Let's start at the very beginning. What is ECS? ECS stands for Elastic Compute Service. This is essentially Alibaba Cloud's virtual machine service, an ECS instance is just a virtual machine. Compared to a physical server, an ECS instance can be deployed immediately and automatically and can be backed up, cloned or replaced at any time.

So what does an ECS instance include? Well, you can break this into the hardware and software layers. On the hardware side, each ECS instance will have some amount of CPU and memory, the amount of CPU and memory being determined by the instance type, some amount of block storage, which we call Cloud Disk. You can easily make point in time snapshots of Cloud Disk anytime you want. And then an Elastic Network Interface, or ENI, which is the network interface that the ECS instance uses to communicate over the network.

On the software side, there are operating system images, which are used to create the ECS instances system disk. These include an operating system, such as CentOS, Windows Server or Ubuntu, and you can also make your own custom operating system images. Then there is the security group, which acts as a stateful firewall for the instance.

Let's discuss some basic ECS concepts. In addition to Cloud Disk or Block Storage, which every ECS instance has, you can also attach ECS instances to other Alibaba Cloud storage services over the network. So you can attach ECS instances to NAS, which is our shared file storage, it's Network Attached Storage, and OSS, our Object Storage Service. Both of these services have private internal network endpoints within Alibaba Cloud, so your ECS instance does not need to connect to the internet to use these network services. Cloud Disk, and again, every instance has at least one because the system disk holding the operating system is a Cloud Disk, can be broken into three types: Ultra Disk, SSD, and ESSD. Ultra Disk is a traditional hard drive, SSD is a Solid State Drive, and ESSD is an Enhanced Solid State Drive with higher IO performance. Newer generation six ECS instances, all use ESSD.

Let's now take a look at networking. Every ECS instance lives inside of a Virtual Private Cloud. This is a private network environment that contains this ECS instance, as well as possibly other ECS instances and connectors to other services like our database service. Inside of the VPC, you will have a VRouter, which is a hub that connects the VSwitches in the VPC and serves as a gateway between the VPC and other networks, as well as at least one Vswitch. A VSwitch is a virtual switch, this is how your ECS instance connects to the VPC.

Let's quickly get some terminology out of the way before we go any further. First, there is the ECS instance, this is equivalent to a virtual machine, and this will include the CPU, memory disk, network and operating system configuration of the ECS service. That is an ECS instance. Then we have the instance family, instances are grouped into families, which map onto different business scenarios. The CPU to memory ratio of your instance will be determined by its family. Some instances may also include additional special purpose hardware, like GPU or FPGA cards.

Then there is the instance size. This is essentially the number of vCPUs or virtual CPU's allocated to the instance. For instance, a '2xlarge' instance type would have four vCPUs. Then there is the image, as we just discussed, this is the operating system and other software that's used to boot the instance. Block storage, this refers to Cloud Disks, but it can also refer to local disks. Some instance families include a local disk, which is a Solid State Drive that's physically attached to the instance itself.

Then there is the snapshot, this is a convenient and efficient cloning and disaster recovery tool. It's a point in time copy of everything on an instances disk or disks. Then we have the security group, this is a virtual stateful firewall, which protects the instance. After that we have the IP address. Each instance will have a private IP address, which allows it to communicate within its VPC group, but it may also have a public IP address, which allows it to communicate with the internet. The public IP address is optional. And then finally we have the concepts of a region and an availability zone.

A region is a geographical area, such as Singapore or Hong Kong, and an availability zone is a data center within that region. So for instance, in Singapore we have three physically separate data centers, zone A, zone B, and zone C. Within a given region, the zones will be linked together using low-latency private network links. I also want to take a moment to discuss block storage. ECS instances utilize Cloud Disk as the primary mechanism for block storage for both system and data disks.

Now, the nice thing about Cloud Disk is that it stores three redundant copies of the disks data separately on different physical machines attached to different network switches. Remember, Cloud Disk is a type of virtual disk that's actually made available to the ECS instance over the network, so it's possible to store multiple replicas of Cloud Disks, and that is in fact what we do, in order to improve reliability and redundancy. 

Newer instance types use our Enhanced SSD as the physical storage backing the Cloud Disk. This allows up to 1 million IO operations per second, and a throughput of up to four gigabytes per second, with a latency in the microseconds, typically between 100 and 150 microseconds. ESSDs are categorized into four performance levels, PL0 through PL3. A PL0 is fine for most applications and for the system disk, this is the storage class we are using to replace our older SSD and Ultra Disk types on our newer ECS instances. So newer ECS instances will typically use ESSD. If you're looking for the cheapest ESSD, then you want PL0. On the other hand, if you're running a very high performance search or relational database application, you might want to consider PL2 or PL3, which offer much better IOPS and much better maximum throughput.