Hard drives have a few different metrics related to their performance statistics. One of those metrics is the drives revolutions per minute - or RPMs.

Consumer systems, especially portable computers, often have hard drives that spin at 5,400 RPM.*

Server systems often have a few different options, including SATA drives that commonly spin at 7,200 RPM, and SCSI or SAS options that spin at either 10,000 or 15,000 RPM.*

(* The newest type of drive is a solid state drive (SSD), which does not spin at all, so doesn't fit into a discussion about revolutions per minute. These type drives are also still rather expensive and limited in system configuration options - especially at the enterprise/server level. They are insanely fast though and I'm sure will work their way into the mainstream over the next few years.)

What do these numbers mean?
Let's crunch the numbers a little bit to see if we can make them useful.

One way to consider the data is to ponder how many revolutions each drive can make in one second - revolutions per second (RPS). Dividing the total RPM by 60 easily gives this number. Here is that number and also the percentage gain in RPS from one speed to the next faster rotational speed.

Those are pretty large increases in spins per second. And of course jumping more than one rotational speed - for example from 7,200 RPM to 15,000 RPM - gives huge gains - an increase of 108% in the 7,200/15,000 example.

Another way to consider the data is the inverse of this calculation. Rather than revolutions per second, let's look at how long it takes each drive to complete one revolution. We'll calculate this by dividing 60 by the RPM of each drive.That's obviously a very small number so let's take it a step further and multiply by 1,000 to reflect how many milliseconds each revolution takes.

As expected, those are fairly decent variances - a 52% faster revolution spin speed if considering the change from a 7,200 RPM SATA drive and a 15,000 RPM SCSI or SAS drive.

Is RPM all that matters?: Let's get one clarification out of the way because I suspect this article will generate a criticism otherwise - the spin speed of a drive is only one of several factors determining individual drive performance. Two other factors are the data transfer rates and how fast the read/write head of the drive can move from one part of the disk to another. This post is specifically about RPM though, so perhaps I'll touch on those other two points another time.

What do these numbers mean to me?
It seems obvious that the RPM speed of drives has some impact on performance of the drives, but let's discuss it a bit further and consider the implications.

To understand the impact of RPM speed specifically, let's assume other factors of the drive are the same.

If the system - whether notebook, desktop, or server - requests data that is just behind the read/write head so that the drive needs to spin one complete revolution for the data to be accessed, this would be a maximum rotational delay. The numbers above reflect those potential rotational delays varying from 4 ms to 11 ms.

Even 11 ms is very fast, so should you care? If your system has light usage or your primary disk operations are big sequential files then you might not care. If you run database services, a wiki, or other type of disk-intensive applications, those speed differences might add up to a substantial performance impact. The more random disk access your server performs, the bigger difference you'd notice.

Consider the difference between SAS/SCSI and SATA in those rotational delay times - a SATA drive can potentially take twice as long to find a specific location (file) on disk. That adds up to substantial performance differences under anything but very light user load.

With the cost differences between SATA and SAS/SCSI diminishing in recent years, I strongly believe it is worth considering a SAS/SCSI disk configuration for your server - for immediate benefit and to handle future growth smoothly.

At ORCSWeb our managed dedicated hosting servers are all deployed with SAS or SCSI disk configurations (and RAID) for the best performance and growth options for our clients.

Despite being a Wintel expert I really don't have any bad feeling against Apple.  I don't own any of their products, but that's a personal choice.  The iPhone plans are too outrageously priced.  Apple computers are too expensive and I can't game much on them.  As for the iPod ... I've always wanted one but could never justify the cost when I have 6GB free to store MP3s on my WinMo phone.

But when the rumors started coming to a head about the now announced iPad I thought, "This could be really good!"  I honestly wanted it to be a great product because I've always wanted a good tablet.  Now that the official product launch has occurred it just seems...lacking.

www.apple.com/ipad/

First off, it doesn't have the sleek "Apple look.” The bezel is HUGE and that one tiny button seems out of place. Maybe pictures add the proverbial 10 pounds to the iPad, but it looks bigger than I thought it would be. The thing that bugs me the most about the look is the interface. It’s just a plain old, stripped down, iPhone/OS X UI. How boring.

Somehow I expected something more.

The most shocking part, being a hardware geek, is the processor. I guess we all kind of expected a custom chip seeing as Apple took the time and effort to buy out P.A. Semi, but that chip is a huge unknown.  It will be interesting to see how it compares to the likes of the Qualcomm Snapdragon, nvidia ION and Intel Atom platforms, all of which are known to be very powerful, energy conservative solutions.  While early reviews do show the iPad is blazing fast, we still need to hear from the dev's how the iPad handles heavy graphics at full resolution.

