[AusNOG] Server/HD Question
Sean K. Finn
sean.finn at ozservers.com.au
Thu Feb 14 18:57:06 EST 2013
+1 LSI over Adaptec.
Sent from my iPhone
On 14/02/2013, at 2:28 PM, "Tony" <td_miles at yahoo.com<mailto:td_miles at yahoo.com>> wrote:
Check the RAID card then, a lot of the ones doing hardware RAID-1 don't write any extra meta-data to the disk or mess the disk stucture up, they just write the data to both disks at the same time. This means that you wouldn't have any concerns about what happens if your RAID card fails as either disk can be put into any machine and used as just a normal disk. I haven't played with server hardware for ages, but the entry-level IBM RAID using LSI Logic controllers was like this, two disks, plain mirroring of the disks to each other. Have to be very careful when replacing a disk which one you mirror from/to as the controller doesn't know because there is nothing "special" about the disk that it can identify the "RAID structure" on the disk.
Whilst we're tlaking about SSD & SAS/SATA, you can also get RAID controllers that have the option to add SSD as cache disks to store frequently used data on to give some of the benefit of fast access to SSD without worrying about them dropping dead quite so much.
Most hardware RAID controllers have a number of features to avoid the issue of having to find the same identical model controller if one dies, but this is a concern. There are several ways to work around this, you can purchase HW maintenance from your vendor and rely on them to have spares for you. If you think RAID controller is a single vulnerability that you don't like, you could purchase two and have a hot spare sitting on your shelf. You could also build in some redundancy at higher layers so that VM's are spread across multiple real machines and recover without downtime if one fails, but you're talking bigger $$ there. Then there is the option of using something like DRDB to give you a RAID-1 set across two machines which is quite interesting:
http://www.drbd.org/
regards,
Tony.
________________________________
From: Skeeve Stevens <skeeve+ausnog at eintellegonetworks.com<mailto:skeeve+ausnog at eintellegonetworks.com>>
To: Jake Anderson <yahoo at vapourforge.com<mailto:yahoo at vapourforge.com>>
Cc: ausnog at ausnog.net<mailto:ausnog at ausnog.net>
Sent: Thursday, 14 February 2013 2:02 PM
Subject: Re: [AusNOG] Server/HD Question
The server only has 2 bays in it. Can VMware do software RAID?
...Skeeve
Skeeve Stevens - eintellego Networks Pty Ltd
skeeve at eintellegonetworks.com<mailto:skeeve at eintellegonetworks.com> ; www.eintellegonetworks.com<http://www.eintellegonetworks.com/>
Phone: 1300 239 038; Cell +61 (0)414 753 383 ; skype://skeeve
facebook.com/eintellegonetworks<http://facebook.com/eintellegonetworks> ; <http://twitter.com/networkceoau> linkedin.com/in/skeeve<http://linkedin.com/in/skeeve>
twitter.com/networkceoau<http://twitter.com/networkceoau> ; blog: www.network-ceo.net<http://www.network-ceo.net/>
[cid:]
The Experts Who The Experts Call
Juniper - Cisco - Cloud
On Thu, Feb 14, 2013 at 2:55 PM, Jake Anderson <yahoo at vapourforge.com<mailto:yahoo at vapourforge.com>> wrote:
On 14/02/13 14:43, Skeeve Stevens wrote:
OK, so today is a day of learning.
As I said.. I am NOT a server hardware guy. Based on some of the responses, I now have learned that SATA drives will fit into a SAS interface (but not the reverse). This is awesome... I thought they were completely different.
So, it is a DELL 1950 with 32Gb ram, 2 x Dual Core 3Ghz processors. I am still unclear what, if any RAID is on board.
I would stay away from "hardware" raid, if you don't need it for ultimate performance use software raid.
When the computer craps itself, slap the drives into a new host and be running again, no need to hunt down the exact same model of raid card.
The server will be used for a dozen VMs. Centos, general purpose, DNS, radius, etc with httpd front end and some mySQL backend, but all low performance. The sort of VMs you commit 128mb of ram to and max at 1-1.5Gb ram.
you can never have too much ram, this applies double with VM's ;->
FYI, I know how to manage Vmware and staggering the boots, and so on.
Oh yeah.. and I did mean 7.2k drives.. ;-)
So.. now that I know I can use SATA drives... it opens things up a bit.
I am looking at Ingram.. still some thing I don't understand.
HP 2TB SATA 6Gb/s 7200 HDD<https://au.ingrammicro.com/_layouts/CommerceServer/IM/ProductDetails.aspx?id=AU01@@2210@@10@@000000000001885395> - $292
HP 2TB SATA 3Gb/s NCQ 7200 HDD<https://au.ingrammicro.com/_layouts/CommerceServer/IM/ProductDetails.aspx?id=AU01@@2210@@10@@000000000001551449> - $342
No idea what NCQ is, and why the faster TP drive is cheaper. Anyone?
