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

Implementation of Emc Storage Virtualization in Shuanghuo data Center

I. deployment

Vplex actual connections are divided into three types of lines:

1. Management port: Ethernet, which is used to manage and establish IPSEC encrypted tunnel with witness

2. COM control line: transmission control signaling

3. FC data line: transfer IO data

The deployment scenario is to build the virtualization of dual-live data storage in the same city, and the storage manufacturer recommends that the link delay between data centers is within 3ms. The measured length of the bare fiber in the data center is 48km, in which the bare fiber transmits about one-way delay 0.32ms after passing through the wavelength division equipment.

Vplex uses write-through mode, only read cache, and the array uses write cache and write-back mode. Set the high priority of Center A VPLEX.

Second, test scenario

1. Scenario test 1

After the whole series of back-end failures and the interruption of FC lines in the back-end connection array of san network, the service can fail over, and the measured transfer time is about 5s.

2. Scene test 2

When the power failure of two SAN directors in a single center, the storage of the dual data center is all hung up, and the service is interrupted throughout the network.

Third, the analysis of problems.

1. IO stream

The EMC engineer explained that the IO stream means that the host sends the IO request to VPLEXA,VPLEXA and forwards the IO request to both the local entire column and the peer data center VPLEXB. At this time, the VPLEX in data center A writes the IO to the full column write cache and writes back an acknowledgement to VPLEXA. VPLEXA receives the write back confirmation and waits for the VPLEXB write back confirmation, and then forwards it to the application for write back confirmation; data center B vplexB writes the IO to the B center array write cache, writes back to vplexB, and vplexB writes back to VplexA for confirmation.

2. Fault analysis

When the san director of data center A fails, as shown in the above figure, when witness fails in center A, it does not judge that the vplex backend fails. In this case, the vplex of AB data center survives, and the VPLEX of A data center is a high priority. In this case, the storage of AB data center will be hung up, and one of the data centers will be manually designated as the main storage, and incremental synchronization will be carried out. Then pull up virtualization, databases and applications.

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.