconsin-Madison by David DeWitt's group and first described at ISCA in 1978 (with a fuller IEEE Transactions on Computers paper in 1979). A controller decomposed relational queries into operators on fixed-size pages and assigned them dynamically to a pool of query processors that shared a bank of page-sized memory modules, so intra-query and inter-query parallelism came from the same mechanism. Benchmarking it against commercial systems led to the Wisconsin Benchmark, which embarrassed several vendors and shaped database benchmarking for a decade. DIRECT itself was never a product; its lessons fed directly into DeWitt's shared-nothing Gamma machine and, through it, today's parallel SQL engines.
Strengths & trade-offs
Strengths
- +Early demonstration of page-level, dynamically scheduled intra- and inter-query parallelism
- +Origin of the Wisconsin Benchmark, a template for later database benchmarks
- +Well-documented in ISCA 1978 and IEEE ToC 1979 papers, useful for teaching database-machine history
Trade-offs
- −Hardware prototype that was never released as software
- −Shared-memory design hit bandwidth limits that pushed the field to shared-nothing (Gamma)
Great for
- +Studying the history of parallel query execution
- +Teaching the database-machine era and why it gave way to shared-nothing clusters
- +Background for understanding the Wisconsin Benchmark
Think twice for
- −Anything beyond study: there is no runnable system
Capabilities
Native or official first-party support · 2 of 25 tracked
Scale & consistency
- Comfortable data size
- Small · ≤ 100 GB
- Horizontal scaling
- Native
- Multi-region
- No
- Transactions
- Limited ACID
- Consistency
- Strong
- Schema
- Strict
- Architecture
- Shared storage
- Query languages
- Relational query trees issued by a host system
Parallelism came from adding query processors that shared page-sized memory modules.
Deployment
Alternatives to DIRECT
Wisconsin's 1980s shared-nothing parallel database machine that proved hash-partitioned query plans.
direct vs gamma →The original MPP data warehouse, now VantageCloud with open table formats and in-database analytics.
direct vs teradata-vantage →IBM's FPGA-accelerated MPP warehouse, now offered as NPS on-prem, in Cloud Pak and as a service.
direct vs ibm-netezza →PostgreSQL-based MPP data warehouse, closed-sourced by Broadcom in 2024 and sold in Tanzu Data.
direct vs greenplum →Java ROLAP engine that answers MDX queries by generating SQL against your relational warehouse.
direct vs mondrian →Google's internal distributed SQL database on Spanner that replaced sharded MySQL for AdWords.
direct vs google-f1 →DIRECT FAQ
What is DIRECT?
DIRECT is a relational database by University of Wisconsin-Madison (David DeWitt et al.), first released in 1978. DIRECT was a relational database machine built at the University of Wisconsin-Madison by David DeWitt's group and first described at ISCA in 1978 (with a fuller IEEE Transactions on Computers paper in 1979). A controller decomposed relational queries into operators on fixed-size pages and assigned them dynamically to a pool of query processors that shared a bank of page-sized memory modules, so intra-query and inter-query parallelism came from the same mechanism. Benchmarking it against commercial systems led to the Wisconsin Benchmark, which embarrassed several vendors and shaped database benchmarking for a decade. DIRECT itself was never a product; its lessons fed directly into DeWitt's shared-nothing Gamma machine and, through it, today's parallel SQL engines.
Is DIRECT open source?
No. DIRECT is proprietary, closed-source software. License: Not released (research prototype).
Does DIRECT support ACID transactions?
Partly. DIRECT offers limited ACID guarantees (for example, single-record or single-partition transactions), with strong consistency.
Does DIRECT scale horizontally?
DIRECT scales horizontally out of the box, with built-in sharding or partitioning. It is comfortable at small scale (≤ 100 GB).
What is DIRECT best for?
DIRECT is a strong fit for studying the history of parallel query execution; teaching the database-machine era and why it gave way to shared-nothing clusters; and background for understanding the Wisconsin Benchmark. Think twice for anything beyond study: there is no runnable system.
What are the best alternatives to DIRECT?
Common alternatives to DIRECT include GAMMA, Teradata Vantage, IBM Netezza, Greenplum and Mondrian. The right choice depends on your workload, scale and hosting; WhichDB's smart match ranks them against your requirements.
Is DIRECT still maintained?
DIRECT is discontinued. Research prototype (c. 1978-1984) at UW-Madison; ideas carried into Gamma and later shared-nothing parallel databases.
How popular is DIRECT?
DIRECT has little measurable public adoption today, so it ranks near the bottom of the WhichDB Score.