Amazon Makes It Almost Impossible To Calculate Their "Virtual CPU" Equivalent (informationweek.com) 114
dkatana writes: AWS started out defining its virtual CPUs as being composed of EC2 compute units, or ECUs, which it defined as an equivalent to a physical Xeon processor. However, a virtual CPU now looks suspiciously variable... A virtual CPU is whatever Amazon wants to offer in an instance series. The user has no firm measure to go by. From the article:
[B]y doing a little math, you could actually compare what you were getting in virtual CPUs in EC2 versus Azure. Also by doing a little math, you knew how to compare one Amazon instance to another based on the ECU count in each virtual CPU. Microsoft didn't look too bad in the comparison.
That is one of the casualties of the nomenclature change.
I have searched for updated information on how a virtual CPU is measured and found nothing comparable to the definition of the 2012 ECU measure. I have questioned Amazon representatives three times between Oct. 27 and Dec. 21, and don't have much of an answer."
Totally true (Score:1)
In fact Amazon doesn't know themselves... I have asked them to compare to the i7 series and they simply don't have a clue.
Re: Totally true (Score:2)
when I did some testing a while back. $20 a month would get you processing of a crappie mobile you could buy for $50. Amazon has never had good cpu.
so it's probably more that they don't want to say.
anyone who's benchmarked them against alternatives has nothing good to say.
http://openmymind.net/Why-I-Di... [openmymind.net]
Re: (Score:3)
That's a good link. My real complaint about AWS is what's mentioned at the end:
It also feels like a lot of services are stuck at version 1.0, lacking that polish and continual improvement
This is what annoys me. SQS is a good 1.0 version of message queues, but the features are the just above the minimum you could possibly call a message queue. DynamoDB is a good 1.0 version of a NoSQL DB, just above the minimum you could possibly call a NoSQL DB.
These services are years old, but look like what most software does at version 1.1 or so: minimal features, no glaring bugs, but nothing great either.
I can't say anythi
Re: Totally true (Score:1)
each insrance is one of who knows how many virtual machines running on old xeons.
Re: (Score:2)
Right, but they clearly promise you a number of cores. The easy assumption is that the cores aren't oversubscribed, and none of the docs suggest they're oversubscribed, except the docs for T2 (seems like it must be). Can't tell for sure, of course, unless someone can find something definitive from Amazon.
Re: Totally true (Score:1)
what they promise is by "virtual core".
that has no L1/2/3 cache.
and will be balanced on the machine based on actual cpu cycles (my "real" 3.5ghz cpus can "turbo mode" to nearly 7ghz on heavy loads - Linux cpu-governor)
it's an extreme version of the old hdd spec lies.
http://tiemensfamily.com/TimOn... [tiemensfamily.com]
300...
compared with 10s of thousands for a "real" core.
Re: (Score:2)
I can't make much sense of your post, and the linked article is a bit confusing as well. But if the claim is that the 2 cores from an m1.large are about what you'd expect from 2009, I can totally believe that - the m1 instance is pretty old. I'd expect them to be sub-2-GHz Xeon cores. Xeon is usually damn slow compared to the consumer cores from the same year, and it's not sensible to compare Xeon performance to consumer performance (or, at least, it's Intel's brain damage, not Amazon's).
From the docs, m
Re: (Score:2)
I believe they use Xen.
Re: (Score:3)
That is not really possible. Amazon is getting to a monopoly position if you want to use all the modern Cloud stuff. Sure, you can get boxes from many different providers, but AWS has a ton of other services that you cannot buy from others and even more importantly, all the 3rd party Cloud services are running on AWS so they are faster if you are on AWS too.
I would not be surprised if in 10 years Amazon would be a verb for running server software like google is now for search.
Re: Quit whining (Score:1)
Re: Quit whining (Score:4, Interesting)
Re: (Score:3)
Re: (Score:2)
Microsoft survives for one reason: They are so well entrenched in the enterprise. Other than small companies that LDAP can work with, AD will be found as the core authentication and management mechanism of most companies out there.
Because of this, if MS sees losses on other fronts, they can just ratchet up Windows Server license fees, and still come out ahead, as they have a captive audience.
Re: (Score:2)
It's odd how much Azure seems to struggle to keep up with AWS. MS has no shortage of cash, and at the very least they could offer "MS SQL instances in the cloud" as cheaply as they needed to go get people locked in, It's not like NoSQL and message queues are rocket science these days, and the AWS versions have fairly minimal feature sets.
In the old days MS was all about lock-in, and while you never wanted to touch v1.0 of anything from MS, by the time 2.0 came out they were usually ahead of the pack in te
Re: (Score:1)
What? Have you even seen MS's offerings? They can't build DCs fast enough because so many people are signing up. Their stuff is at least in par with Amazon if not better.
