4 ms·

Hasn't cloud computing reached the point where local workstations like this aren't required?

by Dowwie 9y ago

drej 9y ago
They still are required, for at least three reasons - UX, cost, and security. 1. For a lot of workloads, it's more convenient to run in on a desktop. Sure, with the right tooling, you can bridge your cloud and desktop pretty seamlessly, but a lot of people can't do that or it's not possible with their tooling. Sure, it doesn't scale, but it often doesn't need to. 2. A Threadripper CPU may cost $1000, but but look at what that buys you in the cloud. The machine pays for itself within a year if used extensively. 3. Sometimes it's not easy to just upload code and data (!) to the cloud. Sure, there's all sorts of security layers and certifications and yes, clouds are very often more secure than desktops, but still, some people can't afford to let their data leave their machines. Two of these were a thing at my former employer (but we still used the cloud for other use cases). So no, I wouldn't say these workstations are obsolete.
fsloth 9y ago
No, it hasn't. I'm not saying we wont reach a stage at some point where compute power is a utility like electricity or water but we are definetly not there yet. I mean, in the least, I can't go to my local PC vendor and purchase a slow terminal with some compute credit allowance to nearest EC2 center or whatnot. When we get there, we can ask this question again.
badosu 9y ago
If your requirement is to have a predictable, responsive environment without having your assets intercepted then no.
dagw 9y ago
First of all no one has really successfully solved the running Autocad etc. in the cloud problem. Secondly the per core performance is generally pretty poor on cloud computers when compared to high end workstations and for many workloads 16 really fast cores beat out 64 slower cores. Finally, cloud computing is still really expensive compared to dedicated hardware. A halfway decent EC2 'workstation' like the p2.xlarge is over $5000 a year (plus storage) if you pay for the whole year up front and that 'only' gets you 4 cores and 60 GB of RAM (and a pretty good GPU). $5000 will buy you a really nice workstation with much better specs, and you get to keep it at the end of the year
- rjsw 9y ago
  Onshape provides CAD in the cloud, does AutoCAD need a lot of local CPU power ?
  - xemdetia 9y ago
    Yes, most engineering applications end up doing all sorts of simulation that can be summed up as 'run as much math as fast as possible.' There's just routing traces on a PCB, there can be heat or stress simulations on a part built with constructive geometry, and a lot of other things that lets an engineer simulate/try something interactively and not simply 'draw a part.' This is where good workstations can definitely make a difference. An example product that is definitely going to need the local beef: https://www.autodesk.com/products/cfd/overview https://www.autodesk.com/products/cfd/overview
robbyt 9y ago
Media creation such as video editing, music, 3d graphics. Plus many researchers like to iterate quickly before sending the work to a larger cluster.
mhw 9y ago
Hmm; for a workstation like the one I'm using now with 32Gb of RAM and 8 cores, EC2 has an m4.2xlarge as an equivalent. $0.4/hour * 24 hours * 365 days = $3,504 per year. You can build the same machine for a fraction of that cost and you'll get more than a year's use out of it. Reserved instances would bring the price to rent down a bit further, but I still don't think you can beat the economics of owning a workstation if you're going to be using it most of the time.
stephengillie 9y ago
Microsoft's desktop licence is more restrictive than their server license. Desktop OSes can't be virtualized alongside server OSes. This is why AWS Workspaces are Server 2008R2/2012 with the Windows 7/10 Experience, respectively. Then there's still latency, and the need for some kind of local client for the remote session. It's also a move away from "micro-computing" and back to the "mainframe-client" model that is so popular with Cloud-and-Phone.
j_s 9y ago
Ask HN: Do you use a Chromebook or Pixelbook as your primary machine? | https://news.ycombinator.com/item?id=15407631 https://news.ycombinator.com/item?id=15407631 (79+ comments right now)
synicalx 9y ago
Only for people who: - Don't care about latency. - Don't need bare metal performance. - Don't mind ongoing costs. - Either don't care, or don't need to care about their data sitting outside their organisation. - Have internet connections that can handle massive data transfers for months on end.
megaman22 9y ago
Running a small VM on Azure full-time for a year will buy a decent desktop. Running something equivalent to a $1000 desktop machine will cost wildly more.

