You're joking, right? Do you really want me to take these apart paragraph by paragraph or do you just want to find some actual study of substantive value, instead of click-bait editorial blogs?
My internet (and 99% of americans) aren't paying any more per byte. Some ISPs instituted caps, which the vaaaaaaaast majority of americans won't ever run in to.
Prices for goods typically go up, so does income, it's called inflation, something we have to always adjust for when we compare prices year to year.
With that said, I've been paying the same price for internet access for 8 years now. Bandwidth hasn't changed and prices haven't gone up. So, relative to cost of living, my cost has gone down.
Meanwhile we have Google Fiber, Verizon FiOS and Municipal Broadband projects rolling out all over the country. I can counter anecdotal isolated examples with the same in the other direction. Just playing devil's advocate here. Do I want 1Gb/s FTTH for $30/mo? Of course I do. Do I think it's possible to provide that service all over the US? Of course not. If you think it is, feel free to start an ISP and make so much money you can swim in it like Scrooge McDuck.
So, you can't provide any metric where we aren't improving. Just wanted to clarify that point.
Intel vs Broadband are completely different. There are lots of providers in the US as opposed to a single chip maker. When AMD came along we saw a massive innovation in x86 CPUs leading up to this point where AMD is beginning to fall behind again. In many places in the US we have several choices for broadband, basically any 1st through 3rd tier city (eg population say, 250k+). I wish our problem was that simple.
I certainly think we have our share of problems, but I think trying to compare the US to other nations is pointless. Unless you can find another nation with similar population densities and total populations, then saying someone is #1 and someone else is #2 is silly. What we should be worried about is our progress.
What I don't understand is, if it's so easy to deploy inexpensive [fiber] broadband then why aren't there tons of companies doing it? Do consumers just not care? Sure the slashdot crowd does, but does 99% of the nation need anything faster than broadband fast enough to stream netflix and youtube? How many people would pay more for a little more bandwidth? Would they even notice a difference in their daily computer use?
Or is it (gasp) maybe not as inexpensive and easy to deploy high speed nation wide broadband networks than the average slashdot user thinks?
Can you provide a metric by which broadband isn't improving in the US? I'm certainly willing to listen.
The US is a tough place because its geographically very large and it's very sparsely populated in some areas and very densely in others. We also have a very high per capita income, and don't subsidize broadband the way some other (mostly European) countries do. My only point in explaining this is to illustrate that it's extremely difficult to compare broadband between the US and other countries. In some areas of the US we do pretty well, in others, not so well. In rural america it's nearly impossible to deploy high speed broadband effectively because the cost per home passed is so large, because of the low population density.
Most med-large SAN use tiered storage. We're seeing SSDs in Tier0/Tier1. We will continue to see mechanical HDD in Tier 2 and near-line for a while, I believe.
Bingo. Reg/ECC RAM is about $10/GB retail. If we're talking about a few GB, just buy a bunch of RAM. Unless you plan on the dataset outgrowing it relatively quickly, then you can either shard/partition (depending on your platform) the DB or move to some type of disk based solution.
Throughput isn't necessarily the most important factor. In fact, with SATA 6 SSD, it hardly ever is. Most people are concerned with latency and IOPS. Even a single SATA SSD can easily do 500MB/s. Throw a few of them onto a PCIe SATA controller and you can easily get several GB/s of throughput, it's been proven time and time again. And usually at a fraction of the cost of the PCIe based SSDs (mostly because the controller tech is expensive, but MORE so that they typically use enterprise level SLC NAND which is far more expensive).
Depends on workload and performance requirements. If you have a very disk intensive workload that requires low latency and high throughput, but low storage capacity requirements, like DB workloads specifically (SQL, Mongo, etc) it can be DRAMATICALLY cheaper to use SSDs versus building massive disk based RAID arrays, and you'll never match the latency. A pair of Intel S3700 drives can outperform several shelves worth of 15K RPM SAS disks. Not to mention power, cooling, management, etc.
Again, all depends on workload and requirements. Making blanket statements like you did are just silly.
This guy basically nails it. SSD will slowly move it's way from "Tier 0" out, and eventually in 20 or 30 years even our near-line will be SSD, quite possibly.
