Page 1 of 2 12 LastLast
Results 1 to 40 of 69
  1. #1
    Join Date
    Aug 2004
    Posts
    1,461

    Do datacenters use new hard drives ?

    I bought and worked with a lot of computers in my live... But the hard drives I used did never die. Now when it comes to dedicated servers it has happened several times that right after receiving the welcome email the server broke down with a broken hard drive. I did some research and found that many DCs are using used hard drives. That means when you cancel a server they take the hard drive, format it and insert it into a new server rather than buying a new hard drive. My theory is that if a hard drive fails they format it and if that works they use it again in another server and that is the reason why hard drives in servers fail so often compared to hard drives in other high usage environments. For example a 60 GB hard drive may have been used for 3 or 4 years before it was plugged into your server... No surprise that this hard drive is likely to fail especially if it did fail before.

    Please post your experiences and opinion: Should DCs use new hard drives ? And do they use new hard drives ?

  2. #2
    Join Date
    Aug 2004
    Posts
    1,461
    Other hardware parts have been buggy, too. In years of working with PCs I never had a broken mainboard or RAM stick but with dedicated servers it happens all the time... Because they use old, used hardware. That is my impression and it is true for many of the large DCs.

  3. #3
    Join Date
    Apr 2002
    Location
    West Yorkshire
    Posts
    1,357
    Good question and I am not sure of the answer there. We now use our own hardware on the majority of our servers, so this isnt an issue for us except the few servers we rent. All is ok at the moment there.

    I am not sure if they are obligated to use new equipment or not as long as they provide the service and the spec of the server and match or exceed that.

    It would be nice to think that they put brand new hardware in the rack for you but in reality I am not sure if they should do that or shouldnt.

    I did purchase from a company once who discounted servers that were previously used and offered the same guarantee.
    -- Matthew

  4. #4
    Well, hardwares does cost the datacenter money so I believe they would inevitably recycle hardware as they can still be used to maximise profits and to prevent wastage. Most of the time the setup fees were unable to pay for the cost of the hardware and with the low prices, it will take several months or years to recoup the cost of the hardware.

    I am sure that as long as they still provide replacements for the hardwares when they fail, I do not really care if I am a second or third user of the hardware.

    Of course, after several years they will retire old hardware when they get obsolete or fails.

  5. #5
    Companies would generally replace faulty hardware (free or for a very low fee), you cant expect them to use new hardware each time, as that makes no sense, why get rid of perfectly good hardware ... its not "second hand" when it goes into your server, and it being used before alot makes no difference provided the person who had the server last wasnt whoring it excessively.

  6. #6
    That's a very broad question and it should be obvious different DCs have different policies but I would imagine that most DC's would have a general policy that falls under two categories:

    - Hardware is to be used for X amount of time before being replaced.

    - Hardware is to be used until broken.

    Usually, when renting a dedicated server I leave it up to the host as it's their job to keep the hardware running. I don't really care what their policy on handling hardware is so long as the end effect is that the system is stable and up. For example, when you use something like PayPal you don't generally worry about how new the hardware on the servers are or how old the hardware is but what you really do care is that the service is available and up.

    You're best bet is to go with a DC or dedicated server host who you feel does a good job of maintaining uptime. There's going to be many different ways this is handled so go with what you're comfortable with.

    There isn't one right answer to this.

    EDIT: I would be extremely surprised if a DC were to replace all hardware even if a few months or a year old just because a new customer signs up. That is a total and blatant waste of hardware if it is working fine.
    Last edited by Snipz; 05-17-2005 at 06:54 AM.

  7. #7
    Join Date
    Sep 2004
    Location
    Flint, Michigan
    Posts
    5,765
    I seriously hope you don't expect a datacenter to be putting in all brand new hardware except for the case whenever a new server is ordered. This would be such an amazing waste of money it's hard for one to even think about. CPUs, Motherboards, cables, hard disks, cd-rom drives, etc... all fail. You probably have not run into these problems as much because your machines are not being utalized like a server and you don't have thousnads of them plugged in and running non-stop In my experience with home PC's alone I've had two HD's fail, one stick of ram, one cpu, and one powersupply. The powersupply failure also resulted in the entire case getting incredibly hot and a handful of different IC's loosing their solder joint and sliding down the board.