NCQ is native command queing, the HDD will re-order buffered reads on the fly for the best average access time.
Can be a decent gain on a heavy disk load.
These look nice:
SEAGATE Constellation CS SATA 2TB 3.5" 7200RPM 64 Cache<https://au.ingrammicro.com/_layouts/CommerceServer/IM/ProductDetails.aspx?id=AU01@@2210@@10@@000000000002145465> - $159 lots in stock
SEAGATE Constellation CS SATA 3TB 3.5" 7200RPM 64 Cache<https://au.ingrammicro.com/_layouts/CommerceServer/IM/ProductDetails.aspx?id=AU01@@2210@@10@@000000000002145467> - $219 lots in stock
I see some NL drives (assuming what someone said was Near Line) - no idea what that is though.
I was also thinking that if the chassis had the space, I should put a little SSD of CF on board to install the ESX onto to keep the OS off the drives... thoughts? My assumption is that once VMware is booted, its disk access is minimal.
I would doubly not do this, or if you do, raid them.
SSD's have the same failure rate as rotating media. Often with worse failure properties, one day they just disappear, no degradation, bad sectors or anything first.
I would put a RAID1 in for boot and put your OS on that.
however you configure the rest of the array, its good if the system can still get booted regardless of the degradation.
Given a number of VM's running partitioning is perhaps a good thing to look at, rather than one bigass raid5 array, put a few smaller raid1's in.
Spread the VM's across the spindles so if a mail system starts thrashing grabbing somebodies emails, when their mail client starts pushing it back onto the file server you aren't running all that on the same set of spindles.
Be sure to align all the VM's with sector boundary’s. Right the way through the chain. IE the VM's internal partitions should start on a boundary of the physical disk.
...Skeeve
Skeeve Stevens - eintellego Networks Pty Ltd
skeeve at eintellegonetworks.com<mailto:skeeve at eintellegonetworks.com> ; www.eintellegonetworks.com<http://www.eintellegonetworks.com/>
Phone: 1300 239 038; Cell +61 (0)414 753 383 ; skype://skeeve
facebook.com/eintellegonetworks<http://facebook.com/eintellegonetworks> ; linkedin.com/in/skeeve<http://linkedin.com/in/skeeve>
twitter.com/networkceoau<http://twitter.com/networkceoau> ; blog: www.network-ceo.net<http://www.network-ceo.net/>
[cid:]
The Experts Who The Experts Call
Juniper - Cisco - Cloud
On Thu, Feb 14, 2013 at 1:10 PM, Skeeve Stevens <skeeve+ausnog at eintellegonetworks.com<mailto:skeeve+ausnog at eintellegonetworks.com>> wrote:
Hey guys,
I know a bit about operating servers, but know bugger all about the hardware, especially when it comes to hard drives.
I am needing to deploy a server for some low performance VM's, and it has 32Gb ram, Dual core dual processor 3Ghz... so all good. Should run a few linux VM's on ESX.
But... hard drives I really don't know about.
The chassis takes SAS drives. In it are some small drives and I want to upgrade.
Will 7.5k speed drives be ok on a VMware server if its not for high performance processing? The cost of 15k SAS drives still seem to be rather expensive. I was hoping for 500Gb-600Gb of space.
Any thoughts on which brand is ok... I will be putting a single drive in mirrored (only 2 bays).
...Skeeve
Skeeve Stevens - eintellego Networks Pty Ltd
skeeve at eintellegonetworks.com<mailto:skeeve at eintellegonetworks.com> ; www.eintellegonetworks.com<http://www.eintellegonetworks.com/>
Phone: 1300 239 038; Cell +61 (0)414 753 383 ; skype://skeeve
facebook.com/eintellegonetworks<http://facebook.com/eintellegonetworks> ; linkedin.com/in/skeeve<http://linkedin.com/in/skeeve>
twitter.com/networkceoau<http://twitter.com/networkceoau> ; blog: www.network-ceo.net<http://www.network-ceo.net/>
[cid:]
The Experts Who The Experts Call
Juniper - Cisco - Cloud
_______________________________________________
AusNOG mailing list
AusNOG at lists.ausnog.net<mailto:AusNOG at lists.ausnog.net>
http://lists.ausnog.net/mailman/listinfo/ausnog
_______________________________________________
AusNOG mailing list
AusNOG at lists.ausnog.net<mailto:AusNOG at lists.ausnog.net>
http://lists.ausnog.net/mailman/listinfo/ausnog
_______________________________________________
AusNOG mailing list
AusNOG at lists.ausnog.net<mailto:AusNOG at lists.ausnog.net>
http://lists.ausnog.net/mailman/listinfo/ausnog
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.ausnog.net/pipermail/ausnog/attachments/20130214/1188e8f2/attachment.html>
More information about the AusNOG
mailing list