That's probably too small of a sample to draw any reliable conclusion don't you think? Even if you had 1 SSD that lasted 20 years, does that really tell us anything, statistically?
For what it's worth, I bought my first SSD, a 30GB OCZ Vertex SSD (original version) on 6/21/2009 (i just logged into newegg and checked) and it's still going strong without a single problem. It's since been "demoted" to my HTPC in the living room, which has been great because the bootup is very "appliance-like" and it's completely silent.
Oh and if you were curious - I paid $139.00 for it in June 2009. For comparison, today you can get a 128GB OCZ Vertex 4 for $119. So 15% cheaper for 2.5x the performance and 4x the capacity in 4 years. That's what I call progress!
There are lots of very large installations using pure SSD (MySpace went all SSD in 2009 for example). However, no one seems to be making the data available. And one reason it wouldn't help is that the lifespan of an SSD is incredibly dependent on the work load, unlike traditional disks. If you're workload is 99% reads and 1% writes, your failure rate would be exceptionally low. But if my workload was 50/50 reads/writes my failure rate COULD BE substantially higher than yours.
I think location is the big consideration. Cost of living. You may very well be able to do the exact same work in another location and make $200k/year. But your cost of living might double as well. That's why specific numbers like $200k don't really matter. I make six figures but I live in an area with an extremely low cost of living. Using a cost of living calculator, I would have to make $250k to live in NYC with the same standard of living.
Another more general way to state this as it applies to most fields: Learn how to apply your programming skill in a way that directly generates income for your employer. If you can prove you make them many times more than you cost, you're going to be well compensated.
Slashdot Mirror
I really wanted to install this on our internal wiki that we use in IT, but then I saw the node.js requirement. What the fuck, man?
Actually laughed out loud at this one. Why do I never have moderator points when I actually need them?
That should be easy enough to fix, you just need a blank word and excel document you can make copies of.
That doesn't make them pro-Israel. That makes them "pro-the-highest-bidder".
You're joking, right? Do you really want me to take these apart paragraph by paragraph or do you just want to find some actual study of substantive value, instead of click-bait editorial blogs?
My internet (and 99% of americans) aren't paying any more per byte. Some ISPs instituted caps, which the vaaaaaaaast majority of americans won't ever run in to.
Prices for goods typically go up, so does income, it's called inflation, something we have to always adjust for when we compare prices year to year.
With that said, I've been paying the same price for internet access for 8 years now. Bandwidth hasn't changed and prices haven't gone up. So, relative to cost of living, my cost has gone down.
Meanwhile we have Google Fiber, Verizon FiOS and Municipal Broadband projects rolling out all over the country. I can counter anecdotal isolated examples with the same in the other direction. Just playing devil's advocate here. Do I want 1Gb/s FTTH for $30/mo? Of course I do. Do I think it's possible to provide that service all over the US? Of course not. If you think it is, feel free to start an ISP and make so much money you can swim in it like Scrooge McDuck.
So, you can't provide any metric where we aren't improving. Just wanted to clarify that point.
Intel vs Broadband are completely different. There are lots of providers in the US as opposed to a single chip maker. When AMD came along we saw a massive innovation in x86 CPUs leading up to this point where AMD is beginning to fall behind again. In many places in the US we have several choices for broadband, basically any 1st through 3rd tier city (eg population say, 250k+). I wish our problem was that simple.
I certainly think we have our share of problems, but I think trying to compare the US to other nations is pointless. Unless you can find another nation with similar population densities and total populations, then saying someone is #1 and someone else is #2 is silly. What we should be worried about is our progress.
What I don't understand is, if it's so easy to deploy inexpensive [fiber] broadband then why aren't there tons of companies doing it? Do consumers just not care? Sure the slashdot crowd does, but does 99% of the nation need anything faster than broadband fast enough to stream netflix and youtube? How many people would pay more for a little more bandwidth? Would they even notice a difference in their daily computer use?
Or is it (gasp) maybe not as inexpensive and easy to deploy high speed nation wide broadband networks than the average slashdot user thinks?
Maybe you shouldn't have had children?
Can you provide a metric by which broadband isn't improving in the US? I'm certainly willing to listen.
