Video Buying New Commercial IT Hardware Isn't Always Worthwhile (Video) 92
Robin Miller: This is Ben and he works for a company called MarkITx. Am I pronouncing it right, Ben?
Ben Blair: Absolutely, MarkITx, yeah.
Robin Miller: They're dealers of used enterprise level IT hardware, which means they have kind of an interesting view on what’s hot and what’s not, and what’s going to be hot and what people are buying and what they’re dumping.
So, Ben, what’s hot?
Ben Blair: We see mostly networking gear – well one small correction. We’re a place to buy and sell. We don’t buy and sell ourselves, so we help the buyer and seller come together, but like you said we see all this order flow, and what’s interesting – so what’s hot is mostly networking gear. We see a whole lot of Dell and HP servers as well, and a lot of other kinds of equipment but that’s sort of like the fat tail of things, right, there’s like thousands of things we see once or twice, but a whole lot of things. It’s Cisco, Gear, Juniper, Arista, Aerohive, HP, Dell, IBM servers, kind of across the board.
Robin Miller: Well, either the last video or the one before that we did was about HP's new... no they’re not blade servers, they’re calling them cartridges or something.
Ben Blair: Yeah.
Robin Miller: Basically they stick a whole bunch of single board servers in a chassis. I remember that as blade servers, but that’s because I’m old. So it’s obviously not that
Ben Blair: Heh. I mean, yeah, it’s just people are looking for new ways to get better power density and better weight density really in their cabinets, and it makes a lot of sense. I mean, you raised an interesting question in the email (that led up to this interview), actually, which was sort of what’s driving a lot of these changes and the OEMs and the vendors I think would have you believe that there are new features. You know, okay, great, HP has got a new cartridge based system. You can have higher density and that’s
Robin Miller: Yeah.
Ben Blair: It improves the calculation incrementally, but if you think, I mean, back to mid-1990s and even into early 2000s, the hardware really did drive things. You’d get the next generation of Intel chips and boy, if you’re running a database that had heavy load, you needed to upgrade it. That was your path to better performance. And since we were in the web world and everything has moved to scale out. You got so much more load you couldn’t possibly fit it on a server no matter how awesome it was, right? So we had a smart and hard software problem of, “Let’s make sure this can run on like 4 or 5 or 10 or 100 or 10,000 servers,” and once you solved that problem this was really the thing that changed. It doesn’t matter so much anymore that you have one server that’s more powerful or even an enclosure that’s a little bit more power efficient. I mean, they’re advertising, I don’t know, 80-some-odd-percent power efficiency, which is fantastic. But it’s not going to fundamentally change what you can do any more. You already have to have things scale out and then it’s just a cost trade off, and this is why I’m not trying to talk about why what we’re doing is interesting, but fundamentally that it's the fact that software is eating the world, that means hardware is becoming a commodity, and so it doesn’t matter if you’re buying servers, like HP wants you to believe that if I buy this server, it will reduce my cost; if I bought cheaper servers that would reduce my cost as well, I don’t necessarily need fancier ones. And as long as my software can run equally across all of them because it’s already scaled out, it’s not this need thing, I don’t need a server that’s twice as good, right. I need twice as many servers and I can have them more dense, less dense, those are all then tradeoffs. As much as I’m a technologist and I want to be in charge of all the stuff that’s almost a finance decision, it’s like, look, I just need 50 cores and 80 terabytes of storage to do this and I need this I/O bandwidth, but exactly how that’s delivered to me, I don’t really care anymore.
Robin Miller: The funny thing is, one thing I hear made much of this power savings. I could see that being important to Google. I could see that being important to IBM or Amazon, but I do not know the people I know who are IT guys or like small and medium sized businesses. You have 200 employees in your business, you have 500, you have 1000. Let’s say you have 150 desktops, you have 200 computing devices in a factory or warehouse, and some of them could be tablets, who knows, and you have a bunch of servers, let’s say 50 servers even. How much money are you going to save on electricity? $100 a year?
Ben Blair: Yeah. It’s a good question. I mean, yeah, on a scale of 50 servers, I mean it may be more than that, but let me give you the answer, slightly different question, but one that I can answer to say I have more firsthand experience.
Robin Miller: Yeah.
Ben Blair: Now suppose you’re in a co-lo facility, right? And there the costs are more concrete, like I’m paying $2,000 a month for a cabinet and however many amps of power are going to that cabinet, so my power costs are fixed, but what I can do with that power is sort of where it matters to me, so if I can have servers that are twice as efficient I can put twice as many servers in that cabinet before I need to buy a whole new cabinet for another $2,000 a month or whatever. So there it can matter, but it’s, I mean, I don’t know, a 2x factor; like in power efficiency it is often more like 20%, 30% as long as you are not talking versus something 8 years old.
Robin Miller: But even if it’s 2x and you’re talking at the most in mid-thousands annually, it’s not a world changer. It’s the difference between buying a cheap used car or not buying one.
