# Larry Land Larry Gadea's random ramblings This is one page of public article previews, not the complete archive. Follow Next page to continue. Summaries are not the original full articles. ## Playing Overwatch on Azure NVIDIA M60 GPU instances DevFeed: [Playing Overwatch on Azure NVIDIA M60 GPU instances]() Original publisher: [Read original article]() Published: 2016-10-12T07:00:00Z Content type: tutorial Language: en Sources: [Larry Land]() Topics: [Azure](), [GPU](), [Cloud](), [Streaming](), [Low Latency](), [Amazon EC2]() Tags: [amazon](), [azure](), [cloud](), [gaming](), [gpu](), [low-latency](), [streaming]() ### AI overview A practical guide to launching an Azure GPU instance, configuring low-latency cloud streaming, and playing Overwatch remotely. It discusses NVIDIA M60 GPUs, image quality, latency, client platforms, and bandwidth costs. ### Source excerpt Updated Dec 17, 2016: Removed mention of the Azure Preview (since it's now GA), linked to easier method to disable monitors, updated driver link Playing Overwatch at 60FPS, 2560x1600, everything on Epic quality, and streaming from the cloud - not too shabby! It's no secret that I love the concept of not just streaming AAA game titles from the cloud, but playing them live from any computer - especially on the underpowered laptops I usually use for work. I've done it before using Amazon's EC2 (and written a full article for how to do it), but this time, things are a little different. Microsoft's Azure was first to give access to NVIDIA's new M60 GPUs, completely new beasts that really set a whole new bar for framerate and image quality. They're based on the newer Maxwell architecture, versus the Kepler cards we've used in the past. Hopefully one day we'll get the fancy new Pascal cards :) Before going through this article, I strongly recommend you at least skim my EC2 Gaming article from before so you can grasp some of the concepts we'll be doing here. Basically it'll come down to this: we're going to launch an Azure GPU instance, configure it for ultra-low latency streaming, and actually properly play Overwatch, a first-person shooter, from a server over a thousand miles away! And yes, it seems I always need to repeat myself when writing these articles: the latency is just fine, the resolution is amazing, it can be very cost-effective (if you don't forget the machine on), and all very practical for those of you obsessed about minimalism (like me). If you're the hardcore gamer-type who's knee-jerk reaction regardless of anything is "omg i dont trust anything that doesnt have a Zalman watercooled CPU fan or has a 'hardware-offloaded' networking card" or if you're in some sort of sunk-costs situation of already having spent loads of money on a gaming rig, I strongly recommend not reading this article - it'll only further infuriate you. :) Note that in this article I ass ## Using Lets Encrypt to secure cloud-hosted services like Ubiquiti's mFi, Unifi and Unifi Video DevFeed: [Using Lets Encrypt to secure cloud-hosted services like Ubiquiti's mFi, Unifi and Unifi Video]() Original publisher: [Read original article]() Published: 2015-12-13T08:00:00Z Content type: tutorial Language: en Sources: [Larry Land]() Topics: [SSL](), [Encryption](), [Ubiquiti](), [Amazon EC2]() Tags: [ec2](), [encryption](), [ssl](), [ubiquiti](), [unifi](), [video]() ### AI overview A tutorial explains how to use Let's Encrypt to obtain and renew free SSL certificates for Ubiquiti mFi, UniFi, and UniFi Video controller services hosted on Amazon EC2. It also notes that the mFi product line has been discontinued. ### Source excerpt Updated Nov 27, 2018: Updated credentials for new unifi versions (uses a new username) Updated Jul 31, 2016: Moved away from letsencrypt-auto and switched to certbot, updated the auto-renewal script, and changed the suggested cron time to weekly. Also made mention that mFi series has been discontinued. Finally, fixed the install instructions for Unifi Video. Wow - I got myself a free signed SSL cert for my WiFi controller! Lets Encrypt recently was released and is definitely super interesting. They basically issue SSL Certificates for free. SSL Certs typically would cost hundreds of dollars per domain and even more for Wildcard certificates. It's insane, it's essentially an entire industry predecated around artificial pricing for something that is essentially zero cost to generate and maintain. Not to mention holding