In the big launch today, EMC launched a whole array of new and groundbreaking things.
I am trying to break these up in a series of posts and explain them in a high-level fashion.
They all have roles to play in the EMC BRS stack:
The first post was a bit of explanation about the new DataDomain systems.
The turn has now come to Avamar, which was announced in it’s latest generation, Avamar 7.0
This is a major release and my objective here is to give a brief introduction to some of the many enhancements.
Avamar 7 brings forth a bunch of things, the first one I am going to touch on is the further integration that is happening with the Protection storage part of the stack, namely the DataDomain systems.
All data center workloads can now be targeted to a Data Domain System using EMC DataDomain DD Boost (except one, I’ll get back to that).
Previously file systems and NDMP workloads were exclusively sent to the Avamar DataStore, or as it is popularly called, “the grid”.
This means even more scalability for Avamar, unstructured content such as files certainly are high growth workloads and many are struggling with containing these.
An excellent strategy for limiting the growth of file backups is archiving, but that is a subject for a later post. 😉
New functionality for virtual environments
You will see a bunch of new and cool functionality for both VMware and Microsoft virtual environments.
On the VMware side, Avamar, when backups are sent to a DataDomain system now offers what is called Instant Access.
Instant Access means that the VM is presented from an NFS mount on the DataDomain and then is accessed there and is up and
running, without any “restore” having been performed.
The performance characteristics of a purpose built backup appliance is radically different than a production storage system, so expect that the VM will run with much fewer IOPS. The good news is that you can immediately start a Storage vMotion, which will move the running machine back to production storage.
Now, on this matter, I will go on a tangent, and describe a bit about how Avamar normally does restores, to show the different tools that a backup admin has at his or her disposal.
Avamar can perform an image backups with VMware Change Block Tracking (CBT) like most other backup software
variants out there.What that means is that at the time of backup, only the blocks changed since the last backup will be scanned, these blocks are then deduplicated against all other blocks Avamar tracks and those that are globally unique are then compressed, encrypted and sent to the protection storage.
Nothing unique here, most backup brands have this functionality today, some only from full backups, etc, Avamar does it every day, every backup is a full backup, but only globally unique data is sent.
Now, on to the really unique stuff. When it comes to restores, Avamar uses the CBT functionality again, querying the vSphere again about what has changed between the point of backup and now. Only the blocks that is needed to rebuild the point in time is then sent back.
Most changes that require a full image restore are trivial, so this means that restores are amazingly fast in most scenarios.
So what we see here is that there are a multitude of possibilities for making restores, from a single file to a whole VM.
Now, on the VMware backup/restore management side, Avamar 7 also has a whole host of new functionality, some of it really redefining stuff.
The first new functionality I will mention here is the vCenter interface integration.
The VMware admin usually feels most at home in the vSphere Client.
Avamar 7 provides a plugin here, so that there is no need for her or him to go learn another interface.
From here, the VMware admin just clicks EMC Backup and Recovery and can immediately configure backup policies, initiate restores or kick off a backup job.
Another new very useful functionality is what is referred to as “Dynamic backup policies”.
What that means is that backup policies can be applied to various different entities within the vSphere hierarchy, such as a cluster, a whole datacenter or what I guess will be most common, a group (folder) of VMs. Any VM that gets added to these entities will inherit the backup policy automatically to ensure new VMs are always protected.
For those out there that are using Microsoft Hyper-V, there are also really good news!
EMC is adding support for Hyper-V 2012 to the already existing Hyper-V 2008 support.
Support for using backup proxies in clustered shared volumes (CSV) has been added, meaning that one host can be pointed out to take care of the I/O generated by backups.
Further, in a Hyper-V 2008 R2 implementation, when a backup was started for one server in a Hyper-V over CSV cluster, all of the other Hyper-V servers in the cluster went into what is called I/O redirection mode. This meant that when the backup was running, all other Hyper-V servers had to send their I/O write requests to the Hyper-V server being backed up, slowing their write I/O performance. This re-direction now eliminated in Hyper-V 2012 backups with Avamar. Thanks for fixing that one Microsoft :).
Hyper-V 2012 got a lot better, and Avamar is right there with it.
NAS system / NDMP backups
Since many many years, Avamar has had really leading functionality for NDMP backups through the innovative use of an “NDMP accelerator”.
This means that after an initial full (level 0) backup, Avamar can continue to only request incremental (level 1) dumps from the filer daily, but still providing a full backup.
Avamar supports the backup of EMC Isilon, VNX and VNXe, Celerra and NetApp systems with this method, and Isilon is the new kid in here.
For now, Isilon has some restrictions, primarily because Isilon can get tremendously big, we want to make sure that the backup solutions are positioned correctly here.
Management of Avamar itself has also gotten an overhaul. A *ahem* less than loved part of Avamar management has been that there has been what is a tad non-poetically named “Blackout Window”, where no backups could be performed.
In Avamar 7, this restriction is removed, only leaving what is called a “maintenance window”, with fewer restrictions.
All deduplication systems need some time to take care of cleaning, expiration, integrity checks, etc. The good news is that the parts which required a read-only system previously now run in the background.
There are tons more to mention about this release, especially on the application integration side, with Microsoft Windows, Microsoft applications, Oracle, IBM DB2 and many many more, all getting major enhancements. I will save this for yet another blog post.
Watch this space!