Slashdot Mirror
Buying New Commercial IT Hardware Isn't Always Worthwhile (Video)
Ben Blair is CTO of MarkITx, a company that brokers used commercial IT gear. This gives him an excellent overview of the marketplace -- not just what companies are willing to buy used, but also what they want to sell as they buy new (or newer) equipment. Ben's main talking point in this interview is that hardware has become so commoditized that in a world where most enterprise software can be virtualized to run across multiple servers, it no longer matters if you have the latest hardware technology; that two older servers can often do the job of one new one -- and for less money, too. So, he says, you should make sure you buy new hardware only when necessary, not just because of the "Ooh... shiny!" factor" (Alternate Video Link)
Used hardware vendor says rack space is free...run your data center on Pentium 3s. News at 11.
John McAfee 'It was like that time I hired that Bangkok prostitute; to do my taxes, while I fucked my accountant'
A guy who is the CTO of a company that deals in used hardware tries to urge people to buy used servers instead of new ones.
I can't see the video but in the summary he mentions using two old servers to do the job of one new server. I appreciate the recycling, but it sounds like he is talking processing or I/O equivalence, and usually it is power that is the dominating factor in data center effectiveness. Are two servers really cheaper than one when you factor in electricity, cooling, and rack space?
"Here Lies Philip J. Fry, named for his uncle, to carry on his spirit"
...is on an "upgrade" cycle. All equipment with red LEDs needs to be replaced with equipment with blue LEDs, at least on the front face of the equipment.
The CEO toured the data center recently and wanted to see blue LEDs on everything.
the guy who's whole business model is dependent on companies buying used hardware.
My karma is not a Chameleon.
What's new about this advice? Was it not as useful and applicable 50, 100, and 1000 years ago?
In Soviet Washington the swamp drains you.
Guy who sells used computer hardware claims that buying new computer hardware is a bad idea, and that you should buy used gear instead. News at 11.
Not what this guy is saying is wrong, but there are other unaddressed issues. They cover issues like "power savings", but not the much more important issue of buying an unknown piece of hardware from an unknown vendor, without a warranty. Aside from that, sometimes there are issues of physical constraints-- like I have limited space, limited ventilation, and one UPS to supply power. Do I want to buy 5 servers, or one powerful one?
Also, it's not true that hardware isn't advancing. In the past few years, USB has gotten much faster, virtualization support has improved, drives and drive interface has gotten faster, etc.
And sometimes, buying "new" is more about getting a known quantity with support, rather than wagering on a crap-shoot.
Seriously? I don't think any self-respecting, halfway competent IT department has purchased new hardware because of "oooh shiny" in I don't know how many years.
You buy new hardware for three reasons:
1)Warranty coverage
2)Need of processing power
3)ROI from depreciation.
Yeah, you can keep old, fully depreciated equipment around, but you're giving up a lot in terms of tax writedowns from depreciation over the lifetime of the equipment. You also don't want to keep old gear around because if it doesn't have a service contract attached to it, and it breaks in production, you are FUBAR. And finally, yeah, you could keep those old servers around and make a big cluster for VMs, but they eat up rackspace, cooling, and power.
If you can afford new, you always buy new.
Oh, great, it's hard enough to replace obsolete equipment as it is. Once management sees this, they'll wonder why we can't keep that old Dell server going a few more years - after all, other companies are buying the same server for this guy. IT will never get another upgrade approved ever again if this gets out. Forget the cost savings of lower-power equipment, and the massive throughput increases in newer drives.
There is one business case for buying old - when a production machine is gone and you can't get parts for it any longer, but you have to use it, you can buy a spare machine or two or three of the exact same model and swap out parts. Used computer brokers are good for something like this.
Little Dell running SQL 2000 on Windows 2000. A whole gigabyte of RAM. Made it through 12 years without an issue. I've been told it's off living on a farm now.
Do you actually feel embarrassed about having posted this, or did you do it willingly?
What I mean is, most businesses keep everything of importance on their servers. Think of the salaries they pay in total all of their employees who spend time in front of computers each day. Everything you pay them to do is, essentially, tied to those servers. If the server runs a hosted application slowly, then all of your people using that application are forced to work more slowly -- making them less efficient. If a server crashes and people lose access to information until it's brought back up -- even more inefficiencies result.
