TheGuyWhoCodes

You didn't define small or medium size and number of messages. From my experience server stream while great for memory and giving the client time to process messages as they come can be slower than unary. One way to solve that is to have a batched message with a repeated field of the underline message you want to send, it's an order faster.

I read sometime ago about about server pushback, maybe you need to override/implement your own StreamObserver to notify the client to slow down as the server isn't ready yet or configure the amount of data the gRPC server can queue there are quite a few configurations (a lot of them obscured) you can make to the server/grpc service.

S3 is a great service but it's a double edge sword. It's easy to start with and "cheap" until you scale and then you come to realize you pay a ton for writes and you have a latency of 100ms+ to retrieve data. Even S3 select which on face value seems great is absurdly slow. Then they offer you Fsx or elastic cache as a proxy.

This isn't really specific to tar. For example if you save data to parquet, sorting the data before persisting it can make a huge difference in the size of the file AND the predicate pushdown performance (as you can skip more pages especially if you sort by the column you filter on)

I remember big corps betting on phonegap and sencha for mobile 7 years ago. I also remember big corps betting on GWT.

Those were painful days...

I find it puzzling that aws uses rust so much yet don't have an official rust aws sdk

I genuinely would like to know what's the issue with a post about a major release of an innovative project. What's political about it? It's not like it's a new release of an enterprise/paid product based on an open source project.

Thanks, I looked at this way back before v1 and had a lot of issues. I guess I need to retest it.

Most of the models I use can be converted to a more optimized formats for inference like https://treelite.readthedocs.io/en/latest/ so the code and interfaces are pretty similar and it makes the architecture less complex to update and deploy.

It really depends on your API clients and your latency. Celery is fine for long running or batching, the queue and the fact you need to store and retrieve the results from somewhere else isn't really ideal. Non blocking events loops (and reactive requests) really like small tasks and even scream at you if you are above their thresholds.

Honestly I see a bright future for apache arrow flight but I think It's a bit immature right now.

But my main points is that 80% of the blog posts about FastAPI are for machine learning and people will just copy the code, run one or 2 requests see that it's fine and move along...

I looked at FastAPI while back for an API for a machine learning inference service (and you'd see a lot of blog/medium posts for that).

The syntax and documentation is really good and the docker examples are great for fast startup.

The only problem is that because FastAPI isn't process based some machine learning libraries (Tensorflow for example) don't really play nice and you'll get random errors or funky results, especially under load.

Obviously one solution is to use Tensorflow severing (and deal with that craziness) and let the FastAPI do the routing/data processing but I bet there are production machine learning products using FastAPI spawning random numbers and the developers are oblivious to the problem.

"Solved" is a strong word.

People suddenly had to pay licenses for at least 7 years with fines and interest that shot up the "price" by a 1000% The public television involved lawyers as collection agencies and threatened people with writ of execution.

Good riddance.

Dell made so much noise about their cooling chamber for the new XPS but forgot to actually do QA. They forgot to provide a working power brick. It didn't provide the advertised wattage (100 not 130) and even if it did actual workstations go with 230W. The trackpad was also broken for most people.

It's worse than that. I was hacking a turret and slowly picking them off one by one while they "approached" me (all of them immediately knew where I was even when I was in cover) and never even engaged with the turret in the middle of the camp blasting them off.

I'm not saying it's all bad. The art style is amazing, the sound track is awesome. The lips sync, facial expressions and body movement are pretty damn good. Yes there some nice side missions and the main story line is ok I guess but it's not really cyberpunk or really engaging. They advertised it as a cyber punk open world RPG and promised so many things and then literally changed the marketing of the game to an action shooter after the game came out...

Having played cyberpunk on PC it's mostly a fancy light show and not a lot of substance. You can't even customize anything past the creation screen, no car paint jobs no hair cuts and no body modification, no transmog system..

They tried to hide the console issues while reaffirming time and time again that current-gen console will be a first class citizen, that's an EA tactic not something that CDPR I thought I knew would do...

Moreover CDPR was given a $7 million grant to develop the artificial intelligence. This game has one of the worst open-world AI I have ever seen.

Cops spawn on you, the drivers and pedestrians AI is almost non existing. Grand Theft Auto: San Andreas from 2004 is miles ahead of this game.

They even announced that they are "very sorry" and that people can get a refund only that Sony won't refund if you played or even downloaded the game not to mention this game is very badly optimized on AMD systems the current gaming king.

