CloudFront 的 BBR 效能提昇

這是在找一些 TCP congestion algorithm 相關的資訊時發現的,看起來 Amazon CloudFront 導入 BBR 一陣子了:「TCP BBR Congestion Control with Amazon CloudFront」。

不過 BBR 被研究的愈來愈多,大家開始發現這個演算法的霸道,跟其他的 TCP congestion algorithm 並不太能和平共存,但這就跟軍事武器一樣,隔壁升級了你就得跟著升級,抱怨沒有用,只會被消滅...

AWS 推出了與 5G 結合的 AWS Wavelength

看到「Announcing AWS Wavelength for delivering ultra-low latency applications for 5G」這篇,裡面提到了與 5G 結合的 AWS Wavelength

裡面提到的東西包括了把 AWS 的架構延伸到各家電信的架構裡面,看起來跟剛推出的 AWS Outposts 有關 (參考「AWS Outposts 總算要開始出貨了」這邊的介紹):

AWS Wavelength brings AWS services to the edge of the 5G network, minimizing the network hops and latency to connect to an application from a 5G device.


AWS is partnering with leading telecommunications companies around the world, including Verizon, Vodafone, SK Telecom, and KDDI, to launch Wavelength across North America, Europe, Japan, and South Korea in 2020.


AWS 的 VPC 在 Routing 上的改善

在這次 re:Invent 發表會上,AWS 也宣佈了一些跟 VPC routing 有關的改善。

第一個是 AWS Transit Gateway 彼此可以互串了:「New for AWS Transit Gateway – Build Global Networks and Centralize Monitoring Using Network Manager」。

第二個是可以拿 EC2 的機器 (實際上應該是 ENI) 當作 routing 的目標:「New – VPC Ingress Routing – Simplifying Integration of Third-Party Appliances」。


AWS 開始推自己的 Machine Learning Chip

除了常見的 GPU 類,以及之前公佈過的 FPGA 外,這次 AWS 推出的是自己做的晶片 AWS Inferentia,以及對應到 EC2 上的機種 inf1:「Amazon EC2 Update – Inf1 Instances with AWS Inferentia Chips for High Performance Cost-Effective Inferencing」。


Each AWS Inferentia chip supports up to 128 TOPS (trillions of operations per second) of performance at low power to enable multiple chips per EC2 instance. AWS Inferentia supports FP16, BF16, and INT8 data types. Furthermore, Inferentia can take a 32-bit trained model and run it at the speed of a 16-bit model using BFloat16.


AWS Inferentia comes with the AWS Neuron software development kit (SDK) that enables complex neural net models, created and trained in popular frameworks to be executed using AWS Inferentia based EC2 Inf1 instances. Neuron consists of a compiler, run-time, and profiling tools and is pre-integrated into popular machine learning frameworks including TensorFlow, Pytorch, and MXNet to deliver optimal performance of EC2 Inf1 instances.

現在看起來類似於 Google 弄的 TPU,專為 machine learning 搞出來的 ASIC,等一陣子應該就會有兩者的比較了...

AWS 提供 Machine Learning 能力的自動 Code Review 服務

AWS 推出了 Code Review 服務 Amazon CodeGuru,使用 machine learning 提供建議:「AWS announces Amazon CodeGuru for automated code reviews and application performance recommendations」。

從界面就可以看出來同時支援 GitHub 與自家的 CodeCommit,看起來可以給不少建議,但網站上沒有提到 security 這塊,本來以為產品的定位不在這邊:

不過 FAQ 裡還是有提到常見的 security issue:

Q: What type of issues are detected by Amazon CodeGuru Reviewer?

Amazon CodeGuru Reviewer checks for concurrency issues, potential race conditions, un-sanitized inputs, inappropriate handling of sensitive data such as credentials, resource leaks, and also detects race conditions in concurrent code.

然後 FAQ 裡提到目前只支援 Java:

Amazon CodeGuru Reviewer currently supports Java code stored in GitHub and AWS CodeCommit repositories.

服務的價位是使用行數計算,不過那個 per month 沒看懂是什麼意思:

Code scan (pull requests)$0.75 per 100 lines of code scanned per month

另外推出的 Amazon CodeGuru Profiler 則是 APM 類的東西,這塊目前市場上產品也很多,看起來也要被 AWS 進來蹂躪...

AWS Outposts 總算要開始出貨了

去年 AWSre:Invent 喊的 AWS Outposts 總算是有東西要出貨了:「AWS Outposts Now Available – Order Yours Today!」。

放在自家實體的機櫃,然後掛到 AWS 上變成一個特殊的 region。目前一個特殊的 region 只能放 16 個機櫃,但預期之後可以更多:

Capacity Expansion – Today, you can group up to 16 racks into a single capacity pool. Over time we expect to allow you to group thousands of racks together in this manner.