    Everytime you re-install windows, or re-image your home test server, you wouldn't but all new hardware in it would you? Of course not, why would you waste the money?
    Mike from Zoodia.com
    Professional web design and development services.
    In need of a fresh hosting design? See what premade designs we have in stock!
    Web design tips, tricks, and more at MichaelPruitt.com

  8. #8
    So you pay $50 USD for a server and expect to get all brand new hardware. Is that what you are trying to say?

    It does not make sense at all, does it?

  9. #9
    Join Date
    Feb 2005
    Posts
    334
    You are comparing apples to oranges. The average computer user, even if they end up owning 5-10 PC's at any one time, still have nowhere near the hardware turnover of a DC. Comparing hardware failures based on "my PC's" is like saying "car accidents shouldnt happen, I've never been in one" Please think of the size of your sample.

    "Because they use old, used hardware. That is my impression and it is true for many of the large DCs."

    please provide examples. Theres nothing wrong with a 2.8 Xeon that a customer was provisioned, then cancelled for whatever reason a month later being re-used for a new customer. There would be something wrong with a customer who was supposed to be getting a 2.8 xeon getting a dual PIII 1ghz.

    Keep in mind if you keep your customers you end up having to buy new hardware to provision regardless. So the key is keeping your customers happy.

  10. #10
    Join Date
    Aug 2004
    Posts
    1,461
    Show me the DC where you pay 50 Dollar for a server one time and can use it forever. I pay around 3000 Dollar per server per year and I keep the servers for a long time. I may pay like 9000 Dollar for an average server in total and I really think they could give me new hardware for this amount of money. For example the hardware does in fact NOT work anymore !! Of all the servers I have ordered MORE THAN 50 % HAD FAULTY HARDWARE initially !!

  11. #11
    Join Date
    Aug 2004
    Posts
    1,461
    >Comparing hardware failures based on "my PC's" is like
    >saying "car accidents shouldnt happen, I've never been in one"

    If I have 5 home computers running 24/7 with heavy usage 24/7 and 5 servers in a DC that run 24/7 and of my 5 home computers none does ever have a problem but out of the 5 DC machines 3 have faulty hardware before I even start using them then something smells fishy, does it ??

  12. #12
    Join Date
    Sep 2004
    Location
    Flint, Michigan
    Posts
    5,765
    Originally posted by thomas.smith
    Show me the DC where you pay 50 Dollar for a server one time and can use it forever. I pay around 3000 Dollar per server per year and I keep the servers for a long time. I may pay like 9000 Dollar for an average server in total and I really think they could give me new hardware for this amount of money. For example the hardware does in fact NOT work anymore !! Of all the servers I have ordered MORE THAN 50 % HAD FAULTY HARDWARE initially !!
    Obviously he was referring to dedicated server providers that offer servers around the $50/month area. A $50/month server is not going to be the same hardware as your $250/month server. $9,000.00 a year means you're paying $750 a month for an average server. But on a capslock got stuck note:

    SINCE YOU HAVE SO MANY CLIENTS AND ARE MAKING SO MUCH MONEY, $250 A MONTH IS NOT A BIG DEAL TO YOU, SO MAYBE YOU SHOULD BE FINDING ANOTHER SOLUTION FOR YOUR SERVERS. I HAVE HAD MANY MACHINES SUB $100 A MONTH AND YET TO HAVE A HARDWARE FAILURE. I THINK YOU ARE EXAGERATING OR LYING THAT OVER HALF THE SERVERS YOU HAVE ORDERED HAVE HAD FAULTY HARDWARE INITIALLY.

    Now how about posting who this provider is that is that you are paying $250/month (or in reality according to you $750/month) yet half the machines they give you don't work initially.

    BTW what are you defining as initially? I find it hard for a server not to work initially due to a hardware failure, i doubt any datacenter would be selling you a server that they couldn't even boot up to a login screen.
    Mike from Zoodia.com
    Professional web design and development services.
    In need of a fresh hosting design? See what premade designs we have in stock!
    Web design tips, tricks, and more at MichaelPruitt.com

  13. #13
    If you're paying that much per server and you have more than a 50% failure rate initially then perhaps that DC is definitely not right for you. I know quite a few people who use dedicated servers in the large quantities and I honestly don't know anyone who has that high a failure rate on initial rental. Sounds like that DC has some serious problems. Best to move on. As I said find a DC you feel comfortable with and it certainly doesn't feel like you're comfortable with your current one.

