Every production database drags a large tail of infrastructure behind it. On average, organizations make 8 to 10 copies of each production database, for development, unit test, cutover test, reporting, and other purposes. Each of these copies requires duplication of the data within the original database, so 8 copies of a 5 TB database requires an additional 40 TB of storage, in addition to any initial provisioning processes. If development or testing require updates from the production system, then additional effort and storage are needed to refresh the copies.
In addition to creating these copies, organizations also create backups and archives of both the original database and the duplicates. Full disk backups typically happen once per week, with incremental backups occurring daily, for both production and some of the copies. At regular intervals, the production system and the copies may be mirrored to alternate sites or archived to tape, for disaster recovery. In short, a single 5 TB production database can easily spawn over 60 TB in copies, backups, and archives. The change control and decision-making processes around this infrastructure delays projects significantly and can prevent firms from bringing useful applications to market in a timely manner.
Delphix eliminates the need for this infrastructure tail and its related bureaucratic delays. Delphix virtualizes the data files within a database by creating a single, highly-compressed copy of the original data blocks, then serving that data to multiple DBMS servers. Each DBMS receives its own fully functional read/write database, and the results are completely transparent to both applications and users. However, each new copy does not create new storage demands, new approval requirements, etc.
In addition to managing updates from each application user, Delphix also manages, via standard, published database APIs, updates from the original production database. As a result, the Delphix virtual appliance contains both a copy of the initial data set and all changes over time to that initial data set. This enables Delphix to provision multiple copies of a production database, at different points in time — a time machine for enterprise databases.
The time machine capability supports better testing and reporting, but also allows a user to recover quickly from application errors. Instead of unloading and reloading a new copy of the database from backup, a Delphix user can simply move a slider to a desired point in time, click a button, and restore a database copy to that point in time. In this way, Delphix significantly enhances the disaster recovery architecture a firm may already have in place.
For example, a firm puts a new application update into production and the update contains a currency-handling bug. Users placing orders for France enter amounts in Euros, and the orders are mistakenly executed in dollars (in this example, a $10M loss).
The firm discovers the mistake after hundreds of orders have been placed, and needs to roll back the database to a correct point in time after the logic has been repaired. It uses Delphix to open up several parallel virtual copies at different points in time and quickly finds the proper point for recovery. Without Delphix, the firm would need to load a backup copy, check it, unload, load an earlier version, check, unload, etc. With Delphix, the process is fast and simple, leading to quick, cost-effective recovery.