Now, WHY would you cheap out over what's probably a $10,000 or less price difference buying a new server with a warranty, and some guy's used one that's less powerful (but probably uses just as much electricity and requires just as much cooling)?
As it is, I've never worked anyplace where servers get swapped out all that often. When it's time to shop for a new one, they've typically gotten a good 4 or 5 years of 24 hour/7 day use out of the old one already. (In bigger places where they get upgraded more often, I suspect they do a higher volume of business too -- and make more profit with each server than the places I worked.)
Used servers are nice to resell at steep discounts so "end users" get the opportunity to tinker with them. They're probably great as someone's home media server, or for the software developer who wants to experiment with hosting his/her own software app. They're probably even an option for the people who couldn't ever afford new systems in the first place (like charities running on shoestring budgets). But for most of corporate America -- IMO, servers should be purchased new and used no longer than their warranty period.
"No shit Sherlock"
And the sad part is some CFO will see the video clip, override the CIO's IT Plan for updating their hardware infrastructure and then complain about a lack of 110% uptime
Captain fucking obvious is looking for a sidekick.
I had to sign up just for this, is this a press release of sorts ?
I know it's a TLA and a corp buzzword but I think it applies here. For most scenarios the actual cost of the hardware is a small part of the total cost of ownership when you factor in things like space, power, cooling, maintenance, support staff, purchasing a warranty, etc.
I'd expect that most of those other factors will be higher for used hardware as I expect them to be less efficient which will require more space, power and cooling. They are used so equipment so I'd expect more failures and maintenance required. Overall I'd expect that these other costs in TCO will eat up any savings on the initial hardware cost and you'll likely end up spending more in the long run. Especially when you factor in that most/a lot of the companies making this stuff requires support/maintenance contracts to even get access to things like updated drivers or firmware updates. Often including critical security fixes so a contract may not be optional if you want to have a secure environment.
Sure there may be cases where it does save you money but you need to look at all of your costs and that isn't easy when you have so many aspects to consider and are forecasting out into the future.
I swear we saw an identical article a few months ago.
Go away.
We do not want your advertisements. Nobody wants your old gear. I pay you guys to haul it away, not sell it back to me on Slashdot.
Next Week; Linux rubbish at server tasks says Microsoft Reseller
There's more truth to that than I'd like to admin. The EMC VMAX, their Cadillac product, has a giant blue light bar across the front of the rack. The light bar is redundantly powered, so that it always looks good if a C-level does a data center walk through.
We consolidated about 20ish old servers (and added new systems) in to two Dell R720xds that are VM hypervisors. Not only does this save on power n' cooling but it is way faster, more reliable, and flexible. It is much easier and faster to rebuild and stand up a VM, you can snapshot them before making changes, if we need to reboot the hypervisor or update firmware we can migrate VMs over to the other host so there's no downtime. Plus less time is wasted on admining them since there are less systems, and they are newer.
On top of that they have good support contracts, and some excellent reliability features that you didn't get on systems even 5ish years ago (like actively scanning HDDs to look for failures).
Big time win in my book. Now does that mean we rush out and replace them with new units every year? No, of course not, but when the time comes that they are going out of support, or more likely that usage is growing past what they can be upgraded to handle, we'll replace them with newer, more powerful, systems. It is just a much better use of resources.
Roblimo has thick skin. He said so in March when he posted an advertisement for this same service in the disguise of a story. This still looks like front page placement of an ad for a friends company. How much does it cost for front page placement ?
Not everyone can buy used hardware, but for those who can, doing so is a huge money saver ($75K worth of hardware six years ago is now selling for thousands). Case and point, we bought a fully stocked 16-blade system for about $4K with Quad-Core Xeons and 4GB of RAM. People might say that is crap, but not when what you're replacing is already crap of the crap and upgrades are cheap as well. When factoring in clustering, etc, running on used equipment is hardly risky. Support-wise, this stuff usually has software available that is rock-solid because it's been around for a while. So yes, you wouldn't want to run mission-critical apps on it, but you can get a way with running a ton of auxiliary infrastructure. Never mind making for great test systems.
When IBM releases new hardware the best time to buy is later. When INTEL release new hardware the best time to buy is now.