This. All of this. More and more clients who traditionally went AWS (advertising campaign back ends, social media startups and so on) are picking Azure. The AWS tools are just crude in comparison and the Azure offerings are typically more complete, more robust and much better documented.
3-4 years ago suggest Azure hosting was suicide for a potential contract, these days it is an advantage and makes the guys pushing AWS look like they aren't keeping up. The perception is shifting fast.
Re: (Score:2)
(They certainly know how to operate at scale as good as anyone.)
Operating at scale when you're bespoke from the chip up is easy.
Well, relatively easy.
Operating at scale when any twat out there can do whatever the fuck they like to your commodity infrastructure (which must be relatively commodity or you can't sell it) is a very different proposition.
I'm confident Google could step into that market if they chose, but it's a deviation from their standard operating model.
Re:Quit whining (Score:5, Interesting)
Alternative: build your own private cloud out of the smallest servers you can find that still suit your need. I did so: a private cloud on HP Microserver (gen 8). The things consume almost no power when idle. Taken together, they provide quite the computing power ( 64 cores, Xeon E3 ) and quite the storage (32 TB). Cost me around € 450 in electric power per year. Am not dependent on Azure or Amazon. Use no bandwidth when doing cloudy things.
Re: (Score:3, Informative)
Alternative: build your own private cloud
The whole point of doing computing in the cloud is to handle variable demand. I spend 98% of my time writing and debugging code, and only 2% running the final model. But when I run it, I want to do it at scale. On AWS I can rent a dozen K80 GPUs for a few hours a week. There is no way it would make sense to own them, even if I could afford to do that.
Re: (Score:1)
Disagree. For a developer, the whole point of computing in the cloud is to be familiar with it. To that end, a private cloud makes eminent sense. And "at scale" then means: 8 servers (HP MicroServer Gen8).
Re: (Score:2)
We have multiple tier 5 data centres. Amazon/Azure is still useful for dynamic scaling for some of our workloads.
Some of the more optimised and specialised services can also really exploit cloud pricing. When part of our business can service millions of consumers for a few hundred dollars a year with Amazon we just can't get close to that price point in house.
But fuck it, The Good, The Bad and The Ugly just started on TV so I'm going to kick back, drink some vodka and enjoy cinematic perfection.
Re: Quit whining (Score:4, Insightful)
I'm not sure I'd call it whining - it's a significant investment for many to set up cloud computing the way they need it so I think it's a fair demand that if your gas station insists on measuring it's product in frackles that it give you a fair conversion rate for litres to frackles. My needs would be something predictable like rendering 3d images, so the first thing I'd be doing is measuring render times for the same image there and here. That's easy and relatively cheap to do, and the numbers scale directly, but I can see this being a serious issue for other uses.
Re: (Score:2)
But they not only measure in frackles but they also break your fuel gauge so you dont know how much they actually put in your car.
Re: (Score:1)
Except there's still a lot of potential market to capture. Microsoft and Google can charge less and get those people before they get locked in to Amazon.
Mostly irrelevant to most people (Score:4, Interesting)
Most people don't care about the exact performance, so it's not with spending the money and effort to precisely define or guarantee it.
Amazon is generic and cheap. Microsoft has really good integration with visual studio and .NET. Those are the factors people choose by.
Re: (Score:1)
Considering that most modern CPUs have performance counters and such, there is no excuse not to be able to measure it, particularly when you are charging for it. How big is this house you want to sell me? Well, somewhere between 100 and 300 sq. m., nobody cares, so why are you whining about it, am I right?
Re: Mostly irrelevant to most people (Score:2)
Re: Mostly irrelevant to most people (Score:5, Funny)
Re: Mostly irrelevant to most people (Score:3)
Re: Mostly irrelevant to most people (Score:2)
Re: Mostly irrelevant to most people (Score:2)
Today's complexity creates a lot of the trouble in getting reliable performance data. That makes perfect sense to me.
Re: Mostly irrelevant to most people (Score:2)
Re: Mostly irrelevant to most people (Score:3)
Re: Mostly irrelevant to most people (Score:2)
I found the fix. ...and it sucks. Apparently, Slashdot now carries over your desktop posting preferences to the mobile site. This didn't used to happen. My desktop was set to HTML codes posts. There is no way I'm imputing proper HTML code using a smartphone touchscreen keyboard!
So, to the fix. You can only do this on the desktop version of the site. There is no mobile accessible setting for this. Login to the desktop site, open any story, then reply to any post. Select options from the post window (not the
Re: Mostly irrelevant to most people (Score:2)
Hey, I've been staying analytical and being honest when I'm outside my skill set. The personal attack is uncalled for.
