A brief evaluation of the "bare metal cloud" of raksmart data center Hong Kong (that is, bare metal server)

raksmart, an old computer room, has launched the "bare metal cloud" business in data centers in San Jose, Los Angeles, Hong Kong, and Japan, which is what we commonly call bare metal servers. So what is the situation of raksmart's bare metal cloud in Hong Kong? Let’s take a look at it. The evaluation data is for reference only.

Official website: https://www.raksmart.com

U.S. company, no real name required, bilingual in Chinese and English, payment via PayPal, Alipay, WeChat, credit card, Bitcoin, etc.

Bare metal servers are 30% off, discount code: Bare-Metal Cloud-30%dis

Hong Kong bare metal cloud features: default 10M bandwidth (premium network, that is, cn2 mixed bgp), unlimited traffic, comes with an IPv4, system disk and data disk are separated, default 100G system disk + 1T SSD data disk;

Value-added items: Bandwidth can be expanded to 100M, IP maximum 64, memory maximum 256G, data disk maximum 10T SSD

Here we have a single-channel e5-2697 (first generation) bare metal cloud for evaluation, but what we got is a second generation CPU. Pay attention to the OS. It is obvious that it is marked KVM instead of dedicated (dedicated). ), the hard disk I/O is approximately 127MB/S:

Change FIO to test the hard disk data reading and writing in more detail:

The default bandwidth is 10Mbps. We used the three network nodes in mainland China of speedtest.net to test how much the uplink and downlink bandwidth can reach and what the data is. It basically meets the official default parameter of 10Mbps:

Test the effect during the evening rush hour. As long as there are no detours in Hong Kong lines and the bandwidth is sufficient, the network effect is actually not much different from that during the day:

Tests on other nodes in Asia (excluding mainland China, tested above):

iperf3 runs on European and American conditions:

The files stored on the test machine were downloaded directly from the Chrome browser during the evening peak hours by Guangzhou Telecom at the following speed:

The same environment remains unchanged, but changed to leisure time during the day, the bandwidth is nearly full

This still shows that the effect in a personal network environment will change with the changes in the backbone network. This is not caused by insufficient bandwidth.

Those who are concerned about latency can take a look. This is the test data of more than 60 nodes in mainland China provided by ping.chinaz.com:

For the outbound journey of China Telecom, take CN2:

China Unicom outbound, China Unicom backbone transfer to CN2:

Move outbound, own CMI:

Return to China Telecom: Pass 59.43, that is, come back from CN2

Return to China Unicom: Return directly from Hong Kong Unicom node

Return, move to the mainland: come back directly from the Hong Kong mobile node

Unix bench Mark：

Test the streaming situation. Since the IP is broadcast, don’t worry about it:

To summarize:

First: The network is well optimized. China Telecom uses two-way cn2, and China Unicom and China Mobile are directly connected to each other on the outbound route. The return trip is directly from the operator's node in Hong Kong. There are few nodes and the direct connection is in place.

Second: Heavy performance loss. From the Unix bench, we can see that a single-channel e5-2697v2 can only run more than 3600 points. This tells us about the performance loss during the virtualization process... If the tasks you run have high CPU requirements, it is best to buy the official raksmart Hong Kong dedicated servers instead of bare metal clouds.

Official website: https://www.raksmart.com

Recommended site searches: virtual space host, vps host, instant solution server, PHP space purchase, external network IP, Ministry of Industry and Information Technology registration system distribution host, US virtual host purchase, Hong Kong vps host rental, IP reverse domain name check,