Challenges for "Go" NAS backup

For many storage administrators, backup of network attached storage (NAS) devices can be a huge challenge because they have to transfer data over the network to the backup server in order to back up the NAS devices. This process requires that the backup agent be installed on the server side to control the NAS device, and this is not very efficient. However, there are some ways to simplify NAS backup, such as Network Data Management Protocol (NDMP) and quasi-continuous protection system (CDP), but are there some limitations to these technologies?

W. Curtis Preston is an executive editor of TechTarget and is also an independent backup domain expert. In this article, he discusses some of the most common NAS backup challenges, and answers the advantages and disadvantages of NDMP, as well as other technologies that can help simplify NAS backup.

　　Q: What are the most common challenges in NAS backup?

　　A : One challenge is that NAS backup actually uses a standard basic protocol, and when we talk about Nas, we're actually talking about NFS, server information interceptors (SMB), or CIFS. In contrast to NAS, a storage Area network (SAN), or a locally attached Das store, is another storage mode in which the file system runs locally on the server. Therefore, in this case, the backup software must access the files to be backed up through the corresponding Protocol, NFS, or SMB, and NFS and SMB are IP-based protocols. So, there is a possibility that the remote device, such as a professional NAS device, itself can meet the performance requirements, but the connection between the network is not good to say, may be directly connected, or use a high-performance switch, or a bad network, such as the very old network equipment. In short, in this case you have to access and back up the corresponding data through the network through the corresponding NAS protocol.

At the same time, it also means that you cannot achieve anything that is unique. For example, a challenge for traditional file systems is the notorious "million-file" problem, where millions of files exist in a single file system. When you make a backup at the filesystem level, you have to deal with millions of individual files, and the system will be very uncomfortable, and the backup and recovery process will become unusually slow. For example, I have seen a 20GB file system Recovery process took 72 hours, it took so long not because of the backup software problem, but because the file system must be to create millions of file nodes and other metadata, this process is very time-consuming. When your backup data is not stored on the NAS, we can do a kind of called "image backup", image backup, that is, direct backup disk or LUN, restore the time directly mount, copy files. Therefore, because of the LUN-level backup, so that the file system data backup speed has greatly improved, if you need to restore the entire file system, then you can directly restore the entire LUN, so that the file system files will automatically appear. In the "Million-document" environment, this approach is 100 times times faster than traditional practice.

However, for Nas, you cannot back up the data on the NAS using the image backup method described above, the only way is to use NFS or SMB to back up the file data. So, this is the core challenge of NAS backup, and the network performance of the NAS is uneven.

　　Q: Is ndmp a good choice for NAS backup? How does it work?

　　Answer: NDMP was introduced to the industry about 10 years ago. That's when NetApp started to become active, and NetApp NAS devices are the kind of equipment that you can back up with headaches. Then people would ask, "How exactly do you back it up?" ", the first answer is:" It is NFS, so just use the NFS backup on the line. ”。 However, many people do not want to back up their data through NFS at all. So, Ndmp was born, and in order to better backup NAS devices like NetApp, many backup software vendors are looking for a way to end up with NDMP.

Many people think of NDMP as a backup format or protocol, but NDMP is actually a management protocol that allows the backup server to communicate with the NAS device. With the NDMP protocol, the backup server has to do is put the tape into the tape drive, or use a virtual tape, or any storage medium you want to store the backup data on, and when ready, the backup software tells the NAS device to do the backup. Then, how the NAS device backs up the data to the media is completely the NAS device's own business.

Another challenge is the backup format, in which a variety of backup software in the market produces a backup set in its own format. For example, NetApp uses the dump format, others use the CIO format, others use the TAR format or enhanced tar, and so on. When using NDMP to connect to a backup server, it says: "I've put a tape here, and I want you to back up the data here." The NAS device then encapsulates some information in the NDMP protocol to tell the backup server: "Well, I've done the backup and the data is here." , the backup software can then access the data.

