Data Recovery and Disk Utilities Forum @R-TT

Micropter:

Thanks for the reply, the most helpful and insightful one yet! Unfortunately, it doesn't sound as though this problem will be fixed any time soon. For now I've just accepted it, luckily with the features of VSS it's no problem to make backups during the business day. Even at the current rate, I'm starting my backups at 7pm and they are finishing at 6am, so it hasn't crossed into business hours anyway. But, this problem should still be addressed as its impact on my business (and others) will only increase. And yes, I prefer RAID 1 and 10 over 5 as well, and would never implement a software RAID. Unfortunately, I wasn't the one who configured these servers.

Todd

This problem is generic with R-Drive, meaning you are not the only one suffering from it. My guess is that it relates to how R-Drive interprets dynamic disks if they are software raided. I have currently been evaluating R-Drive on some of my clients 2008 Servers some have hardware raid and some use Dynamic disks for software raid. However, i did not experience the extreme amount of time you are referring to, but these are pretty powerful machines (dual quadcores HP ML350 with 20GB ram). But yes, it takes considerably longer time to perform then on non Dynamic disks.

On one of the testservers we have a Exchange database on a separate RAID-1 setup (using soft raid-1, dynamic disks). R-Drive sees both disks in the RAID but you cant image one or 2 of them, you can only select the Logical volume in the RAID (this however makes sense, R-Drive probably dont know if its a RAID-0,1,or 5 software raid and imaging single disks in anything but a RAID-1 would seem rather idiotic). The raid array is 250GB, eg it consists of 2 250GB dynamic disks. If you select the option to do a full sector to sector image it appears as R-Drive will actually image both drives even they are identical. Why? Well, the backup archive becomes very big. The ecxhange database is only 17gb big. The resulting archive becomes around 70GB with max compression. Ofcourse, its no wonder it gets bigger since its a full forensic capable sector-2-sector copy. However, a full sector copy of the same raid drive using Paragon Disk Image Server results in an archive that is only 20GB. The difference is, in Paragon you can select any of the dynamic disks individually or both. If i select both, the archive grows 100% which seems logical but the restored (and archived data) is actually the same in both cases. Running R-Drive imaging on the same setup choosing only to backup sectors that has any relevant Data in it results in an archive around 14-15gb with max compression. Im not saying Paragon is a better choice here, R-Drive has several other benefits over Paragon which has made me to choose R-Drive in favour for the production stage i just mention it as a reference. A whole day for a 160GB drive sounds like alot though, but it could be the lack of VSS im not sure, i have had some extremely long backuptimes with Acronis before on rather small , but live systems. All in all, R-Drive is rather fast, and certainly amongst the fastest when using maximum compression in my experience. A hardware raid-10 of 300GB takes R-Drive around 2 hours to finnish on a Dual Quad Xeon w 20GB ram. Image is sent over Gigabit network. It takes the closest competitor almost 4 hours to finnish the same task that i have tested.

However, yes, there is something to look into when it comes to Dynamic disks. Or , i would say probably not the dynamic disks actually but i suspect the "problem" lies in how R-Drive images software-raids or how it interprets the array. For a Raid-5 you have a minimum of 3 disks also, i dont use Raid-5 at all at any of our clients setups as i have found Raid10 far more superior in that it doesnt take any cpu calculation and it has almost no performance hit at all in the case of a disk faulting. Moreover, you dont have to sacrifice performance either since u get both the benefits from the raid-0 and simple mirroring. Rebuilding takes no calculation hits on the cpu either since its just a disk to disk mirror. Ofcourse this is my personal experience and opinion and im sure many would beg to differ. We all have our "personal" pros and cons. But the fact that your software raid is Raid-5 could also relate to the extreme amount of time it takes. The complexity behind a Raid-5 is far more complicated then a RAID-1 or RAID-10 and i have many times experienced that imaging RAID-5 takes considerably longer time, however im still not suggesting a whole day would be normal, but it could depend on the HW. What about the sizes of your final archives, and ofcourse, do you backup only the useful data or do you perform a full sector by sector image?


As for Virtual server. I have imaged disks containing live running systems for Hyper-V. Which also use VHD disks. They will work fine when restored, only that u will get the message from the VM guest why the computer was shutdown unexpectedly when you first boot it again. This is perfectly normal. So im pretty sure there should be no problem imaging MS Virtual server either. You can also image directly from within the guest VM. But that also means you need to purchase 1 license per VM. Considering the price of R-Drive is very affordable this could also be a recommended option.

Hope this sheds some help

I've been using a domain account that is a member of "Domain Admins" since the beginning, so that's the equivalent of running as a local administrator since the domain admins group is a member of the local administrators group.

Todd

Can you perform one test?
Create a task, and select the Administrator user on the User?Password panel (instead of Local system).
Will that speed the process up?

Any update on this? I've been trying all sorts of different tests by varying the settings of the backup, different compression levels, sector-by-sector vs actual data, changing the priority of the R-Drive process once it's started. I didn't notice too much variation when doing actual data backups, which is what I've always done. Finding the balance of compression seemed to have an effect, but what I was using seems to be the best, about level 4 compression. What bothers me is that R-Drive doesn't seem to be able to use all of my available bandwidth on this gigabit network. Like I said, I can transfer files between the servers at around 30 megabytes per second, but R-Drive seems to only be able to transfer at 4 megabytes per second. Is it because r-drive is pretty much consuming all the processing power of the server just creating the backup and leaving nothing for the OS to build the packets? These are dual core, dual processor servers, so I should think I'd have enough processing power. And I use the offload engines on my network cards.

Thanks,
Todd

Yep, I've run bandwidth tests between the two servers and all looks normal. I was able to transfer a single 1.05 GB file in 30 seconds from the server being backed up to the server storing the backup. So that's a transfer rate of about 36 megabytes per second, which is really the fastest speed you can expect to get from gigabit ethernet. And these are real-life hard drive tests, not memory tests. I also stagger my backups, so there is only one server writing to the hard drive at a time to avoid thrashing the backup drive around. The end image size is about 135 gigabytes, so by that math it should be able to transfer that file in about an hour. But in reality it's taking in the 7-8 hour range. Something is acting as the bottle neck, and it seems to be R-Drive.

Thanks,
Todd

I assume you checked that the network from this server to the network backup server connection works normally? Because we never experienced anything like this.
We will try to figure out what the problem might be and post our suggestions here.

If anything it took a little longer, 8 hours to backup 160 GBs using the R-Drive snapshot provider. This backup is saving to a SATA hard drive on another server over a gigabit network. Both servers are on the same Cisco 3550-12T gigabit switch with their NIC and switch interfaces manually set to 1000 Mbit, full duplex. I have other servers backing up similar amounts of data in the same fashion, saving to the same server, which take considerably less. The only difference is the dynamic disk.

Todd

Oh, woops, misread your question. All of my servers other than the one Windows 2000 server are Windows 2003, so I use the VSS snapshot provider on them. I'll try using the R-Drive provider on the one with dynamic disks are report back.

Thanks,
Todd

But in Issue 2 you wrote that was Server 2003.