    Better yet, if you are that concerned then it might be time to look into colocation and provide your own hardware.

    P.S. Personally I would of moved on if the first batch of servers I had experienced more than a 50% failure rate. You should have moved on much sooner.

  14. #14
    Join Date
    Sep 2004
    Location
    Flint, Michigan
    Posts
    5,765
    Originally posted by thomas.smith

    If I have 5 home computers running 24/7 with heavy usage 24/7 and 5 servers in a DC that run 24/7 and of my 5 home computers none does ever have a problem but out of the 5 DC machines 3 have faulty hardware before I even start using them then something smells fishy, does it ??
    All your threads smell fishy.
    Mike from Zoodia.com
    Professional web design and development services.
    In need of a fresh hosting design? See what premade designs we have in stock!
    Web design tips, tricks, and more at MichaelPruitt.com

  15. #15
    Remember that when you pay for a dedicated server, you are not just paying for the rental of the hardware but it would include the cost of the bandwidth, power supply, salary of tech support and every other overheads.

    I believe what samtam was saying is that the $50 paid may be the one-time payment as setup fees. Monthly rental would cover the cost of the bandwidth and day to day running of the datacenter and to ensure the server is up.

  16. #16
    Join Date
    Aug 2004
    Posts
    1,461
    I didn't say I am paying 9000 Dollar per year. Can`t you read ?? I said 3000 per year and 9000 in total meaning if I keep the server for three years.

  17. #17
    Join Date
    Sep 2004
    Location
    Flint, Michigan
    Posts
    5,765
    Originally posted by thomas.smith
    I didn't say I am paying 9000 Dollar per year. Can`t you read ?? I said 3000 per year and 9000 in total meaning if I keep the server for three years.
    I can read very well:

    Now how about posting who this provider is that is that you are paying $250/month (or in reality according to you $750/month) yet half the machines they give you don't work initially.
    3000 per year works out to $250 a month. Now one would assume when you say "in reality I pay $9000" that you would be including the licensing of all your extra software, support costs, administration costs, etc... You did not say $3000 per year and have them an average of 3 years.
    Mike from Zoodia.com
    Professional web design and development services.
    In need of a fresh hosting design? See what premade designs we have in stock!
    Web design tips, tricks, and more at MichaelPruitt.com

  18. #18
    Join Date
    Jul 2002
    Posts
    3,352

  19. #19
    Join Date
    Dec 2002
    Location
    Texas
    Posts
    424
    i'm sure they all recycle, most especially budget providers, this is why its important to have a backup plan


    "what if....."

    I've never had hardware fail on a dc server of mine, but my home computer has died a few times.

  20. #20
    Actually in the life of our business, I think we only 2 harddisk that really fail from what I can remember, and another couple that we proactively asked for a change (and was changed without question). (Fingers crossed it stays that way).

    We do have quite a number of servers, so I didn't think it's more probable, and we have been in business for years.
    Like us on Facebook to qualify for discounts!
    http://www.sprintserve.net
    Offering: | Internap FCP Bandwidth! | Rebootless Kernel Updates! | Magento Optimized Hosting | Wordpress Hosting |
    Services: | Managed Multiple Cores 64bit Servers | Server Management |

  21. #21
    Originally posted by thomas.smith
    I didn't say I am paying 9000 Dollar per year. Can`t you read ?? I said 3000 per year and 9000 in total meaning if I keep the server for three years.
    If you prepay 3 years in advance, I am sure they will be happy to give you a brand new set of hardware, or even 1 year in advance if you request. However on a month to month, I don't think so.
    Like us on Facebook to qualify for discounts!
    http://www.sprintserve.net
    Offering: | Internap FCP Bandwidth! | Rebootless Kernel Updates! | Magento Optimized Hosting | Wordpress Hosting |
    Services: | Managed Multiple Cores 64bit Servers | Server Management |

  22. #22
    Join Date
    Apr 2002
    Posts
    1,018
    Originally posted by 3rdcoast
    i'm sure they all recycle, most especially budget providers, this is why its important to have a backup plan


    "what if....."