back security and encryption on the web since not just anyone can afford hundreds of dollars a year for a cert. This entire industry is holding back progress at a massive scale, so we're going to fix that :) With Lets Encrypt, this is all free now. As cost is no longer a problem, we can encrypt other communication like router config landing pages and other services. No need for self-signed certificates that your browser freaks out about when navigating to. Now we can have real certs! As I'm a big fan of Ubiquiti products, I'm going to show some examples in this article for how to use Lets Encrypt to generate certificates that are compatible with the mFi automation stuff, Ubiquiti's Unifi wifi controller and their Unifi Video series for surveillance. Ubiquiti, as they're an enterprise company, [imo wrongly] expects companies to want to host the backing controller software for these devices on-site. We're going to host them on EC2 though, so we don't need to manage servers or have people tripping over power cables. Since we're on the internet though, we need proper SSL to prevent the NSA and all their shenanigans. Ubiquiti's mFi, Unifi wireless and Unifi ## Real-world benchmarking of cloud storage providers: Amazon S3, Google Cloud Storage, and Azure Blob Storage DevFeed: [Real-world benchmarking of cloud storage providers: Amazon S3, Google Cloud Storage, and Azure Blob Storage]() Original publisher: [Read original article]() Published: 2015-10-25T07:00:00Z Content type: comparison Language: en Sources: [Larry Land]() Topics: [benchmarking](), [Cloud](), [Amazon S3](), [Azure](), [Google Cloud Platform (GCP)]() Tags: [azure](), [benchmarking](), [cloud](), [google-cloud](), [s3]() ### AI overview This article compares the performance of Amazon S3, Google Cloud Storage, and Azure Blob Storage using real-world tests involving large files, small files, and many file operations. The supplied text describes the motivation and testing methodology but does not include the results. ### Source excerpt I really wouldn't mind moving from Dropbox to S3, Google Cloud Storage, Azure Storage, or any other provider Recently at work I've been using the fancy new 4K iMac, yet when I'm at home or out-and-about, I use my equivalently-sexy MacBook. I do the same stuff on both computers though, it just so happens that one is way faster and has a bigger screen, and the other is slower though portable and much more practical for my lifestyle. They both need access to the same files. Typically to solve this problem you'd use Dropbox or a similar service. But one thing I strongly dislike about specifically Dropbox is that it really wants you to sync everything to every computer. To me this seems extremely wasteful as I might not want the storage overhead of syncing my personal photos onto my work computer. Ideally files could be uploaded to a service, but only downloaded on demand. I understand Dropbox has Selective Sync, but that's not what I want. I should be able to mount all my stuff onto a Raspberry Pi for example. What got me curious though is: which of the major cloud services out there actually focused on performance? I have 100mbit+ symmetric connections at both home and work (thx MonkeyBrains and Mimosa! pic here), so I figured that I should optimize for low-local-storage requirements yet be ok with high-bandwidth usage. We'll be testing Amazon AWS S3 Google Cloud Storage Microsoft Windows Azure Blob Storage I would have wanted to also test Backblaze's B2 thingy, but I haven't been invited into their Beta program yet. ;( Testing methodologies For all three services I want to test 3 real-world things: Large file: I created a 100mb file using mkfile 100m ~/Desktop/100m. Small files: I found a git repo on my machine and copied the .git directory. Many instructions: After the above git directory was synced, I removed all the files from the cloud provider. Also, the point of this is that it'll be used from the consumer side. So this was all run on my own laptop and to the se ## How to run a cloud gaming service on Amazon EC2 DevFeed: [How to run a cloud gaming service on Amazon EC2]() Original publisher: [Read original article]() Published: 2015-07-05T07:00:00Z Content type: tutorial Language: en Sources: [Larry Land]() Topics: [Amazon EC2](), [Cloud Gaming](), [Amazon Web Services](), [Streaming](), [GPU]() Tags: [aws](), [cloud-gaming](), [cost](), [ec2](), [gaming](), [gpu](), [linux](), [streaming](), [windows]() ### AI