The US is a tough place because its geographically very large and it's very sparsely populated in some areas and very densely in others. We also have a very high per capita income, and don't subsidize broadband the way some other (mostly European) countries do. My only point in explaining this is to illustrate that it's extremely difficult to compare broadband between the US and other countries. In some areas of the US we do pretty well, in others, not so well. In rural america it's nearly impossible to deploy high speed broadband effectively because the cost per home passed is so large, because of the low population density.
That has nothing to do with engineering at Intel or AMD.
Most med-large SAN use tiered storage. We're seeing SSDs in Tier0/Tier1. We will continue to see mechanical HDD in Tier 2 and near-line for a while, I believe.
Bingo. Reg/ECC RAM is about $10/GB retail. If we're talking about a few GB, just buy a bunch of RAM. Unless you plan on the dataset outgrowing it relatively quickly, then you can either shard/partition (depending on your platform) the DB or move to some type of disk based solution.
Especially for very large virtual workstation installations. Those workloads just absolutely wreak havoc on disks.
Throughput isn't necessarily the most important factor. In fact, with SATA 6 SSD, it hardly ever is. Most people are concerned with latency and IOPS. Even a single SATA SSD can easily do 500MB/s. Throw a few of them onto a PCIe SATA controller and you can easily get several GB/s of throughput, it's been proven time and time again. And usually at a fraction of the cost of the PCIe based SSDs (mostly because the controller tech is expensive, but MORE so that they typically use enterprise level SLC NAND which is far more expensive).
Depends on workload and performance requirements. If you have a very disk intensive workload that requires low latency and high throughput, but low storage capacity requirements, like DB workloads specifically (SQL, Mongo, etc) it can be DRAMATICALLY cheaper to use SSDs versus building massive disk based RAID arrays, and you'll never match the latency. A pair of Intel S3700 drives can outperform several shelves worth of 15K RPM SAS disks. Not to mention power, cooling, management, etc.
Again, all depends on workload and requirements. Making blanket statements like you did are just silly.
This guy basically nails it. SSD will slowly move it's way from "Tier 0" out, and eventually in 20 or 30 years even our near-line will be SSD, quite possibly.
Two possibilities:
1. You're lying
2. You were using consumer level MLC drives in high write workloads (I might believe this)
Also - thanks for the info, very interesting and honestly what I would have suspected. Nice to see it play out in the real world.
When you say it outperforms a RAID10 array of 15K RPM disks - how MANY disks? 4? 100?
That's probably too small of a sample to draw any reliable conclusion don't you think? Even if you had 1 SSD that lasted 20 years, does that really tell us anything, statistically?
For what it's worth, I bought my first SSD, a 30GB OCZ Vertex SSD (original version) on 6/21/2009 (i just logged into newegg and checked) and it's still going strong without a single problem. It's since been "demoted" to my HTPC in the living room, which has been great because the bootup is very "appliance-like" and it's completely silent.
Oh and if you were curious - I paid $139.00 for it in June 2009. For comparison, today you can get a 128GB OCZ Vertex 4 for $119. So 15% cheaper for 2.5x the performance and 4x the capacity in 4 years. That's what I call progress!
There are lots of very large installations using pure SSD (MySpace went all SSD in 2009 for example). However, no one seems to be making the data available. And one reason it wouldn't help is that the lifespan of an SSD is incredibly dependent on the work load, unlike traditional disks. If you're workload is 99% reads and 1% writes, your failure rate would be exceptionally low. But if my workload was 50/50 reads/writes my failure rate COULD BE substantially higher than yours.
I think you mean PCIe interface?
I think location is the big consideration. Cost of living. You may very well be able to do the exact same work in another location and make $200k/year. But your cost of living might double as well. That's why specific numbers like $200k don't really matter. I make six figures but I live in an area with an extremely low cost of living. Using a cost of living calculator, I would have to make $250k to live in NYC with the same standard of living.
I'm absolutely amazed that 10% of developers make that much. I would have guessed 1%. If that.
Another more general way to state this as it applies to most fields: Learn how to apply your programming skill in a way that directly generates income for your employer. If you can prove you make them many times more than you cost, you're going to be well compensated.