Configure a backup using raid drives- NO tapes

Question

I am seeking advise on creating a backup strategy using the SCSI HD in a PE2900.   The configuration: (Which wll remain constant) PE 2900 with Perc5/i 2-73g 15k Raid1 for OS 6-146g 10k Raid5 for Data   In Flex bay are: (Which will vary) 2-300g 10k used for backup (drives A & B) additional 6 300g 10k drives for rotaional backup drives (drives C, D, E, F, G, H)   Goal: To do bakups to the 300 gig drives so they can be removed at regular intervals and replaced with the additional drives as needed. We would want to have a rotaional backup that would look like: Monday = Drive A (Full Backup) Tuesday = Drive B (Differential) Wednesday = Drive C (Replacing Drive A) (Differential) Thursday = Drive D (Replacing Drive B) (Full Backup) Friday = Drive E (Replacing Drive C) (Differential) Saturday = Drive F (Replacing Drive D) (Full Backup)   Monday2 = Drive A (Replacing Drive F which is removed from premises) Start rotation over again with Drives G & H used for rotating Quaterly backups and removed from premises when completed.   Questions: 1) Can this configuartion be achieved? 2) How do I configure the Raid controller so that each time a new drive is inserted, I don't have to reconfigure the controller NVRAM? 3) Would the raid controller see each drive as it is inserted? 4) Do I need to have each drive iniatialized on the raid controller before beginning the backup strategy?   Before I get any feedback about using tape drives, let me explain that the client does not want to have tape drives and deal with tapes at all. There have been issues with tapes in the past. Also, please do not metion USB drives as the speed issue comes up and doing a backup of 300 gig at the substained rate of USB will not provide a reliable bakup in an acceptable time period.   Any ideas on how to configure this on the raid controller?

Mark_A_Smith · Answer

Instead of answering this question I think that I need to point out something about hot-plug drives.  Hot-plug drives can be removed and inserted hot into the system.  This was done so that a failed drive can be easily replaced.  The life-span of the backplane in these systems is not infinite.  In theory you should get 100-200 plugs before the backplane will fail.  In reality it is probably more like 50-100.  This means that with your backup strategy your backplane will most likely fail inside of 3 months.   If you want to backup to disk I would suggest that you look at external SATA storge enclosure.  The MD1000 would work fine for this with no need to constantly remove and replace disks.

DELL-GaryS · Answer

I concurr, please review the liturature on the RD1000 it is a cartrige based hard drive system that is engineered for the multiple inserts and removals your backup strategy calls for. here    Mark said MD1000, that is NOT what you need (14 drive external storage system, has the same drive carrier system as the server, hence multiple inserts and removals will cause failures).   Dell-Garys

Mark_A_Smith · Answer

Gary,   I was not inteding him to do the hot-plug thing in the MD1000.  Just put the drive in and leave them alone and do all of your network backups to the MD1000.   If you are looking to do offsite storage of backups then you will need 'removable' storage and not hot-plug storage.

LaDaraTech · Answer

I have looked at both of these products and concluded that the MD1000 does not allow for the off site coponet without the same issues as the basic server, just additional storage.   The RD1000 was reviewed ad found that the largest HD was 120g. This is not enough space. When I discussed buying the system and just replacing it with a larger SATA drive, I was informed that the drive is sealed in a ruggitised case and could not be replaced.   The sales rep did say they were coming out with a 160g drive sometine in the future, but still not enough space.   So based on what I was told at the time of purchasing the server, (after explaining in great detail the plan for use) I now have several thousand dollars worth of drives that are of no use unless I can get this to work.   I was not aware that the backplane had a limitation to the number of hot swaps that can be done. Thanks for that piece of information.

DELL-GaryS · Answer

Laura,     The 120GB is uncompressed, in a perfect world... you might get as much as 240 GB on it, and would be good enough for your incremental backups.     One problem with using H/W RAID controlled drives for removable storage is that your RAID controller is going to give you an audible alarm upon removal (it can be disabled, but then you will not have the alarm to warn of a real failure), and you will need to force the drive online when reinstalling it (Assuming RAID0 single drive).     As far as the high number of insertions and removals... make sure your customer has the right service warranty: 7x24 hr 4hr response, with a keep your hard drive coverage as well... so that you can send the drives to a data recovery service if needed, and have the replacement in hand the same day. Plan on replacing any drives that get dropped... complete care coverage may be needed in that case.      Plan on getting the RD1000 as soon as we offer the needed capacity.   Dell-GaryS

