Tag Info

Hot answers tagged

50

Your analysis is fine -- to a point -- in that it absolutely will make things faster. You still have to account for a couple of other issues though: Not everyone can afford enough memory; when you have multiple terabytes of data, you have to put it on disk some time. If you don't have much data, anything is fast enough. Write performance for your ...


27

I would ask, what is the value of the data on the disk? If it's more than the cost of a new disk, then my preference would be to destroy the faulty disk and buy a new one. You could spend a lot of time trying to get the disk working long enough that you could do a proper erase, but is it worth it? And do you know that it's definitely worked? What if there ...


14

You could ... degauss it. I mean, it's already broken ... You just need to find a significantly large electromagnet.


14

Just explain to your vendor that the disk failed but that you cannot return it due to sensitive data. Many vendors, especially vendors with business customers, will accept that without asking. If they don't, offer a signed, written statement declaring that the disk failed. If that still doesn't satisfy them, write off the loss and consider buying the next ...


9

Short version: consider the working set size. Long version: How big is your data? If it can fit in memory of a modern server, yes, you're absolutely right. Unfortunately, the biggest Xeon can address 2TB of RAM right now, and that's not that big of a dataset any more. If you can't buy machine big enough to house your entire working set in RAM, you're ...


9

Considering how cheap a 1TB disk is, get another and create a RAID5.. Redundancy and storage.


7

If you want speed: Increase RAM so at least frequently used indexes can entirely fit into RAM (for example, on a system I work on, 32GB RAM is plenty for a 350GB database, because indexes are what you need in RAM, not raw data) Use RAID10 with any disks (faster disks are better) Avoid RAID5 Split mdf, ldf and temp DB onto discrete spindle sets (example: ...


7

The optimal configuration is (usually) as follows: On the host, set elevator=deadline Use virtio and only virtio use raw LVs whenever possible. Qcow2 gives overhead. Files on a FS also have overhead in the VM use the elevator=noop both in host and VM, use noatime,nodiratime in fstab wherever possible Make sure the virtio drivers are up to date, especially ...


5

If the platter sizes are different, then yes. In that case, average seek latency will be lower on the smaller spindle. Don't assume that these drives have different platter sizes, though. Some LFF drives use SFF-sized platters for the increased performance they offer.


5

If the data is so sensitive, you can afford to buy another disk and scrap this one, otherwise look for a specialist in HDD data recovery and ask them what they can do about your problem, as they'll probably know how to handle your request.


4

I'm based in Norway, and we degauss failed drives using equipment from Ibas. Dell, HP and IBM all accept the destroyed media as return for warranty replacements.


4

Supermicro.com do.


4

For the same reason that I thought it was important to point out the problem, I think it's also important to point out that Dell announced that it was rolling back this policy, and that in the 2nd quarter of 2011, they will release a firmware update to the H700 and H800 controllers that allow 3rd party drives. Thank you Dell. I'll be writing a new blog ...


2

The device in /dev doesn't exists because the Linux kernel isn't notified about partition table changes automaticly. You can use partprobe (part of Parted) to re-scan your partition table or reboot the machine to have your device nodes created.


2

I've had pretty good luck with servers from Aberdeen. They are typically re-branded SuperMicro servers. You may also want to look into Silicon Mechanics although they only offer 3 a year warranty. Edit: I should have also mentioned that Aberdeen sells on-site support contracts that are managed by a third party. They can tell you more about that if you ...


2

i just got mail from dell stating: DELL H-SERIES CONTROLLER BLOCKING OF 3RD PARTY DRIVES TO BE REMOVED IN Q2FY11 Dell will no longer prevent customer use of non-Dell hard drives attached to PERC H700 and H800 controllers on our 11th generation of PowerEdge servers. A Q2FY11 PERC Firmware Update is planned to allow non-Dell hard drives to be used with these ...


2

You could build your own electromagnet. Get an iron nail, wrap copper wire around it (the more wraps the better) and connect the ends to the positive and negative leads on your battery (9 volt, 12 volt, etc). However, again this method most likely won't work. Firstly, it's an unchanging current and secondly the strength of this magnetic field pales in ...


2

Greyhole will distribute your files across multiple drives. It will also allow you to specify redundancy, so that certain files have redundant copies stored on multiple drives. It is targeted at the home server or workstation and not as a production enterprise solution.


2

RAM is the new disk, disk is the new tape. In http://www.tbray.org/ongoing/When/200x/2006/05/24/On-Grids . Note that was six years ago. Yes, we have database systems that try (and try hard) to keep the entire dataset in RAM and rather shard to multiple machines than to use the disk because disk is magnitudes slower anyways. You need to write out the ...


2

Problem is that ZFS doesn't allow for removing drives from the pool, you can only add devices. If you want true Drobo-like experience (on any OS) you will have to wait for btrfs reaching mature status or at least RAID5/6 functionality (which is scheduled for 2.6.37).


1

How about following method I wrote: http://nakanoteblog.blog136.fc2.com/blog-entry-3.html This is the way creating same size partitions from hard disks of varying sizes building RAID devices with partitions concatenating RAID devices with LVM as a logical volume


1

It sounds like all you care about is being able to utilize all 2TB of storage without having to manually place files on one drive or another. Either LVM or RAID0 can solve this problem for you at the expense of increased risk of failure. For LVM, you would make each 1TB drive an LVM physical volume and put them both in a single volume group. After that you ...


1

The right way is to use LVM. Personally I just put most of my media collections on one disk, other things on the other disk.


1

This question is similar to a basic one that has led to a lot of research and development in database architectures over the past 5-10 years. Now that it is feasible to store an entire database in RAM for many use cases, the database needs to be designed around working in RAM, rather than simply applying older inherited architectures to RAM-based storage. ...


1

Try setting "deadline" as the I/O scheduler for your host's disks before starting KVM: for f in /sys/block/sd*/queue/scheduler; do echo "deadline" > $f; done If you have I/O bound load, it might be your best choice as this IBM paper suggests.


1

The less free space there is on a drive, the more fragmented newly-created files will be. Imagine two hotels, one mostly empty, one mostly full. Three people want rooms next to each other. Which hotel will be more likely to be able to give it to them?


1

You could also try freezing the hdd. Other have had varying success with it, here one. The good news is that it won't do any physical damage to the drive that the retailer will be able to detect, so it shouldn't void your warranty. Just be sure to remember to put your drive in an Ziploc (or similarly airtight) bag to keep out any moisture.


1

Was this disk the main disk? I have found that when a HDD goes I can usually pull stuff off of it if I put it in another machine as a slave drive and get what I can (or use a live CD). You might be able to run some sort of drive wipe this way. When the OS is running off a failing disk it's nearly impossible to do this.



Only top voted, non community-wiki answers of a minimum length are eligible