Servers & Systems: The Right Compute

HPE Data Management Framework: Making storage bigger on the inside

How do you handle ever-increasing amounts of data without additional capacity drives? Make your storage bigger on the inside with the HPE Data Management Framework (DMF).

Biggeroninside_blog.jpgIf you know the TV series Doctor Who, you’ll recognize the running comedic line: “It’s bigger on the inside!” This line is typically uttered by a first-time visitor to the TARDIS, the space-time vehicle used by the Doctor. Thanks to the space-bending capabilities of the Time Lords, the interior of what appears to be a large blue police box (like the kind you see in the U.K.) on the outside has multiple levels of space on the inside, including a swimming pool.

Unfortunately, such technology doesn’t exist for storage use. So you have no choice but to buy additional capacity drives to accommodate the ever-rising tide of data. This strategy soon collapses under its own weight. Budgets are limited, and even with 200-to-300% over-provisioning, capacity may be insufficient. And with increased capacity, other factors must be considered: How can storage performance be maintained? How can even more data be backed up in the same time window? How can OPEX be controlled?

The solution is data management. Adding capacity without data management is like enlarging an airport without a control tower. Additional runaways may allow more planes to take off and land at the same time, but without a control tower, additional runways just increase the chances for an accident. (A TARDIS doesn’t need a runway, it just disappears.)

The HPE Data Management Framework (DMF) is the control tower for data. Built around an awareness that data is stateful (that is, depending on performance and access requirements), DMF provides for the active movement of data between those states to allow full utilization of storage resources.

Consider this analogy: When planes land at an airport, they may be cleaned and prepared and take off immediately again. Or they may be put in the hangar, and prepared and maintained for their next flight at a later date. Similarly, data may be in an active “hot” state, seeing immediate usage, or it could be in a “warm” state with fairly frequent usage or in a “cold” state when it is stored in archive. Like the plane, data may change states depending on the circumstances, and should be stored on the appropriate storage infrastructure for its state. “Hot” data is stored on SSDs because it needs the quickest access, it shouldn’t be stored on tapes.

DMF provides intelligent tiering of data through a built-in rules-oriented policy engine.

Administrators can create logical tiers of storage based on storage characteristics. That is, a “hot” tier is composed of flash memory and intended for data requiring immediate use, near-line hard drives are used for “warm” tiers, where access is required, but performance requirements are not as stringent, and tape is used as an archive tier for cold data, which doesn’t need immediate access.

Rules provide automation that place data where it needs to be to meet performance and access requirements. Performance requirements are always met as active data can be automatically placed on the fastest tier, using flash memory and SSDs. And as data ages, it is automatically moved to colder tiers, freeing up space for data requiring more immediate access, while lowering data TCO over time.

Conversely, data also needs to be retrieved from archive or colder tiers and made available for active usage. While all data is eventually archived, retrieval mData Management_blog2.jpgay be useful or necessary. A compliance audit may require pulling data from several years ago. A new project may require data from previous projects. Instead of a slow, tedious, manual process to locate and retrieve old data, DMF v7.1 introduces the concept of dynamic namespaces to simplify this process. A namespace is essentially a file structure, like a directory, of data. DMF can create, manage and delete namespaces. This provides great flexibility in managing data as all the relevant data based on specified criteria can be gathered, moved, and archived as necessary. For instance, all the data for a project can be gathered, processed and then archived.

The ability to use dynamic namespaces is based upon the use of a comprehensive metadata catalog which mirrors the file system, essentially maintaining an image of the file system over time. DMF v7.1 can create, delete and archive name spaces, including keeping track of versions of data and allowing fine-grained retrieval of data. Since all metadata information is stored in the catalog, data can be completely deleted, freeing resources for active data. DMF knows exactly where the data is stored. By providing a friction-less movement of data between tiers, it effectively expands the capacity of storage by using tape as an inexpensive archival medium as a bottom-less well to off-load warmer data tiers.

DMF solves the problem of how to backup petabytes of data.

The standard strategy of full backups followed with incremental backups simply can’t accommodate petabytes of data. There isn’t enough time in the backup window to handle the vast quantity of data. DMF allows for continuous backup of data, allowing data to be backed up when not in use, essentially providing an “infinite” backup window. And since the metadata catalog captures information about the location of the data, reconstitution of the data is greatly simplified and can be automated.

With dynamic namespaces and friction-less movement of data between tiers, DMF ostensibly provides larger storage capacity using the cheapest medium available. This allows your budget to be focused on the fastest, most expensive storage tiers—and less on capacity tiers, providing capex relief for storage infrastructure spending as well as minimizing OPEX for storage workloads. With data management from HPE, storage is truly bigger on the inside!

Learn more in this Hyperion Research Report: Technology Spotlight on Next Generation Data Management. (Registration required to download.)

About the Author


As VP & GM for HPC, I lead worldwide business execution and commercial HPC focus for one of the fastest growing market segments in Hewlett Packard Enterprise’s Hybrid IT (HIT) group that includes the recent Cray acquisition and the HPE Apollo portfolio.