Backups can be done in three ways: the NAS device backs itself up to itself, backups between NAS devices and NAS devices are backed up to the server or vice versa. Backup to its own way, typically with a tape drive or tape library directly attached to the NAS device, and then the NAS device itself backs up the data to tape. In the case of NAS-to-NAS backup, one of the NAS is connected to a tape drive, and the other NAS transmits the data over the network to the NAS connected to the tape drive, which then writes the data to the tape storage. In the case of NAS-to-server backup, the server is connected to a tape drive, where the server is actually acting as a device in response to the NDMP request, the NAS transmits the data to the server, and the server then writes the data to the tape for saving.

NDMP is the protocol that provides such a control, and because it can implement itself to its own backup, that is, the backup software can put a tape in the correct location and then tell the NAS device to directly back up the data to this tape or virtual tape, so that the use of external network to transfer data. In this way, the efficiency of backups has improved a lot.

The more cool thing about NDMP is that it not only makes tape backups, but you can use it to view snapshots. You can perceive snapshots in the system in the backup software, and then you can choose to restore them.

In essence, NDMP allows the data on the NAS to be backed up directly to a locally attached tape or virtual tape, which is of interest to many people, and improves the speed and efficiency of backup and recovery.

　　Q: What are the limitations of NDMP?

　　Answer: NDMP does not specify a backup format, which may be a major limitation of NDMP. For example, if you have a NetApp device and an EMC Celerra device, and you want to move all of your data to Celerra. In this scenario, you cannot use NDMP, for example, if you want to back up the data on NetApp from the NDMP, and then back to Celerra, this is not possible. This scenario is not limited to NetApp and Celerra, such as Celerra and Hitachi Data System (HDS), and so on. Different vendors have different backup formats, and each format is incompatible.

Another constraint scenario for NDMP is the legacy of the vendor's own format upgrades, such as EMC's decision to not back up with the dump format after two years, and to use some other new, better-developed backup format, so they are later upgraded. What about the tapes that were backed up before? There will be concerns that manufacturers have to consider back-to-side compatibility, and once the vendor makes a slight mistake, you must keep the old equipment in place to cope with any data recovery that may occur at any time.

Another limitation is related to file-level protocols. Common backup formats such as dump, TAR, CIO, and so on, are file-level protocols. And if you run into a "million file" issue, you will also encounter this problem in the NDMP environment, which is why many backup software vendors provide image backup. Many vendors support image-level backup, when you initiate an NDMP backup, you say: "Back up this volume to this tape, do not use dump, use this new image format." "When you do this, you get an image, and when you recover, you have to restore to an equal or larger storage space, and you can't do file-level recovery at this point."

　　Q: What other technologies can simplify backup of your NAS?

　　A : I am a loyal fan of quasi-CDP technology, which is essentially a combination of snapshots and data replication technologies. If the product you are using is a combination of snapshot and replication technology, you should use this approach when backing up, because this is based on a true incremental backup, and you don't need to do a full backup. The copied data looks like multiple full backups, but in fact each transmission is only a part of the transfer. And, the data you have is not just the data in the current system, you also have the system data images of yesterday, the day before and even last week. In addition, the copied data does not exist in a backup set format, which means that when you want to recover data, you do not need to do like traditional tape backup of the data back-up process, you can directly let your application directly suspend the copied data, while the data back to your primary storage, Once the primary storage data has been returned, you can choose to continue to use secondary storage for your app, or choose to cut the app back into primary storage, and reverse synchronize the changed data on the secondary storage to the primary storage. For example, if the primary storage power failure is only, in this case, the app switches to the secondary store to continue running, and when the primary storage power is restored, the changed data on the secondary storage is synchronized to the primary storage.

So, if you have a NAS device, then I highly recommend that you investigate exactly which is the best way to backup. If your device doesn't support some kind of backup, or you might consider choosing another NAS product. In my opinion, the quasi-CDP approach is the best way to backup.

Challenges for "Go" NAS backup