    I've never had hardware fail on a dc server of mine, but my home computer has died a few times.
    I cant even imagine non-budget providers not recycling, it would be incredibly wasteful and not economically sound to just throw drives aside as a customer cancels and order a new one. I'm sure everyone recycles, and most will replace the drive completely when it fails. We RMA it for a new one when possible, when not we can always pull the circuit boards off in case a customer's circuit board fails and needs a new one so he can get his data, the rest can be scrapped out.

  23. #23
    Why the hardware in dc fail too much???

    I guess coz people dont own the server so they dont care if they rip them off badly. Sorry to sound like an a.. here but i never care how badly I rip the hdd of my dedicated server. It is an IDE and I have around 150 people downloading it constantly. Even the downloaders on slow connections but they r downloading different files therefor your hdd's head still have to move back and forth to provide the data .
    But in my home pc, while iam unraring a file I will never do a copy command at the same time...why not...coz it is pain in the a.. to listen to my hdd screaming even it is a sata.

    But i think the point of the OP is that Dedicated provider buy cheap made or even refurnished, old stuff to build the server...am i right?

  24. #24
    Originally posted by thomas.smith
    Show me the DC where you pay 50 Dollar for a server one time and can use it forever. I pay around 3000 Dollar per server per year and I keep the servers for a long time. I may pay like 9000 Dollar for an average server in total and I really think they could give me new hardware for this amount of money. For example the hardware does in fact NOT work anymore !! Of all the servers I have ordered MORE THAN 50 % HAD FAULTY HARDWARE initially !!
    get a good company then, one that tests hardware before putting it in.

  25. #25
    Originally posted by BSD2
    get a good company then, one that tests hardware before putting it in.
    Iam not sure that there is any DC which test their hardware before putting in =/ except those ECC things.

  26. #26
    Join Date
    Mar 2003
    Location
    Charlotte, NC
    Posts
    2,760
    Originally posted by thomas.smith
    If I have 5 home computers running 24/7 with heavy usage 24/7 and 5 servers in a DC that run 24/7 and of my 5 home computers none does ever have a problem but out of the 5 DC machines 3 have faulty hardware before I even start using them then something smells fishy, does it ??
    That means that the problem doesn't lie with the fact that the hardware is used - Anyone buying large batches of hardware is bound to have a few defective drives after even a few months of use (even if using good quality drives, things can go wrong). Out of the 10 harddrives we have in use in the office, only 1 drive (a Seagate IDE) has ever failed, but have another drive from the exact same batch that's still operational. We have 2 old Maxtor refurbished harddrives that still run perfectly... Who knows why a normal drive would fail before a refurbished one?

  27. #27
    Join Date
    Aug 2004
    Posts
    1,461
    I doubt they check it. I think in some cases if someone reports broken hardware that still works but is just faulty in some way they just replace it and build the old hardware into another server thinking the customer may be wrong and the hardware may be ok. At least that is the only way how I can explain the large amount of broken hardware I have seen in various DCs.

  28. #28
    Join Date
    Sep 2004
    Location
    Flint, Michigan
    Posts
    5,765
    Originally posted by thomas.smith
    I doubt they check it. I think in some cases if someone reports broken hardware that still works but is just faulty in some way they just replace it and build the old hardware into another server thinking the customer may be wrong and the hardware may be ok. At least that is the only way how I can explain the large amount of broken hardware I have seen in various DCs.
    When a dedicated server goes down the client does not know what piece of hardware failed. They submit a ticket/call the datacenter and they investigate it. If there is a faulty piece of hardware it's determined by the datacenter, not by the client.

    Let's hear a few examples of the large amount of broken hardware you have experienced and the datacenters please.
    Mike from Zoodia.com
    Professional web design and development services.
    In need of a fresh hosting design? See what premade designs we have in stock!
    Web design tips, tricks, and more at MichaelPruitt.com

  29. #29
    Join Date
    Mar 2003
    Location
    Charlotte, NC
    Posts
    2,760
    Originally posted by thomas.smith
    I doubt they check it. I think in some cases if someone reports broken hardware that still works but is just faulty in some way they just replace it and build the old hardware into another server thinking the customer may be wrong and the hardware may be ok. At least that is the only way how I can explain the large amount of broken hardware I have seen in various DCs.
    I know that at BurstNET, faulty hardware is tested, and then removed from the server. The dead hardware is then stored in a "broken hardware" closet and then returned if it's under warantee. Why put a faulty piece of hardware back in a server, when it'll only cause you to spend more labor time on removing a drive, adding a new one, transferring data, and possibly issuing an SLA credit?

