My lab just got a sponsorship from Microsoft Azure and I'm exploring how to utilize it. I'm new to industrial level cloud service and pretty confused about tons of terminologies and concepts. In short, here is my scenario:
I want to experiment the same algorithm with multiple datasets, aka data parallelism.
The algorithm is implemented with C++ on Linux (ubuntu 16.04). I made my best to use static linking, but still depends on some dynamic libraries. However these dynamic libraries can be easily installed by apt.
Each dataset is structured, means data (images, other files...) are organized with folders.
The idea system configuration would be a bunch of identical VMs and a shared file system. Then I can submit my job with 'qsub' from a script or something. Is there a way to do this on Azure?
I investigated the Batch Service, but having trouble installing dependencies after creating compute node. I also had trouble with storage. So far I only saw examples of using Batch with Blob storage, with is unstructured.
So are there any other services in Azure can meet my requirement?
I somehow figured it out my self based on the article: https://azure.microsoft.com/en-us/documentation/articles/virtual-machines-linux-classic-hpcpack-cluster/. Here is my solution:
Create a HPC Pack with a Windows head node and a set of Linux compute node. Here are several useful template in Marketplace.
From Head node, we can execute command inside Linux compute node, either inside HPC Cluster Manager, or using "clusrun" inside PowerShell. We can easily install dependencies via apt-get for computing node.
Create a File share inside one of the storage account. This can be mounted by all machines inside the cluster.
One glitch here is that for some encryption reason, you can not mount the File share on Linux machines outside the Azure. There are two solutions in my head: (1) mount the file share to Windows head node, and create file sharing from there, either by FTP or SSH. (2) create another Linux VM (as a bridge), mount the File share on that VM and use "scp" to communicate with it from outside. Since I'm not familiar with Windows, I adopted the later solution.
For executable, I simply uploaded the binary executable compiled on my local machine. Most dependencies are statically linked. There are still a few dynamic objects, though. I upload these dynamic object to the Azure and set LD_LIBRARY_PATH when execute programs on computing node.
Job submission is done in Windows head node. To make it more flexible, I wrote a python script, which writes XML files. The Job Manager can load these XML files to create a job. Here are some instructions: https://msdn.microsoft.com/en-us/library/hh560266(v=vs.85).aspx
I believe there should be more a elegant solution with Azure Batch Service, but so far my small cluster runs pretty well with HPC Pack. Hope this post can help somebody.
Azure files could provide you a shared file solution for your Ubuntu boxes - details are here:
https://azure.microsoft.com/en-us/documentation/articles/storage-how-to-use-files-linux/
Again depending on your requirement you can create a pseudo structure via Blob storage via the use of containers and then the "/" in the naming strategy of your blobs.
To David's point, whilst Batch is generally looked at for these kind of workloads it may not fit your solution. VM Scale Sets(https://azure.microsoft.com/en-us/documentation/articles/virtual-machine-scale-sets-overview/) would allow you to scale up your compute capacity either via load or schedule depending on your workload behavior.
Related
Our development environment has a bunch of virtual machines running different versions of our software. I want to be able to replace the Managed Image that is running on a VM, without having to destroy and recreate it.
The images are created using packer, which provisions them with the correct software and dependencies.
Example of Current Workflow:
Machine A is running on Managed Image v2.5, which runs software with a dependency on Tomcat 10.
To fix a bug in v2.2, which depends on Tomcat 9 and thus cannot run on the same VM without changing the dependencies, I have to:
Destroy the VM
Recreate it using the same arguments (name, size, etc) but based on Managed Image v2.2
Attach the network interface and disks
Restart it
If feel like there should be an easier solution to this, where it is possible to hot-swap the images, without recreating the full virtual machines. I've looked into swapping the OS disk, but I couldn't figure out a solution that would work with Managed Images instead of VHD's.
As per official article, its not Supported.
Microsoft does not support an upgrade of the operating system of a
Microsoft Azure virtual machine. Instead, you should create a new
Azure virtual machine that is running the supported version of the
operating system that is required and then migrate the workload.
Official article : https://support.microsoft.com/en-us/help/2721672/microsoft-server-software-support-for-microsoft-azure-virtual-machines
Instead you can use windows server migration tool that would assist you in roles and feature's migration
Install, use, and remove Windows Server migration tools
Similar issue discussed at Can you do in-place updating/upgrading of Azure VM Operating System?
We're using Managed VMs, and can currently serve files from the local disk in the VM (which is a standard magnetic HD), as well as serving from Google Cloud Storage (which is also backed by magnetic HDs).
https://cloud.google.com/appengine/docs/managed-vms/
As we're working with large files (high-res geo images) in a latency-sensitive context, we'd like to be able to use Local SSDs with our Managed VMs app (it's okay that the data is not persistent, it just needs to be fast and work with large files). At some point, we may want to use other services that are fast and designed for working with large files (e.g. Blobstore?), but we have a workflow already set up to work with files so it should be easiest to set up a faster file system now. Is it possible to use Local SSD storage with Managed VMs?
Here's info on Local SSDs. They need to be created at instance creation time (for Google Compute instances, which Managed VMs are creating behind the scenes). It looks like Local SSDs can be created via command line, gcloud compute, or an API, but it's not clear where we'd configure any of these things since Managed VMs is doing the instance creation for us. Presumably we'd do this in app.yaml, Dockerfile, or in a gcloud command, but it's not obvious how this would work.
https://cloud.google.com/compute/docs/disks/local-ssd
Apologies, this isn't currently available. We don't really let you customize the VM specs, aside from the disk size. If you max out the disk size, you will get the highest IOPS we support. We kicked around the idea of tmpfs, but it's not available just yet.