overview A step-by-step guide to running games on an Amazon EC2 GPU instance and streaming them to a Mac. It covers graphics encoding, SSD storage, OpenVPN, AMIs, Spot instances, and estimated operating costs, with notes about internet requirements and client compatibility. ### Source excerpt Playing Witcher 3, a GPU-intensive game on a 2015 fanless Macbook I've written about using EC2 as a gaming rig in the past. After spending some time and getting all sorts of feedback from many people, I'm re-writing the article from before, except with all the latest and greatest optimizations to really make things better. Now we're using things like NvFBC for graphics card H.264 encoding, using the built-in SSD for better hard drive performance, plus getting rid of things like VNC. I've also made the OpenVPN instructions easier to follow. This is the perfect solution for you fanatics that love to play AAA games but are stuck with one of the new fan-less Macbooks (or similarly slow machines). This is a pretty awesome alternative to building out a new gaming PC. Just make sure you have a good internet connection (ideally 30mbit+ plus <50ms ping to the closest Amazon datacenter). This article will assume you're on a Mac client, though it should work on Linux or Windows with some minor changes in the client tools. TLDR: I have made an AMI for you to use such that you don't need to go through this lengthy step-by-step. Note you become less of a badass if you use the AMI instead of doing it yourself ;). See instructions below. Costs Believe it or not, it's actually not that expensive to play games this way. Although you could potentially save moneys by streaming all your games, cost savings isn't really the primary purpose. Craziness is, of course. :) $0.11/hr Spot instance of a g2.2xlarge $0.41/hr Streaming 10mbit at $0.09/GB $0.01/hr EBS storage for 35GB OS drive ($3.50/mo) You're looking at $0.53/hr to play games this way. Not too bad. That's around 1850 hours of gaming for the cost of a $1000 gaming PC. Note that prices vary for different datacenters. Creating your own AMI with the right config On AWS, create a new EC2 instance. Use defaults everywhere except as mentioned below: Base image should be Microsoft Windows Server 2012 R2 Base (since Windows still has all t ## Using AWS Lambda as a backend for a single-page application DevFeed: [Using AWS Lambda as a backend for a single-page application]() Original publisher: [Read original article]() Published: 2015-05-16T07:00:00Z Content type: tutorial Language: en Sources: [Larry Land]() Topics: [AWS Lambda](), [Back end](), [Single-page application (SPA)](), [Amazon EC2](), [Heroku]() Tags: [aws](), [aws-lambda](), [back-end](), [backend](), [ec2](), [handler](), [job](), [json](), [lambda](), [policy]() ### AI overview The article demonstrates using AWS Lambda to run backend logic for a single-page application through JSON requests, without provisioning traditional server instances. It also discusses restricting AWS credentials with a dedicated user policy and notes a 100-parallel-request limit at the time of writing. ### Source excerpt It's a bit of a sensationalistic title, but hear me out. And sorry for the boring-ish example in this article. This is designed to only excite people who really get it. :) AWS Lambda lets you run NodeJS code or any binary (or code of any language) as a fast starting and stopping EC2 instance. You can use it to, for example, quickly resize an image as a background worker. What I'll quickly show you here though is that you could even possibly use it as a full-on back-end for a single-page app. What's even more awesome? It basically scales by parallelizing infinitely. No need to ever spin up more EC2 instances or Dynos or any of that. Capacity planning? No need! So, lets build a quick static webpage (that you can host anywhere, including file://) All this page does is connect to Lambda and send sup1 as the value to key1 through simple JSON. Here's the lambda code (which you literally can do on the AWS Console): This is even simpler. The handler callback is called by Lambda whenever this Lambda Function is called. context.succeed is used to return a value to the caller. And lets run it: FYI, this is crazy. We just did a backend call with full-on backend logic without setting up any kind of instances. No worrying about starting a new not-even-free-anymore Heroku Dyno. No choosing machine