  30. #30
    Originally posted by square-head
    Iam not sure that there is any DC which test their hardware before putting in =/ except those ECC things.
    I've had TSN, and they are quite pants, test thier hardware before building a server

    If you ask them many hosts would probably do it.

  31. #31
    Originally posted by BSD2
    I've had TSN, and they are quite pants, test thier hardware before building a server

    If you ask them many hosts would probably do it.
    Well at one end you want it fast, on the other you want it tested I am sure if you are willing to wait a week or 2 from ordering, many DC can run the tests for you.
    Like us on Facebook to qualify for discounts!
    http://www.sprintserve.net
    Offering: | Internap FCP Bandwidth! | Rebootless Kernel Updates! | Magento Optimized Hosting | Wordpress Hosting |
    Services: | Managed Multiple Cores 64bit Servers | Server Management |

  32. #32
    Well, the only hardware I've ever asked to be tested is HDD + RAM, as I trust the mobby + CPU to work, equally HDD failures are the most annoying ones, took them an extra 30 odd hours on setup, but I felt better after it wasd one.

  33. #33
    Join Date
    Aug 2001
    Location
    Canada
    Posts
    2,123
    Originally posted by thomas.smith
    If I have 5 home computers running 24/7 with heavy usage 24/7 and 5 servers in a DC that run 24/7 and of my 5 home computers none does ever have a problem but out of the 5 DC machines 3 have faulty hardware before I even start using them then something smells fishy, does it ??
    However, I'm assuming those are not servers in 1U boxes under constant load.... heat will kill components and even with forced cooling it's still an issue I'd say.

    Common hardware issues I've seen are CPU fan failures and harddrive failures (minor ones though but enough to get the drive replaced as it will go completely bad).

    Even with brand new hardware (that we bought) we had 3 drives in a batch that were bad/went bad in a week. Assuming the larger datacenters buy in a batches I can totally understand they'd get a bad batch and go through them all. That's a problem with the harddrive manufacturer I'd say, not the noc.
    www.idologic.com - Reseller, VPS and dedicated hosting - Friendly Customer Service - DirectAdmin - cPanel - InterWorx

  34. #34
    Join Date
    Dec 2001
    Location
    Toronto, Ontario, Canada
    Posts
    5,954
    MTBF is the measurement that most manufacturers use to determine the reliability of a part, it stands for "Mean Time Between Failures" (*not* Mean Time Before Failure). Given this, most parts are actually *more* likely to fail out of the box, thent hey are to fail 6 months down the road (once you get into the years, the story changes granted).

    Parts out of the box fail more often as they've been travelling (shipping, getting rattled around, etc.). Parts in a 6 month old box have assumedly passed the initial setup tests, etc.

    If you want new parts, go with a very new server, very few providers (that I know of at least) would go build out a new lineup of dual xeon's, then load them with old HDD's and RAM, etc. However if you go buy a older Celeron, well what do you expect? The cheap prices are generally subsidized by the fact that previous customers have already had them for several years, covering the majority of hardware related costs (less upkeep).

    Needless to say, when you go to a rental agency, you dont expect to receive a new car, right? When you rent a u-haul, you dont expect the oil to have been freshly changed for your use right? When you rent a modem with your Cable/DSL company, you dont expect the modem to be brand new, right? If you do, theres probably something wrong (with you), as none of those are viable things, as much as they may happen to prove more convinent for you.

    Needless to say, I've yet to ever see someone throw out equipment when moving Colocation hosts (which is essentially the same proposal, especially since the server is physically being moved, typically long distances, from one physical environment to another). If hardware is that much of a concern, definatly build it yourself. There are minor tweaks you can always do, we have several dual opterons that can loose half of the fans in the chassis for example and continue to run (including CPU fans) thanks to absurd airflow levels.