Oh, and they kind of forgot to add multitasking to the iPad. Yes, you heard that right, you can’t listen to music and surf the web at the same time. Or work on say, I don’t know, a research paper and look up a reference online without closing Pages first. [EDIT: I am told that certain applications can multitask iPhone OS 3.2.  Still, I am convinced it is a showstopper if I can only multitask with certain Apple apps.]

I call that a showstopper.

Though I have to give Apple props on two things: price and content. For once an Apple product doesn’t give me sticker shock. Especially with that beautiful LED-backlit IPS LCD display. That should look really handsome. And we all know Apple will deliver on the content.

Overall I am extremely disappointed, and not just because of all the pre-release hype. It’s just…boring, and that’s so not Apple.

EDIT (more thoughts):

I’ve been thinking about the iPad more – it’s hard not to with all the coverage it got today – and I think what disappoints me the most is the fact that it offers me nothing new but a bigger screen. It reminds me of those giant remote controls you find at the corner drug store. The iPad is nothing more than an iPhone with a bigger screen.

While this may make other people ecstatic it is a letdown to me. I was hoping for something revolutionary and instead I got something that is barely evolutionary. In fact, I would argue that it is not even evolutionary.

Apple didn’t release the first portable media player, they released a revolutionary media player. Apple OS X, at its core, is nothing more than a modified FreeBSD distribution, but Apple gave it a pretty, easy to use, arguably revolutionary shell and put it in a pretty package. The iPhone wasn’t the first smartphone, but it was smarter than all the others. The iPad, however, gives me little to nothing new.

Maybe I’m wrong and someday I’ll beat off the throng at the local Apple store with a billy club and I’ll really like the iPad…but I doubt it. Maybe the iPad gen 2 or 3 will bring about the tablet revolution everyone is waiting for. Or maybe Microsoft or Google will beat them to it. Time will tell.

When working with client prospects to understand the differences between ORCS Web's dedicated solutions and cheap host options, many things come to light preventing an easy initial apples-to-apples comparison.

One of the common differences we see is that most hosts include a single hard drive with their standard server packages. Every dedicated solution that ORCS Web deploys has a hardware based RAID solution though. If the term RAID is unfamiliar to you, please see "What is RAID" for some background information.

Having multiple hard drives setup in a RAID configuration benefits both performance and redundancy. Server hard drives spin at a tremendous speeds and are the most likely hardware components of a server system to fail. With a RAID solution, the server stays online with no interruption and the failed hard drive can be hot-swapped while the server is online - no disruption to the hosting at all.

Personally, I would not deploy a production application to a server that did not have a RAID hard drive configuration. But, admittedly, not everyone shares this viewpoint. I've had some interesting dialog with a couple of people who say "who cares?". Their thought is that if their server drive fails, they'll just get the host to pull out the drive, put in a new one, restore the data, and get the server back online.

Well, let's consider some points for a non-RAID drive failure in a production server...

Do you have backups? I'm a HUGE believer in backups for all systems. I'll craft a future post about this point. But, if you are going to consider deploying to server with a single hard drive, you better be sure you have backups. Either manually export all your data to your personal PC on a regular basis, or be sure that your host is backing up your server.

When was the last backup? Any data written or changed on the drive between the most recent backup and the failure is going to be lost. In some situations this can be a huge deal. Lost orders; lost user accounts; lost records - all gone and never to be seen again.

How easy is the restore? If the drive that fails is just hosting standard web content only, or image files, or some other simple file type, the restore process might be fairly quick and easy. If the drive is holding the OS, or data files, or dozens of other types of complex or transactional data, it's a totally different situation. The time to actually run the restore process (getting the data back onto a new hard drive) is only part of the work. There is often a decent amount of manual work still required to get the system fully functioning again with the data current.

Swap the drives... Depending on the type of server you are running, the hard drive swap process can be really simple - pull a drive from the front of the system and slide in a new drive. Or it could be rather cumbersome - un-rack the server, open it up, pull the drive, put in a new one, close everything back up, re-rack the server. Also, if considering this option, best be sure the host can assure you they have the exact physical size, data size, and speed hard drive that you'll need, or you could be stuck without a quick replacement.

Considering those points - If the server was being backed up, and if there are spare drives sitting around, and if the restore process goes smoothly... you would still be looking at a few hours of downtime - maybe many hours depending on the specifics of the situation. And even after the server is restored you could be dealing with the potential fallout of data loss and the impact to your organization. How will your clients react? How about site visitors that stopped by and couldn't access the server?

Is it worth the risk to save a negligible amount of money on your hosting cost each month? Maybe in some situations where the users are extremely tolerant of downtime and data loss is not a concern. In the reality of today's world though, fewer and fewer Internet sites or applications have users that would tolerate such a situation when it could be so easily avoided.

