Home » Posts tagged "instance" (Page 3)

Amazon EC2 推出 4TB 的機器

之前 Amazon EC2 記憶體最大的機器是 x1.32xlarge 的 2TB RAM (更精確是 1952GB),現在推出了 4TB RAM 的 x1e.32xlarge (3904GB):「Now Available – EC2 Instances with 4 TB of Memory」。

現在這個時間點在 us-east-1 的價錢是 USD$26.688/hour (一個月 USD$19215.36),用的到的人應該付得起?

另外值得注意的是,x1e.32xlarge 雖然比 x1.32xlarge 多了一倍的記憶體,但 vCPU 不變 (都是 128),而且 ECU 下降了 (從 349 降到 340)。

這個機器目前在 us-east-1us-west-2eu-west-1ap-northeast-1 四區提供服務:

The x1e.32xlarge instances can be launched in On-Demand and Reserved Instance form via the AWS Management Console, AWS Command Line Interface (CLI), AWS SDKs, and AWS Marketplace in the US East (Northern Virginia), US West (Oregon), EU (Ireland), and Asia Pacific (Tokyo) Regions.

EC2 的 Spot Instance 可以「接關」

Amazon EC2Spot Instance 可以「接關」了:「New – Stop & Resume Workloads on EC2 Spot Instances」。

當 Spot Instance 的競價不足以標到機器時,他會先關起來 (Stop),等到價錢低於競價後就會再打開機器,這時候的狀態就會恢復。另外也提到了必須是使用 EBS 的機器才支援:

Amazon EC2 Spot now allows Amazon EBS-backed instances to be stopped in the event of interruption, instead of being terminated when capacity is no longer available at your preferred price. Spot can then fulfill your request by restarting instances from a stopped state when capacity is available within your price and time requirements.

用法是把 Spot Instance 的關機設定設為 Stop:

To use this new feature, choose “stop” instead of “terminate” as the interruption behavior when submitting a persistent Spot request. When you choose “stop”, Spot will shut down your instance upon interruption.

恢復的時候就會儘量保持一樣地開回來 (連 instance id 都相同):

When capacity is available again within your price and time requirements, Spot will restart your instance. Upon restart, the EBS root device is restored from its prior state, previously attached data volumes are reattached, and the instance retains its instance ID.

是個接關的感覺 XD

在 EC2 上使用 25Gbps 的網路

Amazon EC2 上許多系列最大台的機器現在都可以透過特殊界面跑到 25Gbps 了:「Announcing improved networking performance for Amazon EC2 instances」。

Amazon EC2 instances now provide a maximum bandwidth of 25 Gbps. This feature is available on the largest instance sizes of the M4, X1, P2, R4, I3, F1, and G3 instance types. Using Elastic Network Adapter (ENA) based Enhanced Networking, customers can utilize up to 25 Gbps of bandwidth.

除了 AWS 提供的的 AMI 外,另外在一些官方作業系統也都有支援:

ENA driver is installed in the latest Amazon Machine Images (AMIs) for the following operating systems: Amazon Linux, Ubuntu 14.04 and 16.04, RHEL 7.4, SLES 12, Windows Server 2008R2, 2012, 2012R2 and 2016. ENA Linux driver source code is also available on Github.com for developers to integrate in their AMIs.

如果不在上面的,也可以透過 GitHub 上的 amzn/amzn-drivers (Official repository of the Elastic Network Adapter (ENA) network adapter for Linux and FreeBSD operating systems) 自己 porting... (如果你一定要想辦法用的話)

透過 AWS Budgets 管控 Reserved Instance 的使用率

現在可以透過 AWS Budgets 來看到 Reserved Instance 的使用率了:「Monitor your Reserved Instance utilization by receiving alerts via AWS Budgets」。

Starting today, customers can now monitor and receive alerts when their Reserved Instance (“RI”) utilization falls below the threshold they define.