Not to mention that Timestream inserts time range are dependent on the memory store configuration (https://aws.amazon.com/timestream/faq/), so if you need to insert data past that configuration it's a PITA.

Thanks for answering.

boundary proxy is an ok step but user experience should be streamlined especially if it's for teams and orgs and not just individuals who want to hack scripts but I full understand it's a 0.1 release.

Another thing I couldn't find in the docs is support for multiple installations, let's say I have different vpcs (In different accounts) or I have one on-prem installation and one in a cloud how do I login/switch/configure the cli to work seamlessly with multiple controllers.

Are there any plans or a way to use existing tools? By existing tools I mean winscp or any other tools that use a normal ssh client? RDP etc. I guess for shh and rdp you can just run the Boundary cli with a the predefined target in a terminal embedded into the UI (MremoteNG, MobaXterm etc) but tools like winscp are very much used for sftp file transfers.

A desktop client with a list of services/targets would also be great. Especially for the less technologically inclined individuals.

I know that people have their own opinions on port knocking but I find it as a good tool to remove a lot of noise, some pre built tool for that would be nice but could always just use fwknop-2

I haven't used Flyway but the XML format has some benefits like being DB agnostic and defining context (eg. production, testing etc.)

You can also include multiple change sets in one xml file, this helps with keeping the number of files down.

if you want you can just write pure SQL inside the xml with the <sql> tag.

Liquibase is great, been using it for about 6 years in production.

Liquibase also allow you to run java based change sets not only XML or YAML which is very good for more complex migration logic but you have to be careful not to use your DAOs or some service that might mutate in the future to support a schema change but rather use pure sql hardcode to that change set, as change sets run in order you can be sure of the current state of your schema.

If you are watching on a windows computer the Netflix windows store app can play 4k and 4k hdr titles.

Wow so a very large portion of their spend is on Cassandra. I bet it's heavily modified for their needs but I wonder if Scylladb would be able to replace some of those clusters for a considerable saving in maintenance time and money.

I'm never sure of the relation between the driver, nvidia-docker and the container with a specific cuda version.

Last time I tried it the cuda inside the container tough it was using some old driver version while a much newer version was installed on the host. So I had to manual install the older version, not sure where the issue was but maybe it was because I was using the deprecated nvidia-docker version 2 which is still needed to pass gpu resources to containers run inside kubernetes.

We have a few mysql instances where we only need them for a few days a month (maybe a week). Because a stopped instance will start automatically after 7 days we delete the instance and do a final snapshot.

When we restore we have to go through the process of configuring a new instance from that snapshot. Where you have to select the security group, availability zone, set it to not public (which btw revert to yes once you select some availability zone) etc.

This process is rather error prone and honestly annoying if you do it multiple times.

I don't think it's just the capacity but the power efficiency. G700s has a really low run time, something like a week per charge as a heavy user. The MX 2S can go about 2 months between chargers and the 3 should have an even better run time.

Yes, infinity fabric has limitations. Threadripper suffers in memory intensive workloads because only half of the CCXs have a dedicated memory controller thus some cores need to talk to another CCX to access memory. On EPYC each CCX has its own memory controller so it's less of an issue.

Can't this just be fixed so that Alexa, Siri, etc. will only accept your voice pattern?

1) Blind Test - That's true, but still people who buy check benchmarks, don't forget AMD did exactly that with Bulldozer and we know how that ended up.... If AMD thought that they had a winning horse they would not hide the performance numbers, they would PR that like crazy, like they are doing with Ryzen and ThreadRipper.

2) I'm all for AMD to kick Nvidia's ass in compute and end cuds's reign but it won't sell gaming cards. Vega FE (A 1000$ card) is out for a month and I haven't seen anyone benchmark it and show that it's better, Even if games will use it there is a heck of a difference between building the model and running inference.

Vega is a failure aka bulldozer 2.0 Same performance as 1080, a year later with double the TDP. Only 30% higher than the Fury which was their high end card from 2 years ago. This whole product launch was terrible from AMD laughing at Nvidia with "Poor Volta" to "blind tests" to no live streaming of the SIGGRAPH event.

Which is US only. The only reason to buy was the bundle, which is a good deal. Same performance as 1080 at the same price with double the TDP and will have limited stock. There is no reason to buy Vega as a gamer.