Have you ever wondered how to handle very large databases in the cloud? Look no further than Azure SQL Database – Hyperscale edition. This powerful edition, which became generally available in May 2019, offers unlimited data size and a range of features designed to push the limits of large systems.

Even if you don’t need a 100TB database, running a 5TB system can be expensive and challenging. Hyperscale edition allows you to remove most of the operational costs associated with managing a large system, making it an attractive option for organizations of all sizes.

Hyperscale Capabilities

The key feature of Hyperscale edition is its unlimited data size. While the documentation and literature mention a limit of 100TB, there is no actual size limit when you scale up. You can start playing with Hyperscale edition with a minimum size of 10GB.

With Hyperscale edition, you have the ability to scale out to multiple read-only replicas, allowing you to offload read workloads and improve performance. Currently, you can have up to 4 replicas, but this number may increase in the future.

Another important feature of Hyperscale edition is the option to choose between Gen4 and Gen5 hardware, which determines the number of cores and memory available on each instance. The scale-up of cores and memory is managed by Microsoft and cannot be altered.

One of the standout features of Hyperscale edition is its high volume log structure, which enables quick data recovery. In a 50TB demo, the restore of a database to a point in time in a separate Hyperscale database took just 14 minutes. Additionally, backups are made with file snapshots, eliminating the performance impact of traditional backup processes.

Architecture

Hyperscale edition is built on the same architecture as Azure SQL Database, meaning there are no new syntax or application changes required. The architecture of Hyperscale edition is designed to handle the scale of very large databases.

At the top layer, clients connect to one writable replica and up to 4 read-only replicas. These replicas are SQL Servers built according to the hardware specifications you choose. Each replica uses local SSD storage as a high-speed cache, allowing for quick data retrieval.

If data is not in the cache, there is another layer of servers called Page Servers. These servers hold cache data and serve it to the compute layer as needed. As your database grows, more Page Servers are added to support the load.

At the bottom layer, the actual data resides in Azure Standard Storage, where it is accessed and updated by the Page Servers. The architecture ensures that data changes are propagated to all replicas and Page Servers.

Backup and Restore

In Hyperscale edition, backups are no longer a concern. Log records are maintained separately, eliminating the need for log backups. Data backups are file system snapshots made by Azure storage, and restores are performed by restoring these snapshots and replaying the log records.

Restores in Hyperscale edition are designed to be fast and efficient. Multiple file snapshots are taken at different times, and the closest snapshot before the restore point is selected for each file. The log records are then replayed on the restored files, ensuring data consistency.

The goal of a restore in Hyperscale edition is to take approximately 10 minutes, regardless of the size of the database. This makes restores a quick and predictable process.

Pricing

The pricing for Hyperscale edition is based on the number of cores and the storage required. These are billed separately and increase as you scale up your resources. Hyperscale edition can be more cost-effective than the Business Critical tier, especially as you scale.

It’s important to experiment with the pricing sliders to find the best configuration for your needs. Additionally, testing your workload against Hyperscale edition for a day or two can help you determine its scalability and performance benefits.

Summary

Azure SQL Database – Hyperscale edition is a powerful solution for handling very large databases in the cloud. Its multi-layer architecture and unlimited data size make it an attractive option for organizations looking to minimize the operational costs of managing large systems.

If you’re tired of dealing with the challenges of managing very large databases, Hyperscale edition is definitely worth exploring. Its scalability, performance, and ease of use make it a compelling choice for organizations of all sizes.

References:

• Hyperscale Service Tier for up to 100TB
• What is Azure SQL Database Hyperscale? (video)
• How Microsoft is Building the Biggest Cloud Database
• Hyperscale FAQ