算是個開頭,不過我更想請 AWS 分析有哪些機器是長年開著沒有 discount 的... 或者像是 GCP 那樣自動算 :o

Amazon EC2 推出新的 GPU 系列機器

這次 Amazon EC2 推出了 G3 instance:「New – Next-Generation GPU-Powered EC2 Instances (G3)」。

這次開放的只有這幾區,亞洲目前還沒有在內:

You can launch these instances today in the US East (Ohio), US East (Northern Virginia), US West (Oregon), US West (Northern California), AWS GovCloud (US), and EU (Ireland) Regions as On-Demand, Reserved Instances, Spot Instances, and Dedicated Hosts, with more Regions coming soon.

這次使用的 GPU 是 Nvidia 的 Tesla M60:

Today we are taking a step forward and launching the G3 instance type. Powered by NVIDIA Tesla M60 GPUs, these instances are available in three sizes (all VPC-only and EBS-only)[.]

G3 (M60) 應該會比 G2 (K520) 快,但未必會比 P2 (K80) 快,尤其是 double precision 的部份,可以參考「Nvidia Tesla」這邊的數據。所以還是得看應用程式才能決定要用 G3 還是 P2...

AWS Batch 在東京區可以用了

AWS Batch 支援東京區了:「AWS Batch is Now Available in Tokyo」。

總算是在東京出現了... 在 FAQ 裡面有提到:

AWS Batch uses Amazon ECS to execute containerized jobs and therefore requires the ECS Agent to be installed on compute resources within your AWS Batch Compute Environments. The ECS Agent is pre-installed in Managed Compute Environments.

開起來測一些東西看看...

Amazon EC2 要推出 x1e.32xlarge,4TB RAM 的機器

剛剛 Amazon EC2 公佈的消息,要再推出記憶體大怪物機器 x1e.32xlarge:「EC2 In-Memory Processing Update: Instances with 4 to 16 TB of Memory + Scale-Out SAP HANA to 34 TB」。

Later this year we plan to make the x1e.32xlarge instances available in several AWS regions, in both On-Demand and Reserved Instance form. These instances will offer 4 TB of DDR4 memory (twice as much as the x1.32xlarge), 128 vCPUs (four 2.3 GHz Intel® Xeon® E7 8880 v3 processors), high memory bandwidth, and large L3 caches.

雖然都是拿 SAP 來舉例,但這也是當系統碰到瓶頸時搶時間的方法 (也就是當 happy problem 發生時的 workaround)。

不知道這種機器實際買起來多少錢 XDDD

Amazon EC2 的 RI 降價

這次 Amazon EC2 主要降的是 Reserved Instances 的部份:「EC2 Price Reductions – Reserved Instances & M4 Instances」。

  • New No Upfront Payment Option for 3 Year Standard RIs
  • Lower Prices for No Upfront Reserved Instances
  • Lower Prices for Convertible Reserved Instances

另外一般性的降價以 m4 為主:「Lower Prices for M4 Instances」,如果拿 t2.large 跟 m4.large 比較,us-west-2 (Oregon) 的價錢就很微妙了:

t2.large2Variable8EBS Only$0.094 per Hour
m4.large26.58EBS Only$0.1 per Hour

如果是 ap-northeast-1 (Tokyo) 的話是:

t2.large2Variable8EBS Only$0.128 per Hour
m4.large26.58EBS Only$0.129 per Hour

雖然 AWS 官方的說法是這樣 (取自「AWS t2 burst mode ECU equivalence」),表示 t2 系列有他特別可以應用的地方,但還是有哪邊覺得很微妙...:

Amazon EC2 的 F1 type 開放一般使用

AWS 提供更快計算 Bitcoin 的 FPGA 機種開放一般使用了:「Amazon EC2 F1 Instances, Customizable FPGAs for Hardware Acceleration Are Now Generally Available」。

在 AWS 開始提供服務後,應該會有更多 library 支援吧... 現在現有的應用要上去還得自己先刻些東西,不像 TensorFlow 可以透過 GPU 運算。

F1 instances include the latest 16 nm Xilinx UltraScale Plus FPGA with local 64 GiB DDR4 ECC protected memory, with a dedicated PCI-e x16 connection to the instance. For F1.16xlarge instances, the dedicated PCI-e fabric lets the FPGAs share the same memory space and communicate with each other across the fabric at up to 12 GBps in each direction. The FPGAs within the F1.16xlarge share access to a 400 Gbps bidirectional ring for low-latency, high bandwidth communication.

Archives