Shulou(Shulou.com)06/01 Report--

In this issue, the editor will bring you about how to analyze the Western Digital ActiveScale architecture. The article is rich in content and analyzes and narrates it from a professional point of view. I hope you can get something after reading this article.

This Western Digital ActiveScale is also a historical product, which comes from Amplidata acquired by HGST.

Amplidata, formerly known as B-Virtual, was founded in Belgium in 2008 and moved to California in 2010.

It is said that the erasure code scheme of object storage has been improved through Cleversafe and Amplidata.

In addition, according to people in Western Digital, in 2015, IBM won Cleversafe ($1.3 billion) after a failed bid for Amplidata, while Western Digital / HGST bought Amplidata (the purchase price was not disclosed).

In 2016, Western Digital refactored the code of HGST Active Archive System and launched a brand new ActiveScale 5.0.

The cost of code refactoring is that Active Archive System cannot be smoothly upgraded to ActiveScale and requires a forklift upgrade.

ActiveScale does not sell pure software, the storage node of the P100 system is a disk-control integrated system, and the X100 is a separate (high-density) architecture.

ActiveScale supports scale-up (only extending storage nodes to 3 groups of 6 nodes) and scale-out (expanding system nodes to 3 groups of 3 nodes each).

In China, Ziguang Western Digital Joint Venture Company OEM provides object storage for Western Digital. The model correspondence is as follows:

Ziguang western number P2000 = western number P100, Ziguang western number X20000 = western number X100.

On September 19, 2019, Xisu announced that it had sold IntelliFlash to DDN, while ActiveScale was looking for a buyer. The future of Ziguang West Digital is uncertain, which will greatly affect the possibility of Chinese customers adopting its object storage products.

ActiveScale software function is relatively perfect, often promote the concept of symbiotic design (Symbiotics Design), claiming to be the best understanding of hard disk object storage.

18A5 means that 13 data blocks generate 5 check blocks, which can support 5 hard disk failures without data loss.

Highlights: support for 3 sites to correct deletion. The 2 site is bucket asynchronous replication (bi-directional), or one-way asynchronous replication to AWS. Three site erasure correction and bucket replication can be used at the same time.

The system is divided into two layers: system node / access node (3 nodes) and storage node (6 nodes).

This kind of networking access has high requirements for the network, which affects the system performance, and the general object storage is seldom used.

UDA only supports NFS v3, but other protocols require gateway support (SMB and iSCSI).

It is important to note that it is not any container application, and some of the selected applications need to be verified.

There is a small file packaging function, support to define how big a file (like 0-8MB) belongs to a small file.

Data persistence originally promoted 17 9s, but now it has promoted 19 9s.

As we can see, the function is still relatively rich.

ActiveScale supports the popular cloud analysis functions, and conventional cloud analysis functions, including historical data analysis, capacity prediction, active operation and maintenance, and so on.

ActiveScale advertises its point of differentiation, but in fact, most of the features are available in other object stores.

The latest version 5.5 was released in 2019, with major upgrades to both hardware and software.

The original 12TB hard disk is now supported by 14TB hard disk, so the system capacity has been improved.

The new hardware bandwidth has also been improved, and the bandwidth is advertised to reach 75GB/sec, which is claimed to be the best level in the HDD object system industry.

File performance has also improved, especially for large files.

The three sites finally support asynchronous mode, which can support long-distance deployment and reduce latency requirements.

This feature integrates storage into your workflow. ActiveScale notifies messaging services, such as ApacheKafka ®, of object events, such as creating a new object. This allows Kafka to start an application and then perform operations on the new object for real-time processing. This is similar to NetApp's StorageGRID platform service.

S3 users have the same experience as Amazon S3 management and do not need to learn.

Finally, let's take a look at the third-party comments.

In the IDC report, Western Digital is only the mainstream player. In the Gartner Magic Quadrant, Western Mathematics is in the Challenger Quadrant.

Western Digital ActiveScale as a whole is at the global mid-stream level and is getting farther and farther away from IBM Cleversafe (COS).

There are several evaluations of Gartner's key capabilities for object storage:

1.ActiveScale has a good hardware management function.

two。 Contribute S3A plug-in code and have a deep understanding of big data

3. Support for containers to run ISV applications

4. Upgrading from the old platform requires a forklift upgrade

5. The document is not very mature, so it is recommended to conduct a comprehensive test before use.

Western math ActiveScale does not have a highest score, storage efficiency score is not high, management score is good.

The score of Western math ActiveScale is more balanced in each scene, and there is no obvious deficiency.

ActiveScale, like object storage in EMC, Hitachi and NetApp, has a long history. Although it has a historical burden, its function is still relatively comprehensive. In particular, I think the following features are quite distinctive:

Cross-site EC

Bucket two-way asynchronous replication: 2 sites

Bucket one-way asynchronous replication: on the cloud

Cloud analysis

Container support

Data pipeline service

High density all-in-one machine

However, compared with other newer object storage products, I think the following features are still lacking:

Pure software delivery is not supported, and many users need software decoupling when choosing object storage.

Copies are not supported

The extended cluster is not supported, and the two sites cannot achieve RPO=0

SSD is not used for Cache, and small file merging is not accelerated by SSD

There is no attribute reverse check feature

Mutual access between files and objects is not supported

There is no dedicated HDFS client, and the performance of S3A docking Hadoop is average.

Multi-site only supports 3 sites, a little less

The above is the editor for you to share how to analyze the Western Digital ActiveScale architecture, if you happen to have similar doubts, you might as well refer to the above analysis to understand. If you want to know more about it, you are welcome to follow the industry information channel.

Welcome to subscribe "Shulou Technology Information " to get latest news, interesting things and hot topics in the IT industry, and controls the hottest and latest Internet news, technology news and IT industry trends.