speed and memory and hard drive sizes. It just works, effortlessly, and scales enormously. One thing you might note is that we're exposing our AWS public and secret keys. This is actually not that big of a deal. To make this secure, we actually created a new AWS User and with a Policy specifically preventing everything except the one Lambda Function: Oh and I mentioned that it scales "infinitely", yeah there's a 100-parallel-requests maximum right now, but I'd imagine this will be removed by AWS as the service matures. Meanwhile, you can ask them for more. I guess recent AWS changes makes my lambda-job project kinda useless. It's all built in now. Very cool stuff. Hat tip to Amazon for c ## Run your own high-end cloud gaming service on EC2 DevFeed: [Run your own high-end cloud gaming service on EC2]() Original publisher: [Read original article]() Published: 2015-04-12T07:00:00Z Content type: tutorial Language: en Sources: [Larry Land]() Topics: [Amazon EC2](), [Cloud Gaming](), [Amazon Web Services]() Tags: [aws](), [cloud-gaming](), [ec2](), [gaming](), [gpu](), [ssd]() ### AI overview A tutorial on using an AWS EC2 GPU instance with Steam In-Home Streaming to play high-end games remotely over the Internet. It discusses estimated instance, data transfer, and storage costs, latency and bandwidth requirements, and suggested EC2 setup options. ### Source excerpt NOTE: This article has been superseded by my more recent one here How to use EC2 GPU machines + Steam In-Home Streaming + a VPN to play AAA titles on a shitty laptop You might have tried a service like the now defunct OnLive. Though personally I've played and beat many AAA games on the service, it unfortunately a) had a very limited selection and b) is now gone. I also have a bunch of games on Steam that I've played using my eGPU. With the new Macbook though, I won't be able to continue my low-end-laptop but high-end gaming extravaganza since there's no Thunderbolt. So why am I not concerned? Steam recently introduced In-Home Streaming, which basically creates a mini-OnLive in your own home with all the same Steam games I played with my eGPU. But... let's do it over the Internet! Cost Playing games this way is actually quite economical - especially when comparing to purchasing a full-on gaming rig. Here are the costs you'll need to consider: GPU Instance runs about $0.11/hr (on a Spot instance, regularly around $0.70/hr) Data transfer will be around $0.09/GB, and at a sustained ~10mbit, itll cost you $0.41/hr (4.5GB/hr) Updated Apr 25: Hard drive (EBS General Purpose SSD) storage of 100GB is $12.00/mo, or a bit under $0.02/hr This comes out to around $0.54/hr (Updated Apr 25), not bad, for the cost of a $1000 gaming pc, you get ~1900 hours on much higher-end hardware! The catch? This is all fun and games, but you need to make sure of two things: You are within 40ms to the closest AWS datacenter (test here) and has GPU instances (Updated Apr 25 brought this up from 20ms, many people reporting it's just fine at slightly higher latencies) You have at least a 10mbit connection and it's unmetered Setting it up On AWS, create a new EC2 instance. Use these settings: Base image should be Microsoft Windows Server 2012 R2 Base (since Windows still has all the best games) Use a g2.2xlarge instance (to get an NVIDIA GRID K520 graphics card). Updated Apr 19: There is no point usin ## The Ubiquiti EdgeRouter: Configuring this extremely low-cost, enterprise-grade router for home use DevFeed: [The Ubiquiti EdgeRouter: Configuring this extremely low-cost, enterprise-grade router for home use]() Original publisher: [Read original article]() Published: 2015-01-11T08:00:00Z Content type: tutorial Language: en Sources: [Larry Land]() Topics: [Ubiquiti](), [networking](), [Command-line interface](), [ssh](), [OpenSSH](), [configuration]() Tags: [cli](), [configuration](), [cost](), [enterprise](), [firmware](), [ip](), [networking](), [router](), [ssh](), [ubiquiti]() ### AI overview A tutorial for configuring the Ubiquiti EdgeMAX EdgeRouter Lite for home use, with related guidance for the EdgeRouter-X. It covers command-line administration, initial setup, firmware, SSH access, port forwarding, dynamic DNS, IPv6, and NAT-PMP. ### Source excerpt Updated Jan 17, 2015: Moved the dynamic DNS away from a scheduled task to the new custom- service method Updated Apr 3, 2016: Added dynamic DNS instructions for iwantmyname Updated Jul 24, 2016: Added