Veeam reduces backup times substantially by utilizing Linux XFS reflink. dsmISI/OneFS supports XFS reflink on Scale-out-NAS storage and thus enables linear scalability while minimizing management effort.
dsmISI/OneFS fulfils all requirements for a scalable and easily manageable, economical back-end and is therefore the ideal target for Veeam Backups. DellEMC OneFS is a single file system which scales capacity and throughput linearly by adding nodes.
dsmISI stripes all Veeam backup streams across all nodes of a cluster while OneFS constantly balances capacity. In theory, a single stream of data can therefore be as fast as the sum of all nodes, and OneFS is designed to accept many streams sharing this performance when reading and writing..
dsmISI now supports XFS reflink for synthetic processes which would not be possible on any NAS storage without dsmISI. Compared to block storage, this drastically simplifies deployment and automation
Capacity can be shared dynamically and with a high degree of flexibility. Every XFS file system can be mounted or unmounted to/from any Linux repository server with a single command if a server failure or maintenance downtime should require moving a repository.
By utilizing XFS reflink, random-heavy workloads no longer occur during backup and processes run almost entirely with sequential storage access patterns, which allows utilizing high capacity, economical OneFS nodes.
No longer having to run full backups together with compression and deduplication minimizes capacity requirements.
Using an Dell EMC Isilon Scale-Out NAS cluster with any backup-to-disk solution has obvious benefits such as seamless scalability of the target file system, scalability of performance and availability due to an active-active architecture. However, there are some drawbacks to consider. First of all, the fact that an Isilon’s file system is spanned across all nodes of a cluster make that file system as big as you like (simply by adding nodes to the cluster) – but scaling throughput along with capacity is basically limited by the number of machines accessing an Isilon and the distribution of their data paths across these nodes. By default, any NFS or CIFS transfer between any host and an Isilon is limited to the communication bandwidth of the node selected to be the network target in the cluster and the server initiating the transfer. This leads to reasonably balanced workloads when many servers communicate with relatively few Isilon nodes. But backup workloads are characterized by relatively few servers which have to transfer lots of data in a limited number of streams to a potentially large number of nodes.
Without dsmISI, each of these servers’ throughput would be limited by the bandwidth of the node it happens to communicate with at any given time, which would limit any backup workload’s bandwidth to a mere fraction of an Isilon’s performance capabilities. dsmISI for TSM eliminates that bottleneck by equally spreading TSM workloads across all of an Isilon’s nodes, hence using all the available bandwidth all the time – no matter whether there is just one or many TSM servers using an Isilon as a backup target. dsmISI for Veeam uses the same basic technology to utilize Scale-Out NAS targets for Veeam Availability Suite – but with some additional functionality in order to accommodate the specific challenges Veeam poses to backup target storage.
In contrast to TSM, any Veeam backup job writes data in a single stream which therefore potentially has to satisfy much higher per-stream throughputs than a typical TSM workload which can usually be parallelized almost at will. Another Veeam specific challenge is the size of a target file, which cannot be configured but corresponds directly to the entire size of a backup job or virtual machine. Since virtual machines’ capacities can easily outgrow the file-size limitations of an Isilon Scale-Out NAS cluster (4 TB), natively using an Isilon can hardly be considered a useful solution for Veeam backup. dsmISI for Veeam addresses these challenges by fully virtualizing NAS storage into a standard block-type file system which not only allows practically unlimited file sizes, but uses the entire bandwidth of an Isilon cluster – even for a single data stream.
In summary, EMC’s Isilon Scale-Out NAS in combination with dsmISI for Veeam is the ideal backup target for rapidly growing virtual computing environments because it allows users to quickly and transparently scale to any performance or capacity requirement with unprecedented ease.
Whitepaper: Spectrum Protect and Veeam Availability Suite usage guide abstract, December 2016
Veeam and Spectrum Protect customers often need to store petabytes of backup data. For this, they need an efficient and powerful system that load balances the data stream across multi-target storage controllers, without the negative effects of deduplication storage systems. An example of a combined target disk storage system for Spectrum Protect and Veeam is General Storage’s dsmISI for Spectrum Protect and dsmISI for Veeam. An EMC Isilon Storage Systems is used with the dsmISI load balancer software to get the maximum out of the EMC file cluster nodes. With Veeam, a single target write stream is spread across the nodes for high throughput of more than 1 GBs per stream. General Storage MAGS can help bring Veeam Backup files to TSM tape and organize them.
Technical brief by Andreas Neufert, Solutions Architect, Veeam Software GmbH
General Storage Software GmbH
+49 6206 519 1794