Slashdot Mirror
Cloud Boom Drives Sales Boom For Physical Servers
jfruh writes: The promise of the cloud is that your storage and computing problems will be abstracted away from messy physical objects that you need to maintain, taken care of far way by other people. Well, it turns out that those other people need to buy a lot of servers.
That's because those other people are stupid. They just need to put everything in the cloud!
Get free satoshi (Bitcoin) and Dogecoins
> Well, it turns out that those other people need to buy a lot of servers.
Brought to you by Captain Obvious.
Why don't providers just buy storage from the cloud?
Oh, wait.
Oliver's law of assumed responsibility: If you're seen fixing it, you will be blamed for breaking it.
Sally sells seashells by the seashore?
How many chucks would woodchuck chuck if a woodchuck could chuck wood?
Great Cloud Bubble you got there. When these site go under, BOOM, cheap rack hardware.
Seriously, I stand to make lots of money doing work for these places, so go on ahead and build out those clouds! I'll find new work when the hype is over. No prob.
This is the NSA, we're gonna geet U h@x0r5! Also, what is a h@x0r5?
Well, it turns out that those other people need to buy a lot of servers.
I'm trying to understand the purpose of this statement. Cloud services are great. They are used for backup, collaboration and replacement of in-house servers. I would expect demand to increase.
The point is that those other people are buying less servers than all the people moving to the cloud were buying. If that isn't the case then its a broken business model.
...driven by continued investments in the hyperscale server infrastructures that power public and private clouds.
Duh.
We need a manageable and scalable Java-based enterprise solution to handle the synergistic integration of our para-virtualized LAMP platform.
http://arstechnica.com/informa...
Looks like Slashdot Media's subsidiary is a bad netizen. Shame!
Theory says that the move to cloud should reduce global demand for servers since each individual company won't have to provide for its own compute & storage capacity overhead and can instead rely on both the "elasticity" and the efficient VM packing/balancing of the cloud.
The reality, however, is that the race to the cloud has cloud providers throwing money into cloud infrastructure and charging customers a pittance compared to the capital investment. This has corporate users of the cloud using more capacity than they otherwise would.
A cloud boom is called "thunder."
"A lot of nontechnical people don't understand that \"moving services to the cloud\" is sometimes precisely moving services to someone else's servers in someone else's rack in someone else's datacenter(s)."
as opposed to what? oh, right - putting it in the cloud...
One thing to note is that "hyperscale public clouds" like Amazon, Microsoft and Google don't use off-the-rack HP, Lenovo or Dell hardware. They're using Open Compute Project-style designs contracted out to whitebox vendors. So, where's the demand for name brand servers coming from?
Even though we use virtualization extensively, everything is still in house. I wonder how much of a dent public cloud is actually making in corporate server infrastructure. Sure, some web startup supporting a phone app is a perfect use case for the cloud...but does it meet the needs of most companies?
The promise of the cloud is that your storage and computing problems will be abstracted away from messy physical objects that you need to maintain, taken care of far way by other people that are not well treated for their work.
At least the first mainframe era had some respect for the people involved in the infrastructure. These days, globalization has killed it in favor of mistreatment and abstraction of the workforce.
Twitter supports and protects racists - by smearing their critics with the "Hate Speech" label.
Heroin dealers make the first few hits free or really cheap because when you still have a choice, they need to sell you on it. After you're seriously addicted, the price can be raised because you no longer have the ability to say no.
Similarly, even if it means losing money for a while, cloud providers have to make the cost per unit of compute per hour look very attractive and practically give it away at first, so your IT and Line-Of-Business groups at your firm think cloud is much cheaper than all that physical infrastructure and expensive IT staff you've been paying for. However, some day in the next few years, once you and companies like you have closed down your in-house datacenters and laid off most of your IT staff, you'll find the cloud providers don't need to be competitive with that local choice you no longer possess and the cost of cloud will go though the clouds.
oh c'mon, you knew someone would post it.
Oliver's law of assumed responsibility: If you're seen fixing it, you will be blamed for breaking it.
> Heroin dealers make the first few hits free or really cheap because when you still have a choice, they need to sell you on it. After you're seriously addicted, the price can be raised because you no longer have the ability to say no.
I always wondered if you could circumvent this by getting all of your friends to solicit free or really cheap hits.
Oliver's law of assumed responsibility: If you're seen fixing it, you will be blamed for breaking it.
" HP wasn’t able keep up with its competitors. The company’s revenue share dropped from 25.5 percent to 23.8 percent, while its market share by volume dropped 2.6 percentage points to 20 percent, "
For anyone keeping score, this statement means 'HP is not keeping up because they are still in the lead, but the gap is narrower'.
"Dell increased revenue and shipments, but it too wasn’t quite able to keep up with the market. Its share of revenue and shipments each slipped by just under 1 percentage point to 17.1 percent and 19 percent respectively"
This is a little less blatantly wrong, but Dell is the #2 vendor Strictly true since they said keep up with *the market* which in aggregate grew, but being #2 in the market isn't such a dire thing.
" IBM had the third-largest server revenue, followed by Lenovo and Cisco Systems, while Lenovo was third by server shipments, "
This particular statistic is pretty screwed up because it doesn't correct for the situation that IBM sold of x86 based servers partway through the year in some parts of the world, and at the end of the year in other parts of the world. It mentions this, but fails to recognize that IBM's situation partially included Lenovo still. Lenovo's big year to year growth is mostly a changing of ownership currently.