Ben Blair: Yeah, exactly, and it’s just trading off sort of operational dollars for dollars upfront; like your car example, I can buy a more expensive car today that saves me gas, but how long before that pays off?
Robin Miller: Used equipment, obviously you guys are brokering it, it’s coming through your thing, this is getting hot, you are seeing volumes go up?
Ben Blair: Yeah, volumes are up really significantly. I mean, if you look at it, I am sure you’ve experienced this and probably everyone who's on Slashdot has a closet full of old stuff, right?
Robin Miller: Yeah.
Ben Blair: And this is true whether you are a small shop or if you are a giant bank, except instead of a closet it’s a warehouse, but it’s hard to deal with this stuff and as technologists, this is sort of eye opening for me getting into this, like I always had the view too that I want the thing that’s going to help me do my job and the cool new thing. And then I get the next generation, what do I do with this stuff? I don’t know. I don’t want to throw it away, that’s really wrong, and I got all these other things to do, so just like put it back there somewhere, right.
Robin Miller: Yeah.
Ben Blair: And that’s a fine decision sort of once or twice and then you do it, you know, do it over and over and pretty quickly you got a warehouse. And so it’s crazy. So the good thing for us, it’s a pretty easy conversation to have with someone, hey, all that stuff that you think isn’t worth anything, it’s actually maybe half or a third of it isn’t worth much and we can help you recycle it. But a surprising amount, half, two-thirds, if we’re talking to like financial institutions that are refreshing every year-and-a-half, maybe it’s three-quarters of the stuff they have is perfectly good, and lots of people want to use it. We’re not in the time again -- to go back to like late 1990s, early 2000s – where corporates were increasing fast, you have technically more hardware imrpovements going. But software has changed that. So now if you’re a smart company, if you’re thinking like some of our customers, our buyers are small... mid-size and small companies, the whole business is renting used hardware, right? They may buy it new most of the time, but the minute they have one customer running on it, and then the next one, it’s used in a sense, right? And they’ll run it for 4 or 5 years.
Robin Miller: And we know off-lease, a lot of people lease their computers.
Ben Blair: Yeah.
Robin Miller: And that’s where I first saw used computers, in a commercial and legal sense, was coming off-lease.
Ben Blair: Yeah.
Robin Miller: Like you said, you rent them, and the second you take it home, it’s used.
Ben Blair: Exactly.
Robin Miller: So why not get a used one?
Ben Blair: The economics for buying even one server or one switch, it’s better if we are buying used. But it’s not going to change your world, right. Where it really matters is if you are buying 10 or 30 or 50, right, and then that can make a really huge difference and so I think, you are not talking about like the Windows 95 revolution where everything, suddenly, really was obsolete, you simply couldn’t run the software you wanted on it, and now if you’re running a pizza box and it’s got two quadcores instead of one quadcore or four quadcores or Intel or AMD, I mean, these all make incremental differences, but they don’t say whether you can run a piece of software or not, and either way, if you’re running any kind of sizable infrastructure, your software is already built to be distributed over dozens of machines, and if you have twice as many machines with half the performance, it’s not an even trade off necessarily, in some scenarios, but it’s close to it.
Re:Really? this is news in 2014? (Score:2, Interesting)
If you can afford new, you always buy new.
I would add "size" to your list, sometimes newer hardware can allow you to expand your resources, without taking up as much space. However.....
I don't think your rule is universally true. Buying new is like buying a new car. Yea it's great to drive off the lot, get better gas mileage and chances are slim you will find yourself with a broken car sitting on the side of the road, but making the payments forever can be a bitch.
You may not get depreciation, but that must means it's an expense, which comes off the top of your net before taxes, so you get it all this year.
You may get a warranty, but what's a 24 hour SLA worth to a mission critical system? What you need is a fully capable hot standby, which is easier to afford when you are buying used.
Processing Power? I don't know very many applications that *really* need more processing power, which cannot be (and likely SHOULD) effectively split up over multiple servers. Many people think that faster processing means faster application, but this is extremely rare. What's usually needed is more memory or I/O performance which can be had for a song if you buy old hardware and upgrade memory and disk drives with SSD's.
So, where some folks with money go with new, it's not universally true that this is the best use of one's money. Sometimes, if you think about the problem, cheaper, used hardware can serve you well for a lot less coin.
Re:Duh (Score:4, Interesting)
A cheapie SunFire v200/210 will run like a tank, but you'll be crippled by the server's top speed, and they do put out the heat if you push up the load average (and HVAC costs should always be factored in, yo.)
You'll also need to buy a lot of those pizza boxes to make up for the processing power that you can find in a box half its age, let alone the newer iron.
Sometimes you have to run the old stuff (I work in an environment where we have testbed boxes, and SunFires are a part of that, along with ancient RS/6000 gear, PA-RISC HPUX gear, etc. I can tell you right now that the old stuff cranks out a lot more heat (and in many cases eats a lot more rackspace) than the equivalent horsepower found in just a handful of new HP DL-360's.