Hasn't cloud computing reached the point where local workstations like this aren't required?

by Dowwie 9y ago

Hasn't cloud computing reached the point where local workstations like this aren't required?

drej 9y ago

They still are required, for at least three reasons - UX, cost, and security. 1. For a lot of workloads, it's more convenient to run in on a desktop. Sure, with the right tooling, you can bridge your cloud and desktop pretty seamlessly, but a lot of people can't do that or it's not possible with their tooling. Sure, it doesn't scale, but it often doesn't need to. 2. A Threadripper CPU may cost $1000, but but look at what that buys you in the cloud. The machine pays for itself within a year if used extensively. 3. Sometimes it's not easy to just upload code and data (!) to the cloud. Sure, there's all sorts of security layers and certifications and yes, clouds are very often more secure than desktops, but still, some people can't afford to let their data leave their machines. Two of these were a thing at my former employer (but we still used the cloud for other use cases). So no, I wouldn't say these workstations are obsolete.

fsloth 9y ago

No, it hasn't. I'm not saying we wont reach a stage at some point where compute power is a utility like electricity or water but we are definetly not there yet. I mean, in the least, I can't go to my local PC vendor and purchase a slow terminal with some compute credit allowance to nearest EC2 center or whatnot. When we get there, we can ask this question again.

badosu 9y ago

If your requirement is to have a predictable, responsive environment without having your assets intercepted then no.

dagw 9y ago

First of all no one has really successfully solved the running Autocad etc. in the cloud problem. Secondly the per core performance is generally pretty poor on cloud computers when compared to high end workstations and for many workloads 16 really fast cores beat out 64 slower cores. Finally, cloud computing is still really expensive compared to dedicated hardware. A halfway decent EC2 'workstation' like the p2.xlarge is over $5000 a year (plus storage) if you pay for the whole year up front and that 'only' gets you 4 cores and 60 GB of RAM (and a pretty good GPU). $5000 will buy you a really nice workstation with much better specs, and you get to keep it at the end of the year

rjsw 9y ago
Onshape provides CAD in the cloud, does AutoCAD need a lot of local CPU power ?
- xemdetia 9y ago
  Yes, most engineering applications end up doing all sorts of simulation that can be summed up as 'run as much math as fast as possible.' There's just routing traces on a PCB, there can be heat or stress simulations on a part built with constructive geometry, and a lot of other things that lets an engineer simulate/try something interactively and not simply 'draw a part.' This is where good workstations can definitely make a difference. An example product that is definitely going to need the local beef: https://www.autodesk.com/products/cfd/overview https://www.autodesk.com/products/cfd/overview

robbyt 9y ago

Media creation such as video editing, music, 3d graphics. Plus many researchers like to iterate quickly before sending the work to a larger cluster.

mhw 9y ago

Hmm; for a workstation like the one I'm using now with 32Gb of RAM and 8 cores, EC2 has an m4.2xlarge as an equivalent. $0.4/hour * 24 hours * 365 days = $3,504 per year. You can build the same machine for a fraction of that cost and you'll get more than a year's use out of it. Reserved instances would bring the price to rent down a bit further, but I still don't think you can beat the economics of owning a workstation if you're going to be using it most of the time.

stephengillie 9y ago

Microsoft's desktop licence is more restrictive than their server license. Desktop OSes can't be virtualized alongside server OSes. This is why AWS Workspaces are Server 2008R2/2012 with the Windows 7/10 Experience, respectively. Then there's still latency, and the need for some kind of local client for the remote session. It's also a move away from "micro-computing" and back to the "mainframe-client" model that is so popular with Cloud-and-Phone.

j_s 9y ago

Ask HN: Do you use a Chromebook or Pixelbook as your primary machine? | https://news.ycombinator.com/item?id=15407631 https://news.ycombinator.com/item?id=15407631 (79+ comments right now)

synicalx 9y ago

Only for people who: - Don't care about latency. - Don't need bare metal performance. - Don't mind ongoing costs. - Either don't care, or don't need to care about their data sitting outside their organisation. - Have internet connections that can handle massive data transfers for months on end.

megaman22 9y ago

Running a small VM on Azure full-time for a year will buy a decent desktop. Running something equivalent to a $1000 desktop machine will cost wildly more.