不過要注意的是,需要有 AWS Enterprise Support 才能下單,而且看起來硬體的維修也包在內了:

Support – You must subscribe to AWS Enterprise Support in order to purchase an Outpost. We will remotely monitor your Outpost, and keep it happy & healthy over time. We’ll look for failing components and arrange to replace them without disturbing your operations.

看了一下價錢的頁面,如果以北美的 upfront 來算,最便宜的是 OR-L8IF4WFOR-I0OGL02 的 USD$225,504.81,最貴的是 OR-HSZHMMF 的 USD$898,129.52,暫時應該用不到 XDDD

AWS 在 us-west-2 開 Local Zone

AWS 宣佈 us-west-2 (Oregon) 開 Local Zone,這應該是 AWS 第一次在美國開 Local Zone,上次看到好像是 ap-northeast-1 (Tokyo) 的 Osaka 區:「AWS Now Available from a Local Zone in Los Angeles」。

控制都還是在 us-west-2 的範圍控制,但代碼會是 us-west-2-lax-1a (目前只有一區),之後會開 us-west-2-lax-1b (第二區):

In the fullness of time (as Andy Jassy often says), there could very well be more than one Local Zone in any given geographic area. In 2020, we will open a second one in Los Angeles (us-west-2-lax-1b), and are giving consideration to other locations. We would love to get your advice on locations, so feel free to leave me a comment or two!

剛剛登入進去 VPC 的 Subnets 想要增加看看,沒看到 us-west-2-lax-1a 的選項可以選,看起來還是需要另外申請?

另外算了一下 Oregon (用 Portland 算) 到 Los Angels 的直線距離,大約要 1300km 左右 (比台北到香港還遠不少),光速單趟大約要 6.5ms?這樣對一些應用程式應該是會有感覺...

This Local Zone is designed to provide very low latency (single-digit milliseconds) to applications that are accessed from Los Angeles and other locations in Southern California.


RSA-240 (十進位 240 位數) 成功分解

在「RSA-240 Factored」這邊看到的,RSA-240 前幾天被解開了:

RSA-240 = 124620366781718784065835044608106590434820374651678805754818788883289666801188210855036039570272508747509864768438458621054865537970253930571891217684318286362846948405301614416430468066875699415246993185704183030512549594371372159029236099 = 509435952285839914555051023580843714132648382024111473186660296521821206469746700620316443478873837606252372049619334517 * 244624208838318150567813139024002896653802092578931401452041221336558477095178155258218897735030590669041302045908071447

有效長度是 795 bits,相較於十年前解出來的 RSA-768 (768 bits) 又多了「一些」。看了一下,算法上沒有太多突破,主要是硬體的發展與軟體的最佳化有進展...

Amazon Elasticsearch Service 可以利用 S3 當作二級儲存空間了

Amazon Elasticsearch Service 的新功能,使用 Amazon S3 當作第二級儲存空間 (UltraWarm):「Announcing UltraWarm (Preview) for Amazon Elasticsearch Service」。

UltraWarm 需要不同的機器 (跑不同版本?),機器的規格 (vCPU 與記憶體的比率) 接近 Memory Optimized 的版本,但是貴了不少,所以需要夠大的資料量才會打平回來...

us-east-1 來看,SSD EBS 的空間成本就是 USD$0.135/GB,而傳統磁性硬碟是 USD$0.067/GB (不知道收不收 I/O 費用?),但 storage 的價錢是 USD$0.024/GB。這邊值得一提的是 Amazon S3 是 USD$0.023/GB,看起來是直接包括了 API 的呼叫費用?

Cassandra 也被 AWS 包成服務了

也是剛剛發表的服務 (所以在 Twitter 上看到),把 Apache Cassandra 包成服務,叫做 Amazon Managed Apache Cassandra Service:「New – Amazon Managed Apache Cassandra Service (MCS)」。

而且是個 serverless 服務,直接用服務,不需要管理機器:

Amazon MCS is serverless, so you pay for only the resources you use and the service automatically scales tables up and down in response to application traffic.

從計費的方式也可以看出來這點,是對 Write request units、Read request units 與 Storage 收費,沒有看到機器的費用。

不過稍微算了一下不算便宜,如果沒有用到 Cassandra 的特性的話,比 DynamoDB 貴一些?

目前是 open preview 狀態,是個可以用但是不掛保證的意思:

Amazon MCS is available today in open preview in US East (N. Virginia), US East (Ohio), Europe (Stockholm), Asia Pacific (Singapore), Asia Pacific (Tokyo).