Tengxun Cloud Account: How to design the framework of game going to sea? Tengxun Cloud Global Network Interworking and CVM Selection for Game
In recent years, the domestic game circle of the most volume of the two words, there is no doubt that the sea ".
Whether you are making a single-service million online SLG (strategy), a global MOBA, or a lightweight casual game, as long as you extend your business tentacles overseas, the challenges faced by the technical team will directly rise in several dimensions:
The hundreds of milliseconds of delay brought about by cross-border long-distance networks, the intricate network of operators overseas, the global distribution of hundreds of G patches when the game is released, and how to make global players fight without Caton in the same world?
Many teams that have just gone out to sea are used to using the domestic thinking of "a single computer room covers the world" to set up overseas. As a result, overseas players complain about "getting stuck into PPT" and "getting disconnected every day" in the community every day, and the traffic they bought evaporates instantly.
Today's tutorial does not talk about empty architecture principles, directly from
global network interworking
and
Server Selection
The two core hard-core dimensions take you to use the most down-to-earth technical language to build a standard framework for tengxunyun games that can resist, play and save money.
Hardcore metrics for games out to sea: How sensitive is your game to latency?
Before drawing an easel, "look in the mirror" for your game type ". Different game forms, the cost and design of the network architecture are completely different:
The first category: strong interactive real-time competition (MOBA, FPS, FPS, ball games) life and death line: delay must be within 50ms, and games over 100ms cannot be played directly. Architecture solution: multiple regional deployment (Regional Deployment) must be used to build physical room servers in North America, Europe and Southeast Asia respectively, and players can access them nearby.
The second category: moderate interaction global service (SLG, card, MMO non-core combat) life and death line: delay is acceptable between 150ms and 200ms, allowing short visual delay of data synchronization, but never packet loss. Architecture solution: The global network acceleration line is centrally deployed using the base camp.
Phase I: Design of Tengxun Cloud Global Network Interworking Architecture (Solving "Network Catton")
The Internet is the blood vessel of the game. The most taboo for games to go to sea is to let players directly take the complicated public network international route to connect to your server. Tencent Cloud's core moat in the game circle is its priceless global backbone network.
1. Access Layer: Anycast EIP Global Application Accelerate(GAAP)
The first jump of overseas players to access the server must be "captured" with the fastest speed ".
Seamless Access to the Same Service across the Network (Anycast EIP): As mentioned earlier, Tengxun Cloud's Anycast IP allows you to broadcast the same IP across the network. For example, your SLG game base camp is in Frankfurt. You distribute the same Anycast IP to players in the United States, Japan and Southeast Asia. Players will directly "access the network" in the nearest local Tengxun cloud edge computer room ".
travel
Protocol-Specific Acceleration (GAAP): Many strong interactive games use a custom UDP/TCP protocol. Tencent Cloud's GAAP technology is tailored specifically for games. It has deployed access points in dozens of countries overseas. After the UDP game package of overseas players enters the local GAAP node, it directly cuts into the global high-bandwidth dedicated line of Tengxun Cloud and transmits it all the way back to the domestic or core base camp through the intranet, perfectly solving the problems of high jitter and frequent disconnection of the public network.
2. Bottom layer interworking: multi-computer room series connection of cloud networking (CCN)
If your architecture design is: Guangzhou base camp has full user data, and the United States, West, Tokyo and Singapore each have a set of room servers (Battle Server) to play the competition. How can both sides communicate efficiently?
Do not go to the public network to pull synchronization. The correct approach is:
Establish Tencent Cloud VPC in Guangzhou, US West and Tokyo respectively.
Open Tencent Cloud Networking (CCN) and pull VPCs in these three regions directly into the same cloud networking instance.
Intranet opening: at this time, the room server in west America can directly access Guangzhou's global database through intranet IP such as 10.x. x.x with extremely low intranet delay and 100 zero packet loss rate. The whole process does not need to go through any external network firewall, safe and fast to the extreme.
The second stage: game-specific CVM selection guide (how to buy the most cost-effective?)
The cost of overseas servers is often higher than that in China. If you choose the wrong model, it will not only waste your power, but also make your monthly bill hurt. Tengxun Cloud has hundreds of cloud server (CVM) models. It is suggested to directly lock the following two "core main forces" in the game scene ":
1. Room/Battle Server (Battle Server): Deadstick [Computing C6/C7]]
The battle server is responsible for calculating ballistics, collision volume, damage values, field of view clipping, yes.
