Google Cloud Community

I wrote about cloud run and inference w/ ollama on cloud run -> https://medium.com/google-cloud/ollama-on-cloud-run-with-gpu-in-less-t… performance is okay, ada lovelace has cuda 8_9 support which brings native fp8 support. imo the best aspect is the speed of spinning up new containers and the overall easiness of this service. the live demo at google next 25 was quite something https://www.youtube.com/watch?v=PWPvX25R6dM&t=2140s

I'd recommend learning how transformers work, and the concept of temperature. I don't think I need to cite information that is broadly and readily available, but here: https://medium.com/google-cloud/is-a-zero-temperature-deterministic-c4… I also qualified the requirement of needing the same hardware, due to FP shenanigans. I could further clarify that you need the same stack (pytorch, tensorflow, etc)

I get what you mean. For the record, here's a GCP engineer holding forth on price/performance ratios who also concludes that T2D is optimal. https://medium.com/google-cloud/google-cloud-platform-cpu-performance-… ETA: Since you are using Go and targeting a specific modern CPU, you may also get a measurable benefit from setting GOAMD64=v3, so the go compiler generates code using AVX2, BMI2, LZCNT, etc.

> 1. If you can make voice assistant software open-source and plug-and-play, then it frees people up to tinker with form factors For what it's worth, Google Assistant does have an open API to create new devices. It's not open source, but you can certainly experiment with your own custom form factors. There's even a tutorial: https://medium.com/google-cloud/how-to-build-your-own-smart-speaker-go…

Here's a great blog post on how to use it. https://medium.com/google-cloud/uploading-resizing-and-serving-images-…

Ah this old thing? :) https://cloud.google.com/products/calculator/ Yes, tough to cover it all. I believe Azure has a similar "Azure to AWS" comparison. The GCP one is built by folks expertly familiar with AWS and tried to be as factual as possible. I've previously collected a bunch of Google Cloud customer stories here (not Google PR - customers): https://www.medium.com/google-cloud/i-think-google-cloud-is-the-best-b… Happy to answer any questions! (obv. biased and work on Google Cloud)

It is more like RAID0. VictoriaMetrics shards time series among available vmstorage nodes. I.e. each vmstorage node contains a part of data stored in the cluster. This is usually named shared nothing architecture [1]. As for data replication, VictoriaMetrics offloads this task to the underlying storage, since the replication is hard to make properly [2]. Proper replication must be able to perform the following tasks additionally to copying the data to multiple nodes: * To heal the data (aka to return back replication factor) after a node becomes permanently unavailable. The healing process mustn't degrade cluster performance and it must properly handle other cases mentioned below. * To gracefully handle temporary unavailability of nodes. * To survive network partitioning when nodes are temporarily split into multiple isolated subnetworks. * To handle data corruption. * To continue accepting new data at normal rate when a part of nodes are unavailable. * To continue serving incoming requests with acceptable latency when a part of nodes are unavailable. * To replicate data among multiple availability zones (AZ), so the cluster should continue accepting new data and serving requests if a single AZ becomes unavailable. I'm unsure whether popular systems that claim replication support can handle all the cases mentioned above. The only system that seems to handle these cases properly is GCP persistent disks based on Colossus storage [3]. That's why it is recommended…

For anyone else who is a Google Cloud Customer. I 100% recommend "Exporting Billing Data to BigQuery" [1]. This takes less than 5 minutes and will give you near real-time extremely detailed billing data into BQ where you can explore it via SQL. Second, generate a real-time report off this data with something like "Visualize GCP Billing using BigQuery and Data Studio" [2]. That blog post is good, but it is over built in my opinion, all you really need is a single daily sum of all spend across your account and chart that. Maybe a stacked column with product color or something. This will allow you to quickly see if one product consumption is shooting up unexpectedly at a glance vs checking something at the end of the month and working with your account reps to figure something out. 6k z z 5k z z spend 4k z z z 3k z z z 2k x x x x x x x x x 1k x x x x x x x x x date 1 2 3 4 5 6 7 8 d+ x = expected z = unexpected For example, maybe you left a test env running or something and all of the sudden you're spending an extra few K per day. Maybe you're doing some query loops over a large BQ dataset and racking up tons of $$. Put the chart up on your big TV screen in the Ops area so everyone can quickly check it with all your other metrics. When you're spending 100k+/month, with a few folks digging around the console, it's hard to notice an extra 5k-10k here or there. This quick tip will…

> Most of what you are saying is unsubstantiated. You're welcome to your opinion. > 1/ How easy is it to get data in? Seems standard. It supports a MySQL adapter and a PostgreSQL adapter and then a bunch of Google property adapaters. This is standard. > 2/ Is my data safe? Doesn't use SSL, major red flag. The PostgreSQL adapter doesnt use SSL (https://support.google.com/datastudio/answer/7288010?hl=en&ref_top… This is a showstopper. > 3/ How easy is authoring? [EDIT] seems decent, see demo in reply. You can see Felipe's demo here and make your own assessment. https://medium.com/google-cloud/showing-off-the-new-free-google-analyt… > 4/ Does it support pivots? Can't tell from the website or examples, red flag. If you find an example I'd love to see it. My expectation at a bare minimum is a table view with a dimension down the left, a dimension across the top, and a measure in the cells. I looked through all the examples and did not see a clear example of a pivot. > 5/ Can I model joins? [EDIT] it cant, see reply It can't. https://news.ycombinator.com/item?id=15446616 > 6/ Does it support dashboard-wide filters? Many examples of this in the gallery Most of the examples in the gallery use this feature. > 7/ Does it support complex time comparisons? No examples, red flag. Here's a couple very basic examples of what I'm talking about. - Plug an TY date range into the dashboard filter.…

