Disk Type and Performance
eSurfing Cloud classifies EVS into the following five specifications by I/O performance:
common I/O, high I/O, generic SSD, ultra-high I/O, and fast SSD. You can select the specification that best fits your business needs and budget.
EVS Type and Performance
EVS performance metrics include: IOPS, throughput, and read/write latency.
· IOPS: The number of read/write operations performed by an EVS disk per second. IOPS can be broken down into the maximum IOPS of a single disk, baseline IOPS, etc.
· Throughput: The amount of data successfully transmitted by an EVS disk per second, that is, amount of data read from and written into an EVS disk per second. Generally the maximum throughput of an EVS disk and the throughput per GB of an EVS are measured.
· Read/Write latency: The latency time required by an EVS to process one read/write I/O. This performance is most intuitive for users.
Parameter | Common I/O | High I/O | Generic SSD | Ultra-high I/O | Fast SSD |
Maximum IOPS of a single EVS | 2,000 | 5,000 | 20,000 | 50,000 | 100,000 |
Baseline IOPS of a single EVS | 300 | 1,800 | 1,500 | 1,800 | 1,800 |
IOPS per GB of EVS | 2 | 8 | 8 | 50 | 50 |
Formula for calculating IOPS | min(2000, 300+2*capacity) | min(5000, 1800+8*Capacity) | min(20000, 1500+8*Capacity) | min(50000, 1800+50*Capacity) | min(100000, 1800+50*Capacity) |
Max. throughput (MB/s) | 150 | 200 | 250 | 350 | 750 |
Throughput per GB of EVS (MB/s) | 0.1 | 0.1 | 0.5 | 0.5 | 0.5 |
Throughput formula (MB/s) | min(150, 100+0.1*Capacity) | min(200, 130+0.1*Capacity) | min(250, 100+0.5*Capacity) | min(350, 120+0.5*Capacity) | min(750, 200+0.5*Capacity) |
Single-queue access latency | 5~10 ms | 1~3 ms | 1 ms | 1 ms | Sub-millisecond |
Features | Less commonly accessed workloads | Commonly accessed workloads | Enterprise offices with high throughput and low latency | Enterprise critical business, high throughput, low latency workloads | Ultra large bandwidth and super low latency |
Typical Scenario | ž Applications demanding large capacity, medium read/write speed, but having fewer transactions ž Daily office applications ž Lightweight development and test environments ž Not recommended to be used as system disks | ž General development and test environments | ž Enterprise office Medium-scale development and test environments ž Small and medium-sized databases ž Web applications ž System disks |
ž Transcoding services ž I/O-Intensive scenarios ž NoSQL, Oracle, SQL Server, PostgreSQL ž Latency-sensitive scenarios ž Redis/Memcache | ž Enterprise office, virtual desktop ž Large-scale development and test environments ž Transcoding services ž System disks ž Large and medium-sized databases (SQL Server, Oracle, NoSQL, PostgreSQL) |
Note:
1. The resource pool with the maximum IOPS 5,000 of ultra-high I/O is Hong Kong 2. The maximum IOPS supported by other resource pools is 33,000. Formula for calculating IOPS of a single EVS is: min(33000, 1800+30*capacity).
2. The maximum IOPS and maximum throughput of a single EVS are all calculated based on the sum of read and write operations. For example, maximum IOPS = read IOPS + write IOPS.
3. Take the throughput formula of a single ultra-high I/O EVS for example: The baseline throughput is 120 MB/s. The throughput increases by 0.5 MB/s for every one GB added until it reaches the maximum throughput 350 MB/s.
4. Take the IOPS formula of a single generic SSD EVS for example: The baseline IOPS is 1,500. The IOPS increases by 8 for every one GB added until it reaches the maximum IOPS 20,000.
5. A single queue indicates that the queue depth or concurrency is 1. The single-queue access latency is the I/O latency when all I/O requests are processed sequentially. The values in the table are calculated with 4 KB data blocks.
EVS Performance Calculation Formula
The following uses the IOPS formula as an example. IOPS of a single EVS = minimum value of maximum IOPS of a single EVS and baseline IOPS of a single EVS + IOPS per GB of EVS * EVS capacity. Wherein, maximum IOPS of a single EVS is the maximum IOPS that a single EVS can reach, that is, the sum of read and write IOPS. Baseline IOPS of a single EVS is the baseline IOPS that a single EVS can obtain. IOPS per GB of EVS is the IOPS that can be obtained for each increase of 1GB of EVS capacity in addition to the baseline IOPS.
The following uses the Hong Kong 2 ultra-high I/O EVS as an example. The maximum IOPS of a single ultra-high I/O EVS is 50,000, the baseline IOPS of a single ultra-high I/O EVS is 1,800, and the IOPS per GB of EVS is 50.
· If the capacity of the ultra-high I/O EVS is 100 GB, the IOPS of the EVS is calculated as min (50000, 1800+50*100). The IOPS of the EVS is 6,800, the smaller value between 50,000 and 6,800.
· If the capacity of the ultra-high I/O EVS is 2,000 GB, the IOPS of the EVS is calculated as min (50000, 1800+50*2000). The IOPS of the EVS is 50,000, the smaller value between 50,000 and 101,800.
The performance of an EVS is closely related with the data block size: For a certain identified data block size, when either the maximum IOPS or throughput of a single EVS reaches the maximum limit, the EVS reaches the maximum performance and the other indicator cannot be increased. Generally speaking, for data blocks in small size, such as 4 KB or 8 KB, the EVS can reach the maximum IOPS, but not the maximum throughput. For data blocks in large size, 16 KB or more, the EVS can reach the maximum throughput, but not the maximum IOPS. The following table takes generic SSD as an example to describe the relationship between EVS performance and data block size.
Data Block Size | Max. IOPS | Max. Throughput |
4 KB | About 20,000 | About 78 MB/s |
8 KB | About 20,000 | About 156 MB/s |
16 KB | About 16,000 | About 250 MB/s |
32 KB | About 8,000 | About 250 MB/s |