  35. #35
    Join Date
    Nov 2003
    Posts
    1,093
    You say you pay $3000 a year, but when that's only $250 a month and the provider has no guarantee on the length of your coinbtract, why should they buy an entirely new system for you? Maybe i'm not understanding you correctly, but you're expecting that because your server is $250/month you should get a brand new server, yet the price of the server is probably 3* this when hardware is bought and if you cancel the server after 1 month, what are they supposed to do, get rid of it?

    If what you're asking is, can providers do this when you pay for a year in advance, the answer would be yes. Otherwise, it would be pretty dumb to give every custiomer a brand new server, especially with low-end servers since churn is higher...
    ManageMyServices was sold by me in September 2009. I no longer have any affiliation with this company.

  36. #36
    Join Date
    Jan 2002
    Location
    UK
    Posts
    1,034
    Well I think this topic is interesting.

    Obviously its not viable to buy brand new hardware for every new order but do all datacentres throw hardware away when it fails or do they get tempted to reuse it simply because it passes a test.
    Chris Collins
    Hostingfreak.net
    Directadmin Hosting in europe
    www.hostingfreak.net

  37. #37
    Join Date
    Sep 2004
    Location
    Flint, Michigan
    Posts
    5,765
    Originally posted by Chrysalis
    Well I think this topic is interesting.

    Obviously its not viable to buy brand new hardware for every new order but do all datacentres throw hardware away when it fails or do they get tempted to reuse it simply because it passes a test.
    Typically when a piece of hardware fails it fails and does not come back to life. When a CPU is cooked, ram is dead, or a hard-drive is toast they are unusable and get put in the bad parts bin as mentioned above. I would guess in a few very rare cases they might be reused, but I would not bet on it.
    Mike from Zoodia.com
    Professional web design and development services.
    In need of a fresh hosting design? See what premade designs we have in stock!
    Web design tips, tricks, and more at MichaelPruitt.com

  38. #38
    Join Date
    Dec 2001
    Location
    Toronto, Ontario, Canada
    Posts
    5,954
    Originally posted by Chrysalis
    Well I think this topic is interesting.

    Obviously its not viable to buy brand new hardware for every new order but do all datacentres throw hardware away when it fails or do they get tempted to reuse it simply because it passes a test.
    If its failed, its not going to pass the typical tests (eg. a few hours of memtest86 for memory, or a hdd scanner supplied by WD/Maxtor, etc.).

    I think most places will RMA their equipment, very few would bother attempting to re-use broken hardware (while a HDD may be worth around $100 persay, the amount of man hours dealing with a customer who receives the broken hardware will quickly outprice the potential savings of not buying replacement hardware).

    Needless to say, many brands of memory now come with "lifetime" warranties, Seagate drives now include 5 year warranties, most brand name servers include 3+ year warranties, etc. I'd expect that when a part breaks, it simply gets tossed into the "RMA ME" pile, thats what happens here.

  39. #39
    Join Date
    Jan 2003
    Location
    Chicago, IL
    Posts
    6,889
    What sense does it make to reuse faulty hardware?? It would simply result in needing to replace it again, costing more time and a happy customer. The risk is not worth the reward, so to me it makes no sense that any DC would recycle faulty hardware.
    Karl Zimmerman - Steadfast: Managed Dedicated Servers and Premium Colocation
    karl @ steadfast.net - Sales/Support: 312-602-2689
    Cloud Hosting, Managed Dedicated Servers, Chicago Colocation, and New Jersey Colocation
    Now Open in New Jersey! - Contact us for New Jersey colocation or dedicated servers

  40. #40
    Join Date
    Feb 2004
    Location
    Fort Worth, TX
    Posts
    2,585
    I am sure most datacenters at one time or another use old hardware. It would be a waste to order new hardware each time they sell a server. Think of it this way, expect the server you will receive to have already been used several times..

    I have had 3 hard drive failures in my personal computers in the last 14 years. I have had only 1 hard drive failure in web hosting in the last 8 years.. unfortunately just recently. So this isn't a big number.
    www.JGRoboMarketing.com / We Filter out the Bad Leads and Send you the Good ones!
    █ Office: (800) 959-0182 / Automated Lead Funnel Service

Page 1 of 2 12 LastLast

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •