Common questions about enterprise-level, entry-level, persistent memory-optimized, u1, economy e, ECS Bare Metal, and SCC instances.
FAQ about enterprise-level instances
-
What are enterprise-level instances? What are entry-level instances?
-
What is the fundamental difference between enterprise-level and entry-level instances?
-
Which instances are enterprise-level? Which are entry-level?
-
In which business scenarios should I purchase enterprise-level instances?
-
What is the network performance of enterprise-level instances?
-
What are the limits for upgrading or downgrading the specifications of enterprise-level instances?
-
Can I upgrade an entry-level instance to an enterprise-level instance?
Questions about persistent memory-optimized instances
FAQ about general-purpose compute-optimized u1 instances
FAQ about economy e instances
FAQ about ECS Bare Metal Instances
FAQ about SCC
Appendix
What are enterprise-level instances? What are entry-level instances?
Enterprise-level instances are instance families launched by Alibaba Cloud in September 2016. They provide dedicated computing, storage, and network resources with high performance, stable computing power, and balanced network performance, making them ideal for enterprise scenarios that require high stability.
Entry-level instances are instance types and families designed for small and medium-sized websites or individual users. They use shared resources, so computing performance is not guaranteed to be stable, but they are more cost-effective.
What is the fundamental difference between enterprise-level and entry-level instances?
Enterprise-level instances use a fixed CPU scheduling mode where each vCPU is bound to a physical CPU hyper-thread, preventing resource contention between instances and ensuring stable computing performance with a strict SLA.
Entry-level instances use a non-bound CPU scheduling mode where each vCPU is randomly allocated to any idle CPU hyper-thread. vCPUs from different instances compete for physical CPU resources, which can cause unstable computing performance during high-load periods. Entry-level instances have an availability SLA but no performance SLA.
Which instances are enterprise-level? Which are entry-level?
Among the currently available instance families, the entry-level ones are e, t6, t5, s6, n4, mn4, xn4, and e4. All others are enterprise-level.
In which business scenarios should I purchase enterprise-level instances?
See Instance families and Best practices for instance type selection.
What is the network performance of enterprise-level instances?
Network performance depends on the instance type — larger specifications provide higher performance. See Instance families for details on each instance type.
What are the limits for upgrading or downgrading the specifications of enterprise-level instances?
Can I upgrade an entry-level instance to an enterprise-level instance?
If I use persistent memory as memory, can I directly deploy Redis applications on a persistent memory-optimized instance?
Running Redis on persistent memory-optimized instances significantly reduces TCO per GiB of memory. However, you must modify your Redis application for optimal performance by storing non-hot spot data in persistent memory and hot spot data in standard memory.
To minimize modification costs, the re6p instance family provides Redis-specific instance types. See Deploy a Redis application on an instance with persistent memory.
When you purchase the instance, select an instance type with a name in the ecs.re6p-redis.large format.
I have a Redis application cluster that uses standard memory. How can I migrate my services