You don't need to take my word on it for benchmark cheating. A Google search will do. Here's an older one but, it went all the way to a class-action suit that Intel had to settle: Intel lawsuit [wccftech.com]. The most recent article about CPU cheating was 2009 though GPU cheating is still rampant in the news. On the CPU side they have either cleaned up their act enough to stay out of the news or just ha
Re: Mostly irrelevant to most people (Score:2)
Sorry. That's the hard part of having a long conversation with an AC.
Re: (Score:2)
Re: (Score:1)
There are different types of usefulness. Your post, for example, which attempts to impress the reader with your apparent technical knowledge is only public masturbation rather than adding anything further to the discussion. It has nothing to do with political correctness which was brought up by GP and addressed by myself. However if you wish to find the knowledge you seek I suggest other threads or failing that, perhaps another source than slashdot.
We're all specialists son. In fact I am willing to bet I
Re: (Score:2)
I'm not a specialist. I took a career choice to become a generalist, on the grounds it lets me get involved in solving problems with a broader scope and greater scale. I call this fun.
I have no post-graduate qualifications either. I got on with learning pragmatic useful shit instead.
Re: (Score:2)
Re: (Score:1)
Re: (Score:1)
I try to use the phrase "one's alcoholic uncle" which is less likely to be misconstrued as personal. :-)
Re: Mostly irrelevant to most people (Score:2)
Re: (Score:2)
This behavior is what I (and others) predicted when this whole "cloud" stupidity began: as providers become more entrenched, the prices and complexity slowly increase. Those who become dependent on this particular brand of idiocy will be taken to the cleaners.
Variable Units (Score:2)
what a surprise (Score:3)
So just takes the worst measurement you can find? (Score:2)
Imho there is only 1 valid measurement for cloud solutions, how much will certain performance cost you. And you can only really find out by testing it for _your_ specific goal. Some apps require more cpu, others more disk, others more ram... there's no single number to indicate your price/performance index.
If you need well defined performance... (Score:4, Insightful)
then buy dedicated instances.
I like micro instances/instances which do not occupy full physical processors at Amazon because of availability and price for low-impact/bandwidth applications. For all other use lambda or dedicated instances.
Virtual CPUs are anyway difficult to asses - to me it may be very relevant to have the 1st level cache of the core which i run on undisturbed by other applications (since changing the cache hits is a big deal for specific numerical problems), and for you 20% more share of the CPU may be important.
1% of computation time not spend in my task on a physical processor can do as much damage as 50% change in speed.
A small side remark: the price for the different VCPUs also varies.
Re: (Score:2)
This is exactly the problem, and solution.
From a computer science standpoint, modern algorithms that rely on single-thread speeds are obsolete and should be avoided for new development. From a sysadmin perspective, CPU speed is the least important metric compared to the rest of the system components. For project management, the system management toolset and feature support should be the prime concern.
Amazon's never really tried to hide the fact that their hardware is dissimilar behind the scenes. That will
welcome back to the mainframe (Score:1)
Re: (Score:1)
This is a scam by ISPs. They won't give you a high-bandwidth uploading network connection at your home or office. So you have to pay rent to run some slow computer at some remote facility.
It's not 1990 anymore... why can't you host a server at your own place.
Re: (Score:2, Insightful)
Re: (Score:2)
Seattle's zoning law is anti-utility. For acceptable Internet performance, leave Seattle.
Re: (Score:1)
where can't you get high speed connections? NYC and LA are up to 300mbps and higher for time warner. other markets as well.
The first commonly used modems were a *thousand* times faster than that. Connections here in the US are very slow, like the typical 1.5 Mbps T1 or DSL connections in the Seattle area for businesses, but the claim that it takes more than three seconds to transmit a single bit, is a lie.
Mbps vs. mbps spelling lame (Score:2)
Back when Usenet was popular, members of certain groups in alt. used to call that kind of post a "spelling flame" or later a "spelling lame". If you want to complain about mbps vs. Mbps, please also have something helpful to say about the rest of the post.
AWS (Score:2)
Microsoft makes it almost impossible to be license (Score:2)
Article is wrong on two things (Score:2)
Re: (Score:2)
"A virtual CPU is whatever Amazon wants to offer in an instance series."
No. The vCPU (Virtual CPU) aspect of an AWS EC2 Instance is the county of virtual cores that are exposed to an OS. In desktop computers, a quad core Intel CPU will appear to have four courses when looked at from inside the OS (my go-to way to count them in Linux is to run top and press 1). A quad core hyperthreaded Intel CPU will appear to have 8 cores. The vCPU metric simply tells you what the OS will show you, and tells you how many processor threa
history repeating again? (Score:1)
Deliberately obscure? (Score:1)