I say go RAID and be safeguarded against hard drive failure.

Here is a useful post giving an introduction to Microsoft's WCAT - which can be used to generate load against a web application.

http://blogs.msdn.com/alikl/archive/2008/03/09/stress-test-asp-net-web-application-with-free-wcat-tool.aspx

"What's your server's uplink speed?"

"What?"

"How fast is your server's network connection?"

"OC3 I think."

That's not an actual discussion but I could see it being a common one with many people getting started in dedicated hosting. Non-administrators just assume the server is "connected to the Internet" and can perform at whatever speed the host advertises for network bandwidth. That is of course not quite accurate.

When reviewing differences between our standard managed dedicated hosting options and some cheap competitor options, something that shows up (much more often than many would believe) is the server's uplink speed. This is the speed of the server's connection to the web host's network core. It amazes me that some hosts have default uplink speeds of 10 Mbps then charge for an increase to 100 Mbps and charge further for an increase to 1000 Mbps (gigE).

Now, I can understand why the host would do that because the lower the uplink speed, the lower the total impact on their overall bandwidth. It would also allow them to use older network equipment (10 Mbps wasn't bad 10 years ago).

But how does that impact the hosting client? Well, it can actually have a rather large impact on the client.

Time Warner Road Runner is the ISP I use at home and when things are running well I get about 6 Mbps of download speed. How many concurrent users at 6 Mbps would it take to saturate a 10 Mbps uplink? Not many. It is more than just 1.5 because visitors don't all click at exactly the same second, but still, anything busier than just the smallest of sites is likely to see a performance impact if their server's connection is only 10 Mbps.

100 Mbps is a lot better and might be fine for small sites at an average time during the day. We host servers for many clients though who burst past 100 Mbps on the servers and who would notice performance impact at these levels.

I mention "might" because there are additional factors. Like what happens if your site is medium sized and justifies running web and database services on two different machines? That means that not only do you have visitor traffic using the uplink port, but also the communication and data being transferred between the web and database server... and you want that connection to be fast because applications can slow down quickly if the database becomes a bottleneck.

At ORCS Web all of our servers have 1000 Mbps (gigE) uplink connections to our network core. This provides for excellent burst speeds as traffic increases on the servers, great low-latency connections between the web and database servers, and no network impact when our managed backups are running. Could we save some bandwidth costs with slower uplink speeds? Sure. Could we save some network costs with older equipment? Sure. But our focus is not on providing a "cheap" solution with potentially client-impacting shortcuts just to make more money. Our focus is on providing quality solutions at competitive rates. Our focus is on maximizing value for our clients and providing solutions that best meet their needs both short and long term as their businesses, and therefore their web sites and applications, grow over time.

What's your server's uplink speed?

This is a follow-up to a recent blog post titled "Managed Web Hosting - Compare Features and Services".

John Papa has a cool blog post about capturing images from a webcam using Silverlight. Check it out.

http://johnpapa.net/silverlight/saving-snapshots-to-png-in-silverlight-4-and-the-webcam/

Something we do frequently is work with client prospects to understand the difference between our managed dedicated hosting and other options they might have found. Through these exercises we've grown confident, and the hosting prospects largely agree, that when comparing our solutions to another quality hosting provider apples-to-apples, the rates are comparable or often lower here at ORCS Web.

The potentially tricky part for someone who is looking for hosting is to notice the differences then understand how those differences might impact them. Especially in today's economy it is tempting to see a lower priced option and select it on price alone. Sometimes that cheap pricing isn't such a great deal though if additional services need to be added on to the base, or if the lack of certain services put the solution at risk of failure, poor performance, or even data loss.

Follow my blog for some near-term future posts on specific areas to note and compare when reviewing hosting package differences

I ran into an interesting issue today. We use a dedicated port range for RPC connections through firewall per this Microsoft article. Doing so allows RPC to work through dedicated hardware firewalls. We also enable the local Windows firewall on several boxes as this provides a firewall for any systems not using a dedicated piece of hardware or from other systems behind dedicated firewalls.

While using Shavlik NetChk Configure to scan systems for compliance, I noticed some inconsistencies which I traced back to a firewall issue on the server being scanned. The scans perform some of the checks over RPC. I confirmed that Remote Administration had been enabled using this command:

netsh firewall set service REMOTEADMIN enable

However,  netstat would show the connection in a SYN_SENT state on a port in the dedicated RPC range. Buried in this technet article, I found the reason:

It seems that when setting a custom range of ports for RPC via the HKLM\Software\Microsoft\RPC\Internet key, it "breaks" the Remote Administration firewall rule in the Windows Firewall. This was tested on a Server 2003 R2 SP2 system, but I suspect similar issues would apply to Server 2008.