I was recently visiting a large law firm in Canada and one of their critical pieces of infrastructure was a document repository. With decades of completed cases and supporting materials, the firm wanted to be sure that their intellectual property was protected consistently and reliably and also instantly accessible for use in future litigation. As you can imagine their repository has grown rapidly and consumed about 25TB of NAS storage.
One of the biggest challenges for the customer was backup windows. Their document management system contained 70 million small files and it took hours to protect. They wanted to enhance protection and allow for multiple backups during the day versus the current architecture which struggled to complete one.
NAS systems present unique challenges for data protection particularly when it involves small files. In this example, the law firm relied on a non-EMC protection solution that only allowed agent-based protection. Thus they installed a file backup agent in each server that accessed a NAS share and each server backed up its own network data as if it was a local drive.
The source of their lengthy backup windows was clear. When a job is initiated, a backup application agent scans every file to find those that need to be protected. Analyzing over 70 million small files was extremely slow especially when running over a network. To make matters worse for the law firm, the scan also negatively impacted both filer and production server performance. The result was extended backup windows and reduced application performance.
There is a better way to protect NAS information – NDMP. NDMP stands for Network Data Management Protocol and it enables a filer to backup directly to a protection target. NDMP backs up data more efficiently because it bypasses attached servers and leverages built-in filer functionality. Also because NDMP is an industry standard technology, it will work across vendor NAS platforms and so it is truly heterogeneous.
Implementing NDMP will typically shorten data protection windows, but there are still opportunities for improvement. NDMP backups are similar to traditional protection schemas and there are two common NDMP backup types:
- Level 0 – This is a full backup and so when a Level 0 is triggered, the NAS device will backup all data regardless of what is changed. This is the most lengthy, compute and bandwidth intensive process since it involves moving the largest amount of data. However, it also provides the fastest recovery since all data is contained within the backup.
- Level 1 – This is an NDMP backup that is similar to a traditional incremental. When you perform a Level 1 only changes since the last Level 1 or Level 0 are transferred which results in significant bandwidth and backup window savings. The trade-off is recovery time because restoring a Level 1 will typically require the recovery of the previous Level 0 and all Level 1’s since. The result is that Level 1’s are typically scheduled along with Level 0’s to manage recovery times.
It is clear that Level 0’s and Level 1’s provide strengths and weaknesses and the challenge is determining which level to use and when. An even better solution would be one where you could access a Level 0 equivalent regardless of whether previous backups were Level 0 or 1’s. In traditional backup parlance, this is called a “synthetic full backup.”
Bringing the conversation back to the law firm, our recommendation was that they re-architect their NAS backups. We suggested that they implement NDMP technology to accelerate backup performance and to also implement Avamar. Avamar offers a powerful feature called an “NDMP Backup Accelerator” which allows customers to backup either Level 0 or Level 1 NDMP streams and the accelerator will automatically remove redundancies (e.g. deduplicate) the data for space efficiency and it will automatically create a synthetic full backup. This technology would allow the customer to reduce backup windows by a factor of at least 2x or more by performing only Level 1 backups (after the initial Level 0) and thus enable them to meet their business SLA of multiple full backups during the day. Best of all, Avamar’s ability to create synthetic full backups would deliver restore times similar to a traditional Level 0 or full backup.
I found this meeting to be informative and a great example of how an efficient architecture can change the protection paradigm. As you look at your environment, I encourage you to ask yourself some critical questions:
- Are we using NDMP to protect your filers?
- How are we balancing the use of Level 0 vs Level 1 backups?
- Have we implemented synthetic full functionality to create Level 0’s from Level 1’s?
Answering these simple questions can make the difference meeting and missing critical business protection SLAs for your NAS data.