"Cisco’s year-over-year server revenue growth of 44.4 percent was well above average for the industry, and suggests the company is not done capturing incremental market share in the server market"
Impressive and all, but given *after* that increase they still lag behind 4 other companies, it means that big year to year percentages are likely. Just like the lead experiencing a little crowding in a market shouldn't cause anyone to write them off, a large percentage gain by a relatively small player shouldn't send everyone into an excited state. You could write similarly exciting stories about some of the 'lower tier' vendors, but since those aren't exciting brands, they got omitted.
XML is like violence. If it doesn't solve the problem, use more.
Sounds like I'm in the wrong business....
Cloud services are pretty much fungible... If provider A tries to turn the thumbscrews, provider B will just under cut them in price and customers will switch in droves the next time the service contract comes up for renewal.
What will actually happen (if it's not already) is that a small number of larger cloud service providers will corner the market and drive the smaller and less efficient providers into mergers, consolidation or just plain out of business. You will end up with 3 or 4 major players, maybe more, but all large, who will dominate the market, control prices to keep the small upstarts from getting much of a foothold in the market.
The price gouging won't really start until you get the number of vendors down to near to 2 and everybody else is afraid of trying to enter the market because there is no growth left. Then prices will go up, but only enough to make the big 2 some cash, while keeping it hard for an upstart to undercut them and grab market share.
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
Funny thing is they all have the same shipping address
Ship to: Utah Data Center
Also, heroin becomes twice as addictive every 18 months.
Big data is also going to make these customers extra sticky. When you have exabytes of customer data sloshing around a thousand AWS servers, and you rely on them to deliver critical insights for competitive advantage, how long does that take to migrate out, even if you have the hardware and software all configured right now? Even if everything is set up perfectly and all the technical wrinkles ironed out, it may be a long and ugly process, because you need the big data analytics to be running at full efficacy every day.
You'll find new work taking it all down and palletizing it for the auction?
Last I heard, most "cloud" services were going belly up. The sooner, the better in my opinion.
... occurs when technological progress increases the efficiency with which a resource is used, but the rate of consumption of that resource increases (rather than decreases). [from Wikipedia]
Servers in a cloud are run much more efficiently, because of the economies of scale. Providers concentrate the expertise of running it and make using them easier for the companies, therefore server "ownership" is more economically eficient. Therefore, the consumption of servers has increased.
Not really a problem for two reasons....
1. HUGE data generally requires processing power to sort though, so where it might be cheaper to by shares of some data center's pile of rack mounted servers, once you get to a certain size, building your own makes financial sense. I'm guessing, but it sure seems to me that if you are big enough to be worried about how long it takes to move your data, you will have it locally processed anyway and won't be dependent on a cloud provider.
2. Data has value that is largely age based. Newer data is worth more than the old stuff. So if you *really* have a lot of data laying around that you depend on, you have a design and implementation problem with your big data operation. And if you don't have a design issue and really ARE processing that much data, then transferring the historical data to another provider is largely unnecessary, just switch your data feeds to the new provider and turn off the old provider once the data it has gets stale, which shouldn't be too long.
"File to fit, pound to insert, paint to match" - Aircraft Maintenance 101
The workers that actually have to deal with it are further screwed - contract labor makes things as healthy and stable as being next to a malfunctioning nuclear reactor. Long-term anything goes out the window and compensation is made on worse terms.
Twitter supports and protects racists - by smearing their critics with the "Hate Speech" label.
You should reconsider. Specialization scales. Diversification into areas you have less expertise just lets your competitors exploit a huge advantage.
Your idea about stale data is marginally true for a weekly refresh. There are orders of magnitude of base data below that. Upgrading servers at an existing provider takes months of planning, switching to a new one is much harder. (Not impossible).
We joke about the marketing buzz of the cloud, but the hardware side of things is on fire. Dell went private for a reason. Now all that remains is if SSD manufacturers are savvy enough to leapfrog HDD's.
The crossover between tricky to move is much lower than the crossover for better to run your own datacenter. Not necessarily the case though for better to run your own cloud out of someone else's colo or better to jointly administer a cloud with a colo provider. So this can happen. The cost of multiple good quality data centers is very very high compared to the cost of getting data out of one.
As far as GP's post. He's wrong. First off clouds are designed to scale so adding another copy of parts of the data for replication is not hard. Second you can do crossover networking from one data center to another if you need to move and the cost of using the cloud provider's bandwidth is too high. Also there are devices that can be physically connected to the racks and then trucked (think backup drive moving physically but 20-100x scale). Mainly a data move is the sort of thing an agent can coordinate easily.
There are a couple of problems with your theory, though it could play out that way. Right now the smaller vendors are often more efficient than the larger ones. Smaller players can be more nimble.
Second the larger players all have vastly different models. Just to pick a few examples of the bigger players
AWS -- Generic low quality server experience offered cheaply. Walmart
Sungard -- Highly custom environments quality management lots of value added labor
Verizon (was Terremark) -- Moderately custom environments, mix of high performance cloud IaaS and colo. Some value added services with strong partner service model.
Oracle -- Unified cloud stack offering IaaS plus advanced management especially knowledge of Oracle applications
Azure -- IaaS with Microsoft based PaaS. Good pricing on SQL Server.
How do those consolidate? I think we are looking at a situation more like clothing where stores are genuinely different fulfilling niches for various customers.
Yes. Overbuy and oversell your cloud hosting! Then when the bottom falls out of that market I can profit.
Sadly, a Libertarian cannot force his views on another, and freedom cannot spread as does the cancer known as religion.