UUIDv7 seems popular for Postgres performance improvements, but it causes issues with databases like Spanner. https://medium.com/google-cloud/understanding-uuidv7-and-its-impact-on…

That's not a usable workaround. In most cases it doesn't actually produce full determinism[1]. And even if it did, a certain degree of non-determinism is actually desirable. The most probable tokens might not be correct, and randomness is partly responsible for what humans interpret as "creativity". Even hallucinations are desirable in some applications (art, entertainment, etc.). [1]: https://medium.com/google-cloud/is-a-zero-temperature-deterministic-c4…

There is at least a parameter called Temperature which decides how much randomness to include in the output. It doesn't get you perfectly deterministic output to set it to 0 though, per https://medium.com/google-cloud/is-a-zero-temperature-deterministic-c4… as you don't have perfect control over what approximations are being made on your floating point operations

FYI, a list of vector databases (including announcements). I try to keep the list as up-to-date as I can: https://medium.com/google-cloud/vector-databases-are-all-the-rage-872c…

I find it interesting that it’s quoted random IDs are bad for performance, because it’s actually better for distributed storage systems because you don’t hotspot on a single node. For example see: https://stackoverflow.com/a/53901549 and https://medium.com/google-cloud/cloud-spanner-choosing-the-right-prima…

I recently blogged about: Integrating DockerSlim container minify step on Cloud Build https://medium.com/google-cloud/integrating-dockerslim-container-minif…

Will point to some experiences open source users have had on this journey https://medium.com/hacking-talent/production-code-for-data-science-and… https://medium.com/google-cloud/migrate-kedro-pipeline-on-vertex-ai-fa…

Interesting. I've done dual-writes at the application level to migrate the datastore for a smaller feature (branch by abstraction), but never for an entire application. And the code path was quite simple, so it was easy to think about all of the edge cases at one time in your head. Do you have any resources which talk through the read/write/backfill approach? Here's what I found so far: * https://medium.com/google-cloud/online-database-migration-by-dual-writ… * https://aws.amazon.com/blogs/architecture/middleware-assisted-zero-dow…

Definitely do-able, Google wrote a blog about how to do exactly that on the GCP Cloud Shell a while back so not unreasonable to do the same on AWS: https://medium.com/google-cloud/how-to-run-visual-studio-code-in-googl… And Hashicorp have one on using Terraform from GCP Cloud Shell: https://www.hashicorp.com/blog/kickstart-terraform-on-gcp-with-google-…

Isnt this what google cloud shell + editor [1] already does ? [1]https://medium.com/google-cloud/developing-on-gcp-getting-started-with…

This submission being popular in HN caused an interesting traffic spike in Programming Idioms. Here is a detailed recital of the technical and financial consequences: https://medium.com/google-cloud/surviving-traffic-spike-from-hacker-ne…

Can't you do all this same research in just a few seconds for like pennies, just by sending a few sql queries to Google's BigQuery? I'm pretty sure we've had stories about that here. https://medium.com/google-cloud/github-on-bigquery-analyze-all-the-cod…

It depends on the language you're using and how you manage your dependencies. If you're using Java, see this writeup for how different frameworks can affect startup time: https://medium.com/google-cloud/java-frameworks-performances-on-cloud-… Similarly, if you're using Node or Python, you might want to see if any of your dependencies are enormous with lots of files and slow startup time -- you can check this locally by timing how long it takes to get to the initial listen() call and just print that wall time as you adjust dependencies. If you're building Golang and you're seeing slow cold starts... I have no idea how you're doing that. For development, a lot of us on the open source http://knative.dev/ side are using Go http servers that take tens of milliseconds to start up, so there's probably some other initialization that's slowing you down.

As a Googler who has a lot of technical content on Medium, I have to agree that screenshots are much more flexible than Medium's gist support, but I always add a link to the actual code right under. For example: https://medium.com/google-cloud/kubernetes-configmaps-and-secrets-68d0… In this example, I have a single gist with multiple files, but Medium can't display just a single file from that gist at a time (maybe things have changed, but I don't think so). Medium's gist and code support in general is sub-par compared to other platforms IMO, but as a platform it does make it easy to create and share content that is also easily discoverable. I'm probably going to start hosting the canonical post on my own domain and just use Medium to syndicate it.

Why doesn't this just use the GitHub public dataset available on Google BigQuery to have much more accurate data rather than "scraping GitHub web search results"? https://cloud.google.com/bigquery/public-data/ There are a lot of examples of people analyzing public code on GitHub efficiently for patterns and usages with BigQuery and getting pretty accurate data out of it. https://medium.com/google-cloud/analyzing-go-code-with-bigquery-485c70… If you use GitHub on a daily basis, you are unlucky enough to know that web search sadly can't even find words that exist in your repository.

Related: I wrote this post showing the number of stars that GitHub projects get after being featured on the HN front page: - https://medium.com/google-cloud/big-data-stories-in-seconds-hacker-new…

It works very well and has controls so you don't schedule important workloads on preemptable instances: https://medium.com/google-cloud/using-preemptible-vms-to-cut-kubernete…

This is an interesting article about k8s https://medium.com/google-cloud/kubernetes-from-load-balancer-to-pod-3…

At the risk of self-promotion :). https://medium.com/google-cloud/understanding-kubernetes-networking-po… This is the first of a two-part series, the first dealing with pod networking and the second with services. I plan a third on ingress after kubecon. It's a little GKE-specific in the implementation details, and the whole thing is pluggable and can be configured in different ways (as the OP shows), but I think it covers the fundamentals pretty well.

Cluster federation is an interesting feature that I don't see come up often. I'm curious if anyone is actually running k8s over AWS and GCP. https://kubernetes.io/docs/concepts/cluster-administration/federation/ https://medium.com/google-cloud/experimenting-with-cross-cloud-kuberne…