Answer: The 2-bay is specifically designed to accommodate the “broken mirror” concept for softwareless backup.  What we mean by softwareless is that the backup software and host machine is unaware that an additional copy of the data is being made.   Suppose we had 3 total swap drives (4 hard drives total) and left one drive in at all times as the “primary”. There are several issues we could discuss here:

1) Will the connectors on the back of the removable tray accomodate hundreds or even thousands of plug/unplug cycles?  The answer to that concern is yes.  We’ve been asked why we don’t expose the bare SATA drive and use that as the rear “docking plug” to save costs.  Those SATA connectors are spec’d at only 50 insertions by the committee. If you look at the type of connector we use, you’ll see it is a pin type connector with high insertion ratings.  While somewhat non-traditional, it is this connector that provides reliable daily connection.

2) The primary drive (the one left in place) gets high read activity every day.  We assume you will swap media every day causing a full remirror.  This requires reading every data block on the drive so that it can be written (mirrored) to the secondary drive.  It could be argued that this extra activity creates wear and tear on the hard drive during the daily full backup.  Will the primary drive fail more rapidly for this reason?  Well, we haven’t seen a failure correlation like this.  Our head engineer suggests that if this is a concern there is no inherent reason why you couldn’t rotate the swaps – rotate the right hand (or bottom) drive one day and after the mirror is sync’d rotate the left hand (or top).  The Automatic Mirroring Technology (AMT) doesn’t care and you could balance total read activity this way.  If it makes you feel better, by all means do it. But you will be “fixing” a problem that we’ve never seen happen.

3) The secondary drives (the one swapped each day) will have power removed and applied each day.  This power cycle load is spread out over the 3 swap drives (in this example).  The question is: are hard drives like light bulbs? – Do they often fail when power is applied?   Well, we’ve never seen a drive go “poof” when it was turned on – at least not that we attribute to a power influx. The raw drive has it’s own hot plug ability (Hot plug was added to the SATA II spec) and our trays do have protection circuitry.  We also mitigate this issue as best we can by requiring the key to be turned before the High-Rely classic media is removed.  This additional step provides even more protection.

4) Are there any anomalies (bugs) in the mirroring circuit that could cause corruption to occur after many swaps?  We aren’t aware of any.  It’s been in use this way since we first introduced it back in 2008 and have not seen issues with the re-mirroring process.  It does bring up an interesting point though.  We think it would be a good “best practice” to periodically run CHKDSK /F on your backup media (as well as your source drive).  This could be invoked as a scheduled job or done manually.  Scheduling CHKDSK is a bit scary in that it could actually create data loss or other problems.  If it were scheduled it’d be important to view the logs to see if problems were found and fixed (event viewer, Windows Logs, Applications).  We HAVE seen successful backups (images were created fine by shadowprotect and other programs) in which the source drive was later found to have corruption.  This corruption was merrily imaged onto the xxxx.spf file located on the High-Rely Classic removable disk media.  When the image was successfully restored, the host machine still wouldn’t boot because the original source boot partition was corrupted (and had been for over 30 days so all the removable drives were equally useless).  So that means for the prior 30 days the server was up and running, but it was sick and had anyone tried to reboot it, it wouldn’t have come up.

Clearly, it is reasonable to check for corruption on any drive periodically, whether or not AMT technology is in use.  I hope this helps.  We think Automatic Mirroring Technology is an awesome way to duplicate your backup!

What about using “multiple spindles” or RAID arrays to increase performance?  When measuring IOPS many benchmarks show that multiple drives help.  In fact, more drives equates to more speed.  Our testing shows that with modern hardware RAID, our RAIDFrames and FirstRAID products can  backup slightly faster than stand alone drives.  But the difference isn’t night and day because again, for large sequential writes the speed advantages of multiple drives in small random writes disappears when write operations aren’t as dependent on seek times.  Still,  using our RAID 5 backup products is a good idea for redundancy of your important data.  Plus you may gain 10 to 20% in backup and restore speeds with RAID. As for reliability, we believe slower spinning SATA drives have less heat and last longer than 10,000 and 15,000 RPM drives.  This is true even of so called “enterprise” SAS drives.  Studies of drive failures in large numbers by Google and others have proven that good old SATA drives have about the same failure rate as expensive enterprise drives.  Because the density and sequential writes make up for RPM, they can be about as fast for backup applications too.

Symantec thinks it owns the techniques for successfully doing drive imaging.  These include restoring an image to other hardware (foreign hardware retargeting), disaster recovery using virtual machines, storing disk images, backup catalogs, and using a certain type of GUI screen to perform a restore.  Two separate law suits were filed.  One against Veeam and an almost identical one against Acronis.  This suit clearly has implications for other vendors who do similar things including StorageCraft, VMWare , Microsoft, Computer Associates, Paragon, Double Take, Appasure, VRanger and many others.  This is unfortunate turn of events that could cast a pall on the backup industry.  It will be interesting to see if the courts find that the patents have merit.  If nothing else Symantec may force license fees on other vendors, increasing costs to the end user.  With backup licenses already running in the $600-$1200 range per server,  backup software can cost more than the Windows OS license!  Given the OS is over 50 million lines of code many of us hoped backup software vendors would be bringing prices down.  This lawsuit may continue to prop up those prices for a while.

