For the last couple of weeks I’ve been doing a couple of web-ex sessions with the 3rd party “virtual machine” only backup vendors – esXpress from PHD Technologies, Veeam Backup & Replication and VizionCore’s vRanger. I’ve yet to do the webex with VizionCore just yet – as everyone is pretty busy (including myself) preparing for this year's VMworld 2009 in SanFran.
My plan after completing the webex phase is to download and use each of the big 3 vendors' technologies and report back my findings. But I thought I would begin with a “heads up” approach first, with quick webex sessions where I could see the various products demonstrated, and ask a few pertinent questions along the way. The kind of thing I had in mind was seeing how mature the various vendors' technologies had become, what their future directions might be, and generally to figure out where are we in the world of VM backup. I was also kind of inspired to do this research because of vSphere’s own vDR appliance – and the question of what this development might mean for the 3rd party VM backup companies.
My first webex was with PHD Technologies and their esXpress product.
I was chatting to Mike Luca. He’s got a great job title “Sr Complexity Simplifier”. I kind of like that, as that’s something I think I have to do every time I teach a VMware course. I have to make what might seem complicated, seem simple. Often my students tell me how easy everything is, to which I always respond, tongue firmly in my cheek – “yes, that’s because you’ve got the best VMware instructor on the planet”. That always makes the guys laugh, hopefully without too much irony.
esXpress ships into formats – Professional & Enterprise. The nice thing about this is that the two versions have no features lost between professional and enterprise. They are just a way of bundling esXpress into different types of customers. I like when vendors don’t hobble or remove features simply because it can cause false SKU types. Instead the different SKU types merely reflect that businesses are of different sizes and requirements. It means you don’t end up buying an enterprise product for a small business just to get a particular feature you want. If only other software vendors took the same approach. So Professional can backup 4 VMs concurrently, whereas Enterprise can do 16 VMs concurrently. At the moment, the esXpress product does have “agent” that is installed to the ESX “Service Console”. But Mike told me that ESX4i support is on their roadmap. esXpress does offer 24/7 based in the US. English only.
Increasingly, the 3rd party vendors are offering de-duplication of backup data and the ability to back up the entire VM once and then merely backup block changes after that. esXpress is no different in that respect, offering 25:1 de-duplication ratio. I asked Mike whether he found customers concerned about the reliability of scalability and of de-duplication. Mike said most customers are happy with the benefits of de-duplication for archiving purposes and that it was more production data (de-duplication of live data) that worried customers. The advantages of de-dupe for backup seems to be such a compelling one, that no-one disputes its usage. My worry is if multiple backups reference the same data and that data turns out to be incorrectly backed up or corrupted, it does have knock on affect. However, it became clear that these algorithms have been in circulation for some time, and the verification by calculating long checksums is one that can be relied on. So that’s pretty much a standard. So I was keen to see what made esXpress different from a feature spec. What did stand out is how esXpress has its own method of triggering a backup on the fly without having to create a specific job for VM. Simply by including specific wildcards into the name of VM you can trigger a backup. For example:
[x0] in the VMs name indicates to exclude a VM from a backup
x22 indicates to backup a VM at 22:00hrs
xNOW] indicates to backup the VM immediately
[o1] [o2] indicates what order to backup VMs
Then we moved on to a product demo. All the vendors did this – and it allowed me to hit them with my top questions:
- If a snapshot is not removed from a VM after a backup. What systems are there in place to resolve that? Is it logged and am I alerted?
- Do you communicate by vCenter exclusively to manage the backup and restore the VM?
- How easy it is to restore an individual file from a backup set?
- Do you use Microsoft VSS calls?
- Do you support vSphere4?
- What do you think about VMware’s request to ask Veeam not to support the free ESXi release?
Restoring individual files is relatively easy in esXpress. Of course, where most 3rd party backup vendors shine is in complete backups and restores of a lost VM. But you must ask yourself how common is that? Except in extreme cases where a VM is riddled with a virus; an idiot powers off a VM and then chooses “Delete from disks” OR you have lost entire VMFS volume, and you don’t have array snapshots – then the most common reason to trigger a restore process is an individual file such as a word or excel document that has become bad and needs replacing with the last good backup. Historically, 3rd party VM only backup vendors have struggled with this. This is because there isn’t an easy way to open a VMDK file, navigate the disk contents, and then fire it back into the original VM. For the moment esXpress allows you to download the files you want to restore from a .zip file, and then copy this .zip file to the affected VM using Windows Shares – from which the .zip file can be extracted, and the files restored. It’s pretty neat in that respect – but it does mean some work for the operator at the end of the day.
As for vSphere4 support –my webex, PHD technologies have announced they fully support vSphere4. In fact, they have gone down the road of creating virtual appliances to assist in the backup and restore process. That sounds eerily familiar to VMware’s vDR, in fact some of my sources tell me – that it may be that idea of vDR actually came from PHD Technologies in the first place…
It does seem that this approach is the way forward for the 3rd party backup vendors to make file-restores easier. Basically, the concept is that the backup appliance can mount the VMDKs of the VM to start the backup process but also additionally, use this process to facilitate the restoration of files – with restore data being re-assembled from de-duplication files to present back to the VM a virtual disk. The next thing that’s needed is a simple wizard in the VM which can be used to select files and their restore location. Anyway, you can see more of this new esXpress virtual appliance below in this youtube video.
esXoress does support MS VSS. This is important because historically VMware’s own File Synch driver hasn’t always performed well for certain types of systems. This one will take more research on my part, because it’s my understanding that the VSS API is a particularly rich and complicated one – and therefore how well it is implemented is going to vary from vendor to vendor. So it wasn’t something I could clean from a web-ex.
Well, that’s it for this quick overview on my webex with PHD Technologies. I will be writing up my webex with Veeam next. And once that is done I will do the same for VizionCore. Of course, the real work will begin after VMworld, when I will begin downloading and evaluating the products for real.