Huawei cloud gaussdb (for redis) unveiling issue 19: gaussdb (for redis) comprehensive comparison with CODIS

---

• gaussian Redis database ：bbs.huaweicloud.com/blogs/248875

---

One 、Codis How hot it is ？

say Codis It is the most popular self built in China Redis Cluster solution , Not at all . The customer once said ：

“Codis Don't you know? ？ Our company has used it for many years , Good stuff .”

 

You bet , Compared to open source with decentralized architecture Redis Cluster,Codis such proxy Agent clusters are obviously more “ To use ”, Even even RedisLabs Officials have launched similar functions redis-cluster-proxy To make up for tradition Cluster shortcoming .

But in fact, from N Years ago ,Codis There is no one to maintain the community , Our client project manager, Mr. Haipeng, often helps clients build their own projects Codis Moving to the cloud GaussDB(for Redis).

.

Codis It's really excellent , But we can responsibly say ：

Hua Wei Yun KV database GaussDB(for Redis) Better ！

 

Two 、GaussDB(for Redis) vs Codis： Architecture Competition

First , Let's look at the difference between the two products from the perspective of technical architecture .

Pictured , The differences in architecture are clear at a glance , We can analyze it from the top down ：

1、 Non-standard SDK vs standard SDK

Public cloud users generally hate “ Make special ”, However Codis We have to do something special .Codis Kidnapping users must use non-standard Jodis SDK To solve the single point of failure of the agent layer . But if users prefer to use standards SDK（ As famous as Jedis） Well ？Codis Express ： I'm sorry , Say goodbye to Gao Ke .

contrary ,GaussDB(for Redis) It is very friendly in use , The business side uses any standard SDK Will do , At the same time, it is highly available 、 Automatic load balancing capabilities are all in . When moving to the cloud ,GaussDB(for Redis) Compatibility is also very mature , No matter what the customer's original code is StandAlone, still Cluster、Sentinel Equal mode , Fine 0 Transform the cloud , Really use it out of the box , Excellent experience .

 

2、 No load balancing vs Professional ELB

Codis Although the agent layer of provides horizontal expansion capability , But it's not easy to use ： Or sacrifice high to use standards SDK Connecting a single point ; Or use Jodis SDK, But the price is that the business is going to be non-standard .

and GaussDB(for Redis) Provide ELB, It is convenient for users to access uniformly , While simplifying client complexity , Also brought Professional load balancing capability . With this unified access IP, Users are equivalent to using a single machine with super large capacity Redis, You don't need to worry about the bottom details .

 

3、 Storage and calculation coupling vs Deposit is separate

It can be said that ,Codis It is in multiple sets Redis A layer of packaging is made on the chip , The architecture model is the traditional memory computing coupling . At large key、 Data consistency 、 Slice failure 、 In typical scenarios such as expansion and contraction , Both continue open source Redis Long term “ Pain points ”.

GaussDB(for Redis) Adopt the architecture of separation of storage and calculation , The computing layer focuses on data services , The storage tier is based on high-performance distributed shared storage pools , Make sure The storage of three copies of data is strongly consistent , Ultra high reliability . At the same time, the hardware cost is greatly saved , Help reduce business burden .

 

3、 ... and 、GaussDB(for Redis) vs Codis： Feature competition

From the comparison of architecture dimensions mentioned above, we can see ,GaussDB(for Redis) There are ELB Let the business side “ Better to use ”, There is a storage pool for data “ More reliable ”, Compared with self built Codis The cluster is too good .

 

Let's talk about it from the feature dimension ,GaussDB(for Redis) What value can it bring to the business side .

1、 Compatibility

GaussDB(for Redis) Fully compatible with Redis agreement , Compatible with StandAlone/Cluster/Sentinel Many other kinds SDK Pattern , Facilitate business relocation to the cloud . and Codis The compatibility of is very poor .

2、 stability

build by oneself Codis Cluster stability strongly depends on open source Redis Ability , Whether it is expansion or expansion key scene , All have the same pain points , No more details here .

GaussDB(for Redis) The self-developed architecture completely solves the problem of open source Redis Of fork problem , Greatly improved stability ; meanwhile , Under the deep cooperation with the distributed storage pool Architecture , The reliability of the data is also better than Codis.

 

3、 cost

Codis It is still traditional pure memory Redis Encapsulation , The business will touch the cost pain point if it does not save much . With the long-term development of the business , Gradually, there will be data cooling （ Such as APP Cold users of ）, here Codis The proportion of really valuable thermal data in the market is getting lower and lower , The cost of full data in memory can be quite high .

GaussDB(for Redis) The calculating power of is selected according to needs , Low storage prices , Very suitable for large capacity KV Storage scenarios .

 

Four 、 summary

Codis It was once favored by many students in development and operation , It is the memory of the times , however , Technology is developing rapidly . ad locum , It is suggested to use Codis You can consider using GaussDB(for Redis), Feel the enterprise class of huaweiyun KV The new value of databases .

 

5、 ... and 、 appendix

• The author of this article ：

Huawei cloud database GaussDB(for Redis) The team

• Hangzhou / Xi'an / Shenzhen resume delivery ：

[email protected]

• More product information , Welcome to the official blog ：

bbs.huaweicloud.com/blogs/248875