• Some users have recently had their accounts hijacked. It seems that the now defunct EVGA forums might have compromised your password there and seems many are using the same PW here. We would suggest you UPDATE YOUR PASSWORD and TURN ON 2FA for your account here to further secure it. None of the compromised accounts had 2FA turned on.
    Once you have enabled 2FA, your account will be updated soon to show a badge, letting other members know that you use 2FA to protect your account. This should be beneficial for everyone that uses FSFT.

Why do Web/File servers have fast drives?

I

imzjustplayin

[H]ard|Gawd
Joined
Jan 24, 2006
Messages
1,171
I was wondering, why do web and file servers have SCSI and other fast drive technologies in there? With fibre channel and the whole bit, I wonder, why would you need wicked fast reads when all you're doing is hosting a website/file server? I ask this because isn't the real bandwidth limitation in the internet connection and the networking connection? It was only recently did we get 1gigabit and 10gigbit connections for the network, so why would in the days of 100Mb networking, would you need fibre channel or SCSI 320 drives? Please explain to me so that it would make more sense.
 
lots of people getting different files from the file server wich would be on different areas of the drive?
 
Forums and other dynamic web content requires a DB backend. With more than 1 user, it will need fast disk access to quickly provide the content.
 
Wouldn't the performance factor for database access in a case for a website be disc latency/access times and not total bandwidth/burst rates? Cause why would you need a burst rate that is significantly higher than your internet output?
 
imzjustplayin said:
Wouldn't the performance factor for database access in a case for a website be disc latency/access times and not total bandwidth/burst rates? Cause why would you need a burst rate that is significantly higher than your internet output?
Click to expand...
Maybe you can be a bit more clear about what you mean by "have SCSI and other fast drive technologies"?

From what I recall SCSI is rather 'good' in multi-user scenarios, if you look at SR, you will notice that SCSI drives shine in multi-user scenarios, i.e. when servicing lots of concurrent requests.
 
Let’s say you have 2000 people browsing your forums and around 100 or so click on any random link within a second of each other. You now have 1 or more web servers polling your database server for information at the same time. This information is scattered all over different sectors of the drive. The important thing with this scenario is seek / access time, not necessarily the speed at which the drives transfer the data since most of the the amount of data is minimal. SCSI has the fastest seek / access times because the spindle is spinning at 10 or 15k rpm. The data is passing the head almost 2X as often compared to a 7200 rpm drive. SCSI are also way better at handling multiple requests at the same time, even compared to Raptors.
 
GeekBoy2U said:
Let’s say you have 2000 people browsing your forums and around 100 or so click on any random link within a second of each other. You now have 1 or more web servers polling your database server for information at the same time. This information is scattered all over different sectors of the drive. The important thing with this scenario is seek / access time, not necessarily the speed at which the drives transfer the data since most of the the amount of data is minimal. SCSI has the fastest seek / access times because the spindle is spinning at 10 or 15k rpm. The data is passing the head almost 2X as often compared to a 7200 rpm drive. SCSI are also way better at handling multiple requests at the same time, even compared to Raptors.
Click to expand...

Thats what I was thinking... So basically a fast drive (Lots of bandwidth) is unnecessary for webpages and they really need is just a drive that that has fast seek times, correct? What about a datacenter, would it still be a 'seek' related issue or would hard drive bandwidth play a factor? I'm having trouble wrapping my head around the concept of why companies (not video editing specialists) would use Raid 0 or Raid 5 when they could likely get away with JBOD? Reason I see Raid 5 or 6 is for uptime and capacity combining but for Raid 0, just seems like for their purposes they'd only need JBOD, right?
 
Honestly I'm making a semi educated guess. ;) All I know is my buddies DB servers for his site loves the SCSI (site regularly has 5000+ people browsing forums) :D

Why is probably better answered by the DB server experts :cool: (there are a few on the [H])

I honestly do not see alot of Raid 0 if any in our production enviroments at work but I am just the lowly helpdesk (who handles dispatching the broken servers) I mostly see Raid 1 or 5 and I can understand why, I see at least one or two dead server drives per day.
 
As mentioned, it is the seek rates. Throughput doesn't matter in practice.

Another factor is that if you want many drives then SATA isn't going to make you happy with it's maximum of 2 drives per channel when SCSI has 16.

Also, error recovery on the SCSI drives is better and you have more reliable control over the write cache.
 
uOpt said:
Another factor is that if you want many drives then SATA isn't going to make you happy with it's maximum of 2 drives per channel when SCSI has 16..
Click to expand...
Ide is 2 drives per channel. Sata is only one, it's a point-to-point technology. And SCSI is really 15; one LUN is going to be used by the controller. Last nitpick (I promise) is that Sata/SAS actually has higher expandability - you can add expanders to SAS domains or Sata port multipliers to sata controllers and connect more than one drive to a port.

Raid 0's advantage over JBOD for databases is that it spreads the reads and writes more evenly over two disks when the database is small. The DB would tend to be accessing only one of the disks otherwise, and having an idle spindle is bad. Which is where the SAME philosophy came from - Stripe And Mirror Everything. All transaction logs, data, etc, go on one big stripe+mirror volume.
 
GeekBoy2U said:
I am just the lowly helpdesk (who handles dispatching the broken servers) I mostly see Raid 1 or 5 and I can understand why, I see at least one or two dead server drives per day.
Click to expand...

1st ... don't sell yourself short, you're in IT ... don't worry about the position. Everyone starts somewhere.

2nd ... I really hope your company either has a hell of a lot of hard drives in said servers, or just a hell of a lot of servers, because you should *NOT* see 1 or 2 dead server drives per day ... someone doesn't know what the fsck they're doing :p

I work in IT for one of the quote / unquote "Big Three" auto manufactures ... they have well over 2000+ servers, and drive replacements are rare at best. Granted most of the hardware is mid-range to high-end ... but still ...

Cheers!
 
sn3ak3rp1mp said:
1st ... don't sell yourself short, you're in IT ... don't worry about the position. Everyone starts somewhere.

2nd ... I really hope your company either has a hell of a lot of hard drives in said servers, or just a hell of a lot of servers, because you should *NOT* see 1 or 2 dead server drives per day ... someone doesn't know what the fsck they're doing :p

I work in IT for one of the quote / unquote "Big Three" auto manufactures ... they have well over 2000+ servers, and drive replacements are rare at best. Granted most of the hardware is mid-range to high-end ... but still ...

Cheers!
Click to expand...

Let's just say I work for a large government entity and leave it at that :p

I'm sorry maybe its 2 or 3 server drives \ controller cards per week
Definately at least 2 or 3 workstation drives per day :eek:
 
You must log in or register to reply here.
Back
Top