For my system, I have a back-end process that uses a 3rd party command line tool to do some occasional processing. This tool writes to and reads from the file system (I point it at some files, it works its magic, and then it writes out the results to another file).
This is obviously easy to do with an Azure Virtual Machine. Just write a Windows Service to manage this command line tool and have it read from a Queue to get the processing jobs.
To this point, however, I've been able to do everything in Azure without having to resort to a full blown VM. I like that. I like not having to worry about applying patches and other maintenance, downtime and the like.
So, my question is, is there something in Azure that would let me have this service without resorting to a VM? Would a "Worker Role" be able to accomplish this? Can it read and write to/from the file system? Can it handle 3rd party tools with a bunch of arbitrary dependencies? Can I launch another process from C# code within the worker role?
Would a "Worker Role" be able to accomplish this?
Absolutely! Remember that a Worker Role is a full blown VM also (with same OS powering Azure Virtual Machine).
Can it read and write to/from the file system?
Yes. However there's a catch. You can can't read/write to any arbitrary location on the VM. You would have full access to a special folder on that VM called Local Storage. You can read more about it here: http://msdn.microsoft.com/en-us/library/azure/ee758708.aspx
Can it handle 3rd party tools with a bunch of arbitrary dependencies?
Yes, again! Again, there's a catch. Since these VMs are stateless VMs, anything you install after the VM is stood up for you by Microsoft is not guaranteed to be there in case Microsoft decides to tear down that VM for whatever reasons. If you need to install any additional software, you would have to install them via a process called Startup Tasks. You can read about them here: http://msdn.microsoft.com/en-us/library/azure/hh180155.aspx.
Can I launch another process from C# code within the worker role?
Though I have not tried it personally but I think it is possible because you get a VM running latest version of Windows server.
i have an existing program that i would like to upload to the cloud without rewriting it and i'm wondering if that is possible.
For exemple can i upload and run a photoshop instance in the cloud and use it?
Of course not the GUI but photoshop has a communication sdk so web program should be able to control it!
As far as i can see, Worker roles looks good but they have to be written in a specific way and i can't rewrite photoshop !
Thanks for your attention!
As long as your existing program is 64bit compatible and it has installer that supports unattended/silent install; or your programm is xcopy deployable, you can use it in Azure.
For the programm that requires installation and supports unattended/silent install you can use StartUp Task.
For the program that is just xcopy deployable, just put it in a folder of your worker role, and make sure the "Copy to Output" attribute of all required files are set to "Copy always". Then you can use it.
However the bigger question is, what are you going to do with that "existing programm" in Azure, if you do not have API-s to work with.
Here's the thing, the Worker role should be what you need - it's essentially a virtual machine running a slightly different version of Windows, that you can RDP to, and use it normally. You can safely run more or less anything up there, but you need to automate the deployment (e.g. using startup tasks). As this can prove a bit problematic, Microsoft has created a Virtual machine Role. You create your own deployment and that's what gets raised when you instantiate the machine.
However! This machine is stateless, meaning that files it creates aren't saved if it gets restarted. So you need to ensure the files are saved somewhere else, e.g. in blob storage (intended for just such a purpose).
What I would do in your case, is create a virtual machine role, with Photoshop installed, and a custom piece of software next to it, accepting requests via Azure Queues, that does the processing, and saves the file to blob storage, then sends the file onwards to whoever requested
For the first time, I am writing a web service that will call upon external programs to process requests in batch. The front-end will accept file uploads and then place them in a queue. The workers on the backend will take that file, run it through ffmpeg and the rest of my pipeline, and send an email when the process is complete.
I have my backend process working on my computer (Ubuntu 10.04). The question is: should I try to re-create that pipeline using binaries that I've compiled from scratch? Or is it okay to use apt when configuring in The Real World?
Not all hosting services uses Ubuntu, and not all give me root access. (I haven't chosen a host yet.) However, they will let me upload binaries to execute, and many give me shell access with gcc.
Usually this would be a no-brainier and I'd compile it all from scratch. But doing so - not to mention trying to figure out how to create a platform-independent .tar.gz binary - will be quite a task which ultimately doesn't really help me ship my product.
Do you have any thoughts on the best way to set up my stack so that I'm not tied to a specific hosting provider? Should I try creating my own .deb, which contains Ubuntu's version of ffmpeg (and other tools) with the configurations I need?
Short of a setup where I manage my own servers/VMs (which may very well be what I have to do), how might I accomplish this?
The question is: should I try to re-create that pipeline using binaries that I've compiled from scratch? Or is it okay to use apt when configuring in The Real World?
It is in reverse: it is not okay to deploy unpackaged in The Real World IMHO
and not all give me root access
How would you be deploying a .deb without root access. Chroot jails?
But doing so - not to mention trying to figure out how to create a platform-independent .tar.gz binary - will be quite a task which ultimately doesn't really help me ship my product.
+1 You answer you own question. Don't meddle unless you have to.
Do you have any thoughts on the best way to set up my stack so that I'm not tied to a specific hosting provider?
Only depend on wellpackaged standard libs (such as ffmpeg). Otherwise include them in your own deployment. This problem isn't too hard too solve for 10s of thousand Linux applications over decades now, so it would probably be feasible for you too.
Out of the box:
Look at rightscale and other cloud providers/agents that have specialized images/tool chains especially for video encoding.
A 'regular' VPS provider (with Xen or Virtuozzo) will not normally be happy with these kinds of workload, but EC2, Rackspace and their lot will be absolutely fine with that.
In general, I wouldn't believe that a cloud infrastructure provider that doesn't grant root access will allow for computationally intensive workloads. $0.02