Dolphin Browser Cloud Road

Dolphin Browser in February 2010 officially released the Android version, nearly a year after the official release from a pure client products began to iteratively evolve, gradually joined the functions of cloud services, Dolphin browser in the cloud of the road to depart. At the beginning of the venture, because of the shortage of resources and manpower, the deployment of natural cloud services became preferred. The Amazon cloud platform was so popular among small entrepreneurial companies abroad that we did not hesitate to choose AWS as a service provider, and did a massive attempt at the first release of Dolphin Reading (Dolphin webzine), which followed the launch of Dolphin synchronization, Dolphin sonar and other cloud services.

First contact with the cloud platform is actually more than the cloud platform as a common IDC host rental services in use, experience the advantage is relative to the physical host, Cloud Host (instance) on-line/offline is more convenient. And unlike many domestic host services are prepaid or prepayment consumption patterns, AWS platform payment directly linked to the credit card, how much to pay how much, very flexible. With the passage of time, the understanding of the cloud platform is shallow and inadequate, leading to a lot of our suffering, of course, has accumulated a lot of experience and lessons. Up to now, most of the services in the AWS Cloud Platform (see Figure 1) have practical experience.

Figure 1:aws Service Stack

Scalability on the cloud platform

When it comes to the use of cloud platforms, you have to talk about horizontal extensions (scale out). Before the service was officially released, we had to deploy a number of same-role machines (such as front-end machines and application servers) to ensure that unexpected traffic growth could be addressed. The deployment and use of these standby machines is not really necessary on the cloud platform. In AWS, a flexible load balancer such as ELB can be used to automatically extend the level of service, ELB supports multiple protocols, and can customize the conditions for horizontal scaling, which is a lot of development for the service developer, and for ordinary load balancing scenarios, It can completely replace nginx or haproxy.

For vertical expansion (scale up), there are two points is more important, one is the cloud host Take-off and landing type choice, the second is to understand the life cycle of the cloud host. EC2 on the cloud host has a fixed number of types, the first choice is generally 64-bit machine, so that convenient memory expansion, if the CPU consumption is higher (such as HTTPS connection), then the preferred high-cpu type, if the memory requirements are relatively large (such as MongoDB), Then choose the High-memory type first. Most of the Dolphin's machine selection focuses on Micro/small as monitoring and front-end, Small/medium deployment application services, message queues, Medium/large database and off-line computing. XLarge are used very little, and are basically solved by horizontal expansion. For the life cycle of the cloud host, restart and Stop/start are different, the take-off and landing must be a stop cloud host, the lifting level is completed and then started, the machine's internal IP will change (IP through DHCP allocated), This can often cause problems with IP-dependent service configurations. There are two solutions, one is through VPC to control the IP address allocation, the other is the use of elastic IP such static address.

Storage on the cloud platform

Like computing resources, storage resources are one of the core elements of a cloud platform. Storage on the cloud platform is divided into three major types according to usage scenarios:

• Temporary storage. The instance storage of AWS is a kind of temporary storage, mainly used to store cache and some intermediate results. Note that the contents of the temporary storage will be emptied after the cloud host stop, because the DF command often does not see this, so there have been instance storage as a persistent storage experience, the loss is very heavy.

• Persistent storage. Persistent storage most often refers to the physical hard disk, on the AWS platform, EBS is such a can be mounted at any size of the "hard disk", in fact its implementation is a network file system, so its access rate is limited by network bandwidth, and not so stable. It is usually possible to troubleshoot I/O rate problems by doing raid on the EBS standard volume or by using the latest provisioned IOPS volume. In addition, although EBS is a persistent storage, does not mean that it will not be the loss of data, EBS has an annual bad rate of 0.1%-0.5%, so for a single storage node need to do regularly to the image and backup of EBS to prevent accidental occurrence.

• Large-scale redundant storage. S3 is a sample of this type of storage. S3 is not a filesystem architecture, I/O rate and latency is not as good as EBS, but its advantage is that one can provide 99.999999999% reliability at a relatively low price (similar to EBS), on the other hand can access very large, such as PB-level data, This data can be used anywhere, not limited to AWS. Dolphins use S3 to store almost all mirrors, data backups, and various logs. In addition, S3 and CloudFront (AWS's CDN Solutions) are highly integrated, so dolphins often use S3 as a conduit for distributing content such as APK.

(Responsible editor: The good of the Legacy)