NAT-PMP for UPnP, up-to-date dynamic DNS methods, IPv6 instructions, and EdgeRouter-X mention I've gotten a new, inexplicable, love for Ubiquiti. They fit in my favorite category of companies: they make high quality products that cost nothing compared to the old-boys-club equivalents. Oh and these products look spectacular. I'm actually quite surprised how UBNT is able to do all the things they do... I couldn't support them more. After finding out about them from MonkeyBrains, I looked through their product line to be quite surprised at the feature sets, yet still low cost. Being a huge fan of networking equipment, I decided to buy their cheapest router (at the time), the EdgeMAX EdgeRouter Lite. A 3 port, gigabit-capable router, that can really only be configured by commandline. Today, they have an even cheaper version of it, for $50! The EdgeRouter-X. For this article, we're going to configure the EdgeRouter Lite for home-use. Instructions should be similar if not identical for the EdgeRouter-X. A Challenger Appears! and it's only $100! Getting going First off, like I mentioned, this thing is only really configurable via commandline. It uses a forked version of the opensourced edition of Vyatta 6.3 (now maintained as VyOS post their acquisition by Brocade). So to learn about how this thing works, you'll need to read pages upon pages of documentation from the old Vyatta docs keeping in mind the Ubnt has modified a bunch of stuff too. A lot of help is available on their community page. Oddly enough, they don't document anything they do, so you're stuck reading release notes and harassing the employees that troll the forums. A great way to get started is to read the EdgeOS CLI Primer and the Basic System [PDF] docs. When you buy and receive yours, plug into Port 0, assign ## GTX 570 eGPU on a 2013 11" Macbook Air DevFeed: [GTX 570 eGPU on a 2013 11" Macbook Air]() Original publisher: [Read original article]() Published: 2013-07-28T07:00:00Z Content type: tutorial Language: en Sources: [Larry Land]() Topics: [GPU](), [intel](), [Windows]() Tags: [article](), [gpu](), [graphics](), [how-to](), [intel](), [windows]() ### AI overview A step-by-step guide to connecting an external GTX 570 graphics card to a 2013 11-inch MacBook Air through Thunderbolt and ExpressCard adapters. It describes the required hardware, Windows 7 setup, and reported graphics performance gains of approximately five to seven times. ### Source excerpt Note: This is a repost of my TechInferno Forum article from July 28th, 2013. Methods have changed and are somewhat easier these days. Make sure to check out that forum for more info. TLDR: By buying around $250 in commonly available parts, plus a video card, you can make the graphics of your 11" Macbook Air from 5X to 7X faster. Demo video at end of post. Step-by-step, here's how to exactly do it. Warning: not for the faint of heart! Hey everyone! This is my third article here on this forum, though it's the first that the process can be done by anyone with off-the-shelf parts. No more discontinued exotic parts like the $180 BPlus TH05 are required. All you need is a macbook air, a graphics card, a power supply, Windows 7, and ~$250 to buy some adapters and software online. All these parts are readily available for anyone. Like usual, I really want to thank nando4 for his help in doing all this. He's the mastermind behind the technicals, I just like writing articles and making stuff easier for everyone. He's super dedicated and eGPUs wouldn't be anywhere near where they are today if it wasn't for him! Thanks! So what are we doing? We're going to make a Macbook Air accept an external video card via Thunderbolt! Yes, you might have read in the news that real commercial solutions are just around the corner. We've been promised by these companies over-and-over again, with youtube videos, hands-on reviews, press releases, etc, but nobody is releasing anything. It's been like this for over a year. Intel even openly admits its bias against GPU usage where it's listed as unsupported in their Thunderbolt Certification Application. Talking to one of their thunderbolt guys, here's what that "Not Supported" means: The "Not supported" means that Intel won't neither certify your product nor deliver, at the moment, any Technology License for this kind of usage. As you know, this Technology License is required to develop a Thunderbolt device in the market and Certification is a must