CPU single-core explosive power
Demanding as high as a pervert.
Selection recommendation: Tencent cloud computing C6 or the latest generation C7 instance.
Why: These two models are equipped with the latest generation of high-frequency Intel or AMD processors (full-core turbo boost is usually above 3.5GHz). More importantly, their memory ratio is usually 1:2 (such as 4-core 8G,8-core 16G), spending every penny on pure CPU computing power. When the game server is faced with a large range of skill release and a sharp increase in the number of people on the same screen, the high main frequency can ensure a stable frame rate and there will be no metaphysical Caton.
2. Game Login/Hall/Microservice (Lobby Server): Embrace [Standard S6/SA3]]
The hall server mainly deals with the logic of player login, backpack management, mall trading, card pumping, etc. The characteristics of these businesses are:
Not computationally intensive, but requires reading and writing a large amount of memory cache (such as Redis).
Selection Recommendation: Tengxun Cloud Standard S6 or AMD Standard
Quasi type SA3.
Why choose: the standard memory ratio is 1:4 (for example, 4-core 16G,8-core 32G). It can exchange more memory space at a lower cost, and is very suitable for running Java and Node.js game backends with large memory footprint, or deploying local Redis cache nodes.
🚨"Gold Pit Avoidance Guide" for Game Sea Selection"
We must not use "sudden performance type (T series)": when the game business is released or engaged in activities, the traffic will soar instantly. The sudden performance server will instantly run out of points and limit your speed to death, causing all players to drop the line together.
Note Network Throughput Metrics (PPS): Strong interactive games (e. g. MOBA/Shooting) are characterized by extremely small packets but a particularly large number of packets per second (high PPS). When selecting the model, pay attention to the "network packet throughput" of the instance. if PPS is too low, even if the CPU is idle, the bottom layer of the network will jam the packet due to queuing.
The third stage: global standard architecture diagram (actual combat copy operation)
In order to make you more like the landing of this system, here is a set of the most classic "global service, nearby battle" architecture to build the route:
Plaintext
[Player Side]-(Nearby Access)-> [GAAP Nodes Around the World]
│
(tengxunyun backbone network dedicated line intranet is soaring)
│
▼
[Cloud Networking CCN Unified Network]
-------------------------------------------------------------------------------------------------------
▼ ▼
[Tokyo/US-West VPC (Combat Suit C7)] [Guangzhou Core VPC (Hall Suit S6)]
Responsible for local players 50ms speed battle processing recharge, card withdrawal, core MySQL database
Global core: Deploy the global hall service and core account database (MySQL Redis) in the domestic core computer room of Tencent Cloud (such as Guangzhou) or the overseas core hall closest to your R & D team (such as Singapore).
Edge Battle Unit: Deploy pure computing C7 instances as room combat suits in core player gathering places such as the United States and Japan.
High-speed interworking: Connect the Singapore Hall and the global theater through the cloud networking CCN. Players.
When recording, go to the hall. Once the match starts, the hall will dispatch the player to the nearest Tokyo combat suit through the intranet cloud network. After the results of the competition came out, the Tokyo Combat Suit sent the settlement data back to the Singapore Hall through the intranet.
Through this architecture, you not only preserve the unity and security of core data (no need to do complex distributed database synchronization), but also perfectly solve the player's extreme requirements for latency in wartime.
Summary and technical tips for going to sea
The game is not a simple "buy the server overseas", but a fine management of network delay and server computing power. Finally, I will send you four pithy formulas. I wish your game will shine brilliantly overseas:
Regionalization of Action Competition: Don't be lazy in strong interactive games, build computer rooms at more points, and the network delay is greater than the sky.
The strategy card is the same in the world: centralized deployment saves costs, GAAP dedicated lines pull cross-border, and zero packet loss is not a card.
Computational type is used to fight: C7 is selected for room fighting, single-core high-frequency PPS is large, and the number of people can be lower.
The standard type is left to the hall: S6 is used for logging in and drawing cards, with a ratio of 1 to 4 equipped with large memory, which is easy to use, cheap and worry-free.