A short story with links to the .pdf complaints can be found here: http://www.storagenewsletter.com/news/business/symantec-patent-lawsuits-acronis-veeam

Backup Virtual Machines

Computer Restoration Methods

Method of providing Replication

Selective File and folder snapshot creation

Here is the 2 Bay AMT premier shoebox size direct attached storage unit (also available in rackmount).

2 Bay Premier with Automatic Mirroring Technology

This is how it would work:  Plug a 2 bay unit like this (or 2 Bay RAIDFrame or FirstRAID AMT for more capacity) into an eSATA or USB3 port on the server.  Turn the internal mirroring on (we call it AMT for Automatic Mirroring Technology).  Now, the server will only see one drive (not two). Setup your image software to backup to the high-rely removable drive with it’s incremental and collapsing chain magic.  Since the external drive will never go away (you always leave the the main drive in place), the job can continue to run each night.  No struggles trying to get drives to re-appear after a swap.

The user swaps top drive every night.  Appasure, Symantec System Recovery, or ShadowProtect (whatever software) has no clue the drive has been swapped because it always sees a drive online (just like as if it were a NAS).  When the user puts in a new drive the AMT board starts a re-mirror.  It takes about 6-10 hours to mirror depending on whether you use 2TB or 3TB drive sets (Need more storage? The 2 Bay RAIDFrame AMT can backup and mirror up to 9TB).  An LED tells the user when the mirror is sync’d.  The mirrored drive can then be used for physical off-site transport.   Voila.

A low cost Slimline can be used to restore older archived trays without disturbing mirror setup.

Consider an extra $225 slimline removable drive system to simplify restores (You wouldn’t want to stuff a 6 month old drive with archival data into the bottom slot and accidentally over-write it when the mirror kicked in).  Not strictly necessary to use a separate device if you can remember to flip the mirroring switch off before attempting a restore from an older drive.  But….for the price it’s an easier way to avoid mistakes.

Direct Attached Storage or DAS refers to connecting storage directly to the server.  For Highly Reliable Systems our removable backup products use one of three types of DAS connection:

1. USB 2.0 (0.5 Gbps [480Mbps] data rate)   Backup speeds to SATA disk are up to 120 Gigabytes/Hr. 
2. eSATA (3.0 Gbps data rate) Backup speeds to SATA disk are up to 460 Gigabytes/Hr.
3. USB 3.0  (4.8 Gbps data rate) Backup speeds to SATA disk are up to 440 Gigabytes/Hr.

Note the USB 3 port also supports USB 2 connections. From the data rates above, you might expect USB 3.0 to achieve the highest backup speeds but as you can see, USB 3.0 can be slightly slower than eSATA despite the higher data rate of the cable.  This is because the hard drives we use are native SATA and a bridge chip must be used to convert to USB 3.0, meaning the higher data rate doesn’t always mean faster backups.  However, any speed difference is usually negligible and USB 3.0 controllers are often more compatible with servers that may already contain a storage controller.  For this reason USB 3.0 can be the preferred interface if eSATA conflicts are encountered.

Network Attached Storage or NAS devices are attached to the network (as opposed to directly to the server), usually with a 1 Gigabit Ethernet card.  NAS devices are usually not quite as fast as DAS devices for backup.  However, NAS devices are more intelligent, do not require a dedicated controller in the server being backed up, and typically have a “motherboard” with CPU and RAM running an operating system that shares the hard drive using a higher level network protocol. The numbers below assume a standard 1500 byte ethernet frame (Jumbo frames and bonding or teaming multiple NICs can sometimes improve speed).  

1. Gigabit Ethernet (1 Gbps data rate) and connected via a Windows Share – Backup speeds up to 180 Gigabytes/Hr.  Windows shares use the Server message Block or SMB protocol. 
2.  Gigabit Ethernet (1 Gbps data rate) and connected via iSCSI- Backup speeds up to 300 Gigabytes/Hr.  iSCSI uses a low level block protocol that bypasses the network security and thus achieves faster speed.  However in our iSCSI implementations only ONE host being backed up attaches to each removable drive in the NAS.  While this drive can then be “re-shared” for other servers, doing that creates a potential slow down.  
3. Gigabit Ethernet (1 Gbps data rate) and connected via NFS up to 240 Gigabytes/Hr.  NFS is a networking protocol like SMB that is typically used by Unix/Linux and VMWare systems.  It is easier to implement on these operating systems than SMB and has good speed.   Backup speed depends on many factors so your mileage will vary.

In the case of the WBA this means that if you have multiple Windows 7 machines that have the same files (for example in the \windows operating system folder) the WBA backs these files up only once.  So the backup drive(s) can be much smaller than the sum of the machines backed up. This functionality applies to the Microsoft workstation backup storage on the WBA, not for general NAS shares.  This appliance can could be used to backup both workstations and servers assuming you have enough space on the 2TB internal drive.  The way this is achieved is by using Microsoft built-in backup software  for the workstations and 3rd party software for the server backup. if you share a folder for server backup be aware that deduplication won’t apply to that data.

Windows Storage Server Essentials for fast Backup of Workstations

http://www.mspnews.com/msp/articles/257528-storagecraft-gives-managed-service-providers-new-pricing-option.htm

In addition to using block level imaging, products like Shadowprotect allow the  use of incremental backups that can be periodically rolled up into a single image file (often referred to as synthetic backup).  This overcomes the old concern of having to restore from multiple tapes or files and verifies data integrity in a single file.