LaDaraTech · Answer

I will consider this option once the capacity is closer to my requirements. In the meantime, here is what will work after many test:   Set Flex Bay to Raid1 with (2) 300 gig drives. (Drives A & B) Create backup jobs for each day of Week and cofigure the backup software to back up each day  with the name of the day in the backup job; with the overwrite option.   Set complete backup on Monday - After completed remove one (1) drive from Raid (Drive B) Insert another drive in bay (Drive C) and wait until raid rebuilds (approx 2 hours) Set a Differential Backup on Tuesday and Wednesday leave drive in bay until completed on Wednesday, then remove Drive A and insert Drive D; wait for Raid to rebuild. Set complete backup on Thursday. When complete rermove Drive C. Insert Drive E and wait for raid to rebuild Do differential backup on Friday and Saturday. Remove Drive D on Monday and insert Drive A.   Repeat.   While this is not a perfect solution, and yes you will need to ensure you have a valid Gold Level 4 hr 7 x 24 support contract to cover possible hardware related issues arising from the constant removal of the drive; it does provide the solution that the client wants.   I can only hope that, soon, Dell gets on the bandwagon and creates a backup solution that does not involve tape drives; with the capacity,  walk away ability, and functionality  to work with their servers.   (Opinion Follows) Anyone who has ever had a tape drive fail and/or had to rebuild from a failed tape back up is not intested in sinking an ever increasing amount of money into a solution that does not provide a solid backup solution.   This after sinking $4000 in Tape Drive $1190 for 14 tapes for rotational daily backups ( to be replaced every 3-6 months) $1190 for replacement tapes for daily's (maybe x 2 or 4) $1020 for 12 tapes for monthly archives $425 for 5 tapes for Qtr and Yearly archives $7825.00 - total cost for year   I can buy a whole bunch of, more reliable, drives for that kind of money   The only way to know if each of your tapes are functional or not, is to do a restore occasionally and see if you can SUCCESSFULLY recover your entire tape. Then, and only then, can you be assured that they are working correctly.  This does not speak to the degration that happens over time from a celulose base tape aging or the fact that the archive from 2 years ago on a DDS4 tape that can not be used in a DLT drive.   Of course, each time you test the restore abilty of the tape you degrade it further, kind of a catch 22 if you ask me. I have thousands of dollars of tapes in storage that probally can never be used to recover data from,maily because we don't have the drives to read them anymore. These are not available or not practible to have installed.   The only way to keep them currently available is to restore to disk and back up on the 'Current' tape drive. This then modifies the archive date and renders it useless for snap shot time stamping.   All server manufacturers have the same issue to deal with and none are taking it to task yet. I can get a solution for about any workstation I want, but nothing for the servers.   How much longer do we wait? My $0.02 (End of Opinion)

pcmeiners · Answer

I was just curious to the number of removal/insertion cycles SCA connectors are rated for, from all search results the design is for 500 minimum cycles.

DELL-GaryS · Answer

Hi,     A little discussion of what a design life of 500 insertions and removals means for this customer.   A single SCA connector has a 50% chance of failure after 500 total insertion or removal actions. He will be performing 4 cycles a week on two SCA connectors. To determine his chance of not having a failure... 1-((1-(cycles experienced/design cycles*.5))*(1-(cycles experienced/design cycles*.5)))   In 125 weeks he'll have 500 insertion and removal cycles on his backplane, his total chance of a failure is 75%. He'll have a 35% chance of a failure in the first year. (56% and 19% respectively if the MTBF spec is a complete insertion and removal cycle rather than a single insertion or removal). In addition, remember the drives are also getting insertions and removals, and have SCA connectors, then there is the chance of dropping a drive...      As the total service life of a server is 5 years, I can tell you he will be needing his service contract several times, and letting him know this up front can save him a lot of headache if he prepares his client for this expectation.   Dell-GaryS

pcmeiners · Answer

I was just surprised the SCA connectors had such a low 500 cycle rate. Personally I would not plan on removing drives multiple times per week, just too much to expect of hardware. With multiple cycles per week, the backplane would not be an issue to me, as they can be replaced easily enough/economically , but he drive connector wear would be my issue.

Message Edited by pcmeiners on 03-05-2007 03:19 PM

PowerEdge HDD/SCSI/RAID

Was this post helpful?