Nov
4

Should You Upgrade or Replace Your Server?

Should You Upgrade or Replace Your Server?

November 4
By

One of the best features of running a Windows or Linux server is the ability to prolong its life by performing upgrades to it. Some components are easier to upgrade than others. Network cards, modems, RAM, and graphics cards are usually painless upgrades that can breathe a couple more years of life into your server. OS and motherboard upgrades are more complex because either may require upgrading other components. For example, installing a new motherboard maybe also require a RAM upgrade and so forth.

At Puget Systems we’ve been building workstations and servers for over a decade. While most of our customers hire us to build new servers, we have a fair number of customers who look to us to provide guidancSlide Ine on upgrading their older servers. Sometimes we’ll build a server for a customer who will return to us a couple of years later seeking recommendations on upgrades for a server that’s being re-purposed from web to SQL server. If the server is still in good condition, there are a number of upgrades one can make to a server to extend its life.

This week, I want to take a look at when it makes sense to upgrade the hard drive in your server, and when it makes sense to purchase a new server. We’ll cover a few other components as well.

Adding Additional Drives

Let’s start out with a simple solution. If you’re running a server that’s only a couple of years old and simply need more storage, consider adding one or more modern hard drives. This upgrade assumes you have an open SATA port on the motherboard or an open channel on your RAID controller. Let’s look at few scenarios:

You need more storage

If you’re running RAID, you’ll want to add the same brand and model of hard drive to your array. This is helpful if you have an 8-channel RAID controller but are only using four or six channels. This option will probably require rebuilding your RAID array, but it’s a great option if you have open channels on your controller card. Instructions for rebuilding your RAID array will vary based on your controller and/or you server management software. Please don’t forgot to back up your data before performing any drive upgrade for RAID rebuild.

LSI 9361-8i
The LSI 9361-8i is a popular 8-channel RAID controller

Another more involved option is to replace your current hard drives with newer, larger drives. Maybe you have older 500GB drives you’d like to replace with 1 or 2TB drive. For sheer performance in a large capacity drive, we’ve had great luck with Western Digital Black and RE drivers. The RE drives are built especially for servers running RAID. Depending on the RAID type and controller, you may be able to replace one drive at a time  and rebuild the array until each drive has been replaced. Or you might want to back up all your data to an external device, replace the drives and rebuild the array from scratch.

You need better performance

There are a number of different avenues to take when looking to increase the performance of your server, but for this discussion, I’m going to limit it to hard drive performance. If your server needs are simple, you may want to consider upgrading your primary drive running Windows or Linux. Maybe you already have a RAID array in place, but are running Windows Server off a mechanical drive. Upgrading your OS drive to an SSD will result in a substantial performance gain, and many commercial SSDs from Samsung and Intel include migration/image software to make the transition nearly painless.

SSDs to consider range from consumer-grade drives such as the Samsung 850 Pro line to the server class Intel DC 3700 series. These SSDs range anywhere from 256GB to 2TB, which provide plenty of room to run a Windows or Linux. If you’re not running your OS off an SSD you’re giving up a lot of performance in terms of the responsiveness of the OS along with boot and shutdown times. Patching the OS is also faster on an SSD.

Intel 750
PCI-E SSDs deliver ultimate drive performance. At a cost.

If your needs demand the ultimate in drive performance, consider the newer PCI-E based storage drives. These drives, such as the Intel 750 series, require PCI-Express 3.0 to run at full speed, and they can only be used as boot drive on modern chipsets like Z97 and X99. They will run on older chipsets, like Z87 but can only be used as secondary storage. PCI-E provide blistering speeds and are much closer to the CPU from an electrical standpoint making them incredibly fast. They cost a lot more today, but if your applications demand top performance, they can’t be beat.

The upgrades you make to your server will largely be determined by the server’s configuration and age. I know IT managers like to talk about how long they are able to keep a server running with proper care and maintenance, but there are limits to what you can upgrade if you’re running a server that’s more than a couple generations behind on CPU and drive technology. Our lab manager at Microsoft used to say that a server 2-years old or less was worth the effort to upgrade. Servers between two and three years, were taken on a case by case basis. But servers over three years old were replaced. Now, before you flame me, I understand Microsoft is a company that can afford to replace its servers anytime it needs to. I’ve worked for companies that didn’t have that luxury, and fully understand the frustrations involved when asked to keep old hardware, especially servers, running smoothly. You do the best you can with the hardware provided.

When to Replace your Server

This is the question IT managers have asked for years. The question is made even more complex today when you add virtutalization and cloud services to the mix. Many companies are moving applications and data to the cloud and are able to retire some of their servers, but that’s discussion for another day. Virtualization and clustering technologies can be used to extend the life of a server, or at least mitigate risk, but there comes a time when when the risk of running an old server outweighs the benefits of keeping it in operation. So when should you replace your aging server?

When your server is 3+ years old

According to IDC’s customer-based research, failure rates began to significantly climb as servers aged into their fourth year and beyond. It found that upgrading resulted in an ROI of more than 150% over three years. Replacing older servers that are performing critical functions makes option this even more sensible. Can your company continue to operate if your network or AD server goes down? If the answer is no, consider a replacement over upgrade.

When your server is out of warranty

There’s not a lot of value in adding gigs of RAM or terabytes of storage to a server that’s not under warranty. Most commercial servers are sold with a 3-year warranty, and you shouldn’t wait until a crash forces your hand. A good rule-of-thumb is to check to see if your model of server is still being built. If so, you might be able to get replacement parts without much hassle. But the PC market changes quickly, and you might find it difficult to find replacement motherboards that are more than a couple years old. Instead of installing a new, but untested motherboard with your current components, you might be better off replacing the whole server.

When your server is unstable

This should be common sense. Don’t assume that an upgrade will cure a flaky server. It’s one thing to replace a bad stick of RAM, but if Windows or your RAID array is acting up, installing new drives probably won’t fix a thing. If you must upgrade your server, get everything in order before performing an upgrade. That could mean reinstalling the OS or rebuilding the RAID or swapping out the offending component.

Conclusion

Each scenario, each company and each server requires a case-by-case evaluation on whether it should be replaced or upgraded. The time to replace your server is when it’s still running. Waiting for a crash can cause panic and more down time compared to a planned server refresh. New servers come with advanced power saving features that help save money. Lately, we’ve been building a lot of dual and quad-CPU servers that allow IT managers to combine the functions of several servers under VMs running on the new, more powerful box.

But not everyone is able to run out and purchase a new server, even if there is a business need to do so. Knowing what you can easily do to add more storage or better performing storage will allow you to extend the life of your server. Just make sure the PSU and case fans can handle the upgrade. And always make a backup!

Photo by bigpresh