BeeGFS Storage Servers

BeeGFS

BeeGFS has been developed with performance and design as a priority as well as simple installation and management. If you expect to be dealing with I/O intensive workloads, BeeGFS is the way to go.

BeeGFS is a leading parallel cluster file system, that was specifically designed to manage I/O intensive workloads in environments where performance is critical. The system has received global acclaim from leading enterprises, universities and researchers for its ease of installation, maximum scalability, robustness and amazing flexibility. BeeGFS improves productivity through delivering results faster, and enabling new methods of data analysis that were impossible before its inception.

Features of the Filesystem

Broadberry storage solutions configured with BeeGFS transparently spread user data across a multitude of servers. Through increasing the number of servers and disks in the system, you can easily scale performance and capacity from small clusters all the way to enterprise-class systems with thousands of nodes.

Maximum Usability

All BeeGFS components can be installed and updated without having to reboot. For installation and updates, you can find the available rpm/deb package repositories available; for the start-up mechanism, super-simple-to use system service scripts are included.

BeeGFS was designed for easy administration, the graphical administration and monitoring systems allows for you to deal with the usual management tasks in a straightforward and intuitive manner, while also offering a command link interface. This includes cluster installation, health monitoring, storage service management, live load statistics (including for individual users) and more.

Broadberry engineers can install and configure BeeGFS on your storage solution, or if you prefer to do it yourself there is an abundance of documentation available to help you have everything up and running within an hour.

High Availability

Mirroring storage and metadata with high availability is based on what is referred to as “buddy groups”. Generally, a buddy group is a pair of targets that manage data replication internally between each other. This “buddy group” approach allows for all data to remain accessible even in the unlikely event of half of all servers within a system failing.

It’s also used to place buddies in different failure domains or fire domains (for example, different racks or differing server rooms).

Built-in BeeGFS Buddy Mirroring automatically replicates data, deals with storage server failures transparently for actively running applications, and grants automatic self-healing once a server comes back online, only resyncing the files that have been altered while the machine was offline.

You can also use storage server buddy mirroring with an odd number of storage servers. This works due to buddy groups being made up of individual storage targets, independent to their assignment to servers.

Fault Tolerance

Broadberry’s enterprise-grade storage solutions configured with BeeGFS are generally used with an underlying RAID, in order to transparently handle potential disk errors. It’s also possible to handle server failures by using BeeGFS with shared storage. The Buddy Mirroring approach, which is built-in, takes an additional step through tolerating the loss of complete servers including all of the data on their RAID volumes.

Quota Enforcement

When you configure BeeGFS on Broadberry storage appliances you’re able to define system-wide quotas of disk space allocation and the number of chunk files either on a per-user or per-group basis. This can be used to organise users in different access layers with differing levels of restriction, and also to prevent individual users from overusing resources.

The BeeGFS quota management mechanism is made up of two features:

The first is quota tracking, which enables querying the amount of data and the number of chunk files that users/groups are using within the system. It does not impose any restriction.

The second is quota enforcement. It allows for the definition and the application of quota limits. When this feature is enabled, the BeeGFS management daemon collects quota reports from all storage targets periodically. It checks for exceeded limits and then informs the rest of the system which users will no longer be allowed to consume any additional resources.

Storage Pools

For users to fully harness all-flash performance, the BeeGFS storage pools features makes flash drives transparently available to users, allowing them to request (via the beegfs-ctl command line tool) moving the current project to flash drives. When finished, the user can then move the project back to spinning disks.

BeeGFS introduced the storage pools feature in order to enable the cluster admin to group targets and mirror buddy groups together in different classes. For example, there can be a single pool consisting of fast solid-state drives, and a separate pool for bulk storage utilising large capacity but slower spinning disks. Pools can be given descriptive names, making it a lot easier to remember which pool to use without having to look up the storage targets.

Advantages of Using BeeGFS

There are a wide variety of advantages you can enjoy when configuring BeeGFS on Broadberry storage solutions. One fundamental concept is the meticulous evasion of architectural bottle necks. Another key advantage is the distribution of file system metadata over a multitude of servers, which larger metadata-expensive systems and applications can greatly benefit from.

File system nodes are built on very efficient and highly scalable multithreaded core components that feature native InfiniBand support. They can serve both InfiniBand and Ethernet (or in fact, any TCP-enabled network) connections concurrently, and will automatically switch to a redundant connection path in the event that one of them fails.

This parallel cluster file system is very easy to use and utilises a patchless kernel module, negating the need for frequent patching. It comes with its own graphical cluster installation tools and enables you to introduce additional clients and servers to the running system whenever you wish.

BeeGFS has been optimised for highly concurrent access. NFS and other simple remote file systems can potentially suffer serious performance problems in scenarios involving highly concurrent access. Beyond that, data can become corrupted as a number of clients write to the same shared file. Specifically designed for performance and robustness, BeeGFS addresses these issues, making it an ideal choice if your application involves highly concurrent access.

BeeGFS can run on any enterprise Linux distro and allows client and servers to run on the same machine, enabling high performance for smaller clusters. BeeGFS also does not require any dedicated file system partition on the servers - taking advantage of existing partitions, formatted with any one of the standard Linux file systems (for example ext4, zfs, xfs). For larger networks, it’s possible to create many distinct BeeGFS file system partitions with differing configurations.

Open Source

BeeGFS is open source and free to download and use with no limitations, making a Broadberry storage solution configured with BeeGFS extremely cost-effective.

There are a large number of advantages to this filesystem being open source. Open source software comes with less errors and bugs as the source code is visible to everyone, there is less chance of any developmental errors getting through.

Broadberry CyberStore Storage Appliances

The Broadberry range of CyberStore Storage Appliances features wide ranging compatibility with all major storage server software.

Many leading traditional storage vendors will try to lock you into their own ecosystem in order to increase their business through limiting your options. When you order a Broadberry CyberStore storage appliance, you never have to worry about vendor lock in, providing you much more freedom over your data centre and enabling you to make the additions you want on your CyberStore storage server.

In addition to no vendor lock-in, the Broadberry CyberStore storage appliance features complete compatibility with all major storage server software. Your CyberStore server can be configured just the way it needs to be to perfectly it the needs of your business.

This wide-ranging compatibility is a rarity as the majority of storage servers available from most vendors are only compatible with a single storage software operating system.