Otherwise you get no HA. If you mean X new modern servers crammed with RAM, etc, versus 6X old servers though, then yes, absolutely and without question. Better, more reliable, easier to support, smaller footprint, more efficient, and quite frankly, cheaper in the long run.
For your primary servers, power is a very important cost consideration of course.
On the other hand, I buy Raritan 16 port IP KVMs that are BETTER than their new models at 90% lower cost. I use them a few times power year. Their better than the new ones because they have a perfectly good web interface I can use from my phone to take care of a server that it down, rather than having to drive to office to use their proprietary control software for the new ones.
Similarly, I use some very popular 16-bay storage boxes that I get for around $100 used. It's nothing more than a metal box with a SAS expander in it. There's darn little that can go wrong with what is essentially just a case and sleds, so why would I want to pay $X000 each for them?
The people talking about tax depreciation obviously haven't thought it through. You pay lower taxes by having lower profits. Sure, spending $20,000 on equipment means you can (slowly) deduct $20,000 from your taxable profit, thereby reducing your tax by $4,000. You just spent $20,000 to "save" $4,000. That's not exactly a brilliant move, especially since that $4,000 is depreciated over at least five years. You want to spend $20,000 now to get $4,000 back five years from now? I see why you're a computer geek and not an accountant (or manager).
It entirely depends on what you are doing with it. If the task is not CPU bound on an old box you don't need a lot of them.
I've got one old sparc box for occasional use for some legacy software from 1996 and 2002 - it flies on a machine from around 2008. Another has a pile of old tape drives of various types hooked up to it, once again, for occasional use. The only gain in either situation from replacing them is theoretically increasing longevity. Neither case lends itself to a virtual machine unless the thing running that VM has a sparc processor, in which case there's no point for a VM.
It doesn't work like that.
To use a car analogy that would be like assuming each car is moving ten tons all the time just because the motor can do it.
When you have SMB type customers then refurbished hardware is great value. They're usually not willing to fork out for a new server. When there is refurbished hardware for a fraction of the price -- still new enough to be reasonably efficient and to add a HP Care Pack or whatever -- why not? Having hardware that is up to scratch is both good for you and good for your customer. Out of dozens of customers of this nature we've never been bitten (and yes, the customer knows the server is refurb + Care Pack).
It's really great when you get a strong business relationship going with your local refurb business. Getting the pick of the litter really gets your geek juices flowing!
We did have a reasonably strong virtualization setup too, and that helps as the article suggests.
The laptop I am typing this on right now is a refurb model that I got for an excellent price a year and a half ago. It's probably the best laptop I've ever had including brand new ones.
At one point the interviewer asks "how much money you gonna save on electricity for 50 computers, $50/year"? It's clear he's never even attempted to do the math. An extra 100 watts in California is going to cost $314.91 per year at the typical rate (above baseline) of 35.949 cents per year. That's just the savings on one computer system much less 50 computers.
This guy is ignoring two very important factors here involved in purchasing of IT hardware in any enterprise.
- Hardware is a capital cost whose depreciation can be written off every year on your corporate income tax. After 4 years or so, your hardware actually now has near zero actual capital value to the company. Thus, as long as a company believes they will be around to see the depreciation of the asset fully written down, it is of little advantage to sacrifice performance in order to save some inconsequential amount on the hardware. This is why companies always buy the latest and greatest.
- The money you spend on the system is just the one-time capital cost. The on-going costs - the electricity used, the maintenance costs, the costs of extending the warranty - these will all be substantially higher per unit computed with older systems than newer systems.
HP's weakness is their prices.
The point being that the capital depreciation can be offset making the TCO lower. It is an important factor in the math. You want to compare TCO of your current kit vs. TCO of optional kit, and a 4K swing might change the winner.
To me this sound like blatant advertising and marketing for a Gray Market vendor. When did Slashdot turn in to a sales rag?
With respect, basing a precise number on a wild guess (eg. $13k vs $10k or $1k) is pointless numerology even if it is a common bad habit.
In general terms you have a point but in specifics you are probably out an order of magnitude or two, especially in terms of little servers that probably draw under 100W while fully CPU bound despite having a 500W rated power supply.