- ページ 29

サーバー Cisco UCS S3260 M5のPDF 設計マニュアルをオンラインで閲覧またはダウンロードできます。Cisco UCS S3260 M5 35 ページ。 Server with cloudian hyperstore object storage

Design Criteria

Nodes in the DC



Replicated EC

This configuration requires a minimum of two Data Centers (DC). Each DC consists of 3 nodes each. This

supports the minimum data and parity fragments of (2+1) where 1 is the parity. Table 4 lists the default

replicated EC configuration and the default number of nodes per DC.

Default Replicated EC Configuration and Default Number of Nodes

Nodes Total

Each object is encoded into equal parts and parity fragments are replicated on each node. Each DC is a

mirror image. For configurations greater than 2 DC, Distributed EC configuration is recommended. This

configuration mirrors the encoded data and parity fragments to the other data centers in the configuration.

The choice among these three supported EC configurations is largely a matter of how many Cloudian

HyperStore nodes in the datacenter. For a replicated EC configuration, a minimum of 3 nodes per DC are

required.



Distributed EC

Cloudian's Distributed EC solution implements the new ISA-L Erasure Codes that is vectored and fast. ISA-

L is the Intel library containing functions to improve erasure coding.

The Cloudian Distributed Data Center with EC configuration requires a minimum of 3 data centers with 4

nodes each.

Data stored: DC1: 4, Dc2: 4, DC3:4, Metadata stored: Data stored: DC1: 4, DC2: 4, DC3:3

Distributed EC configuration offers the same level of protection as the replicated EC configuration with 50%

less storage. The Distributed EC configuration is recommended if number of DCs involved are 3 or more.

Flash Storage

Flash Storage with SAS SSD's or NVMe's are used to store metadata for faster performance. The standard

capacity requirement for Flash are less than 1 percent of the total data capacity. Standard design also calls for

having a ratio of 1 SSD for 16 HDD. When using NVMe, one PCIe card is sufficient for all the drives in the S3260

systems.

For example, a dual node configuration has 28 disk slots for each node, so 2 SSD's are recommended. Note that

using 2 SSD's reduces the number of slots available per node in S3260 to 28. However, using 1xNVMe as

mentioned in the BOM above provides the additional 2 slots when compared with SSD's. This slightly increases

TB/RU of the servers.

DC1

9+3

12+4

DC2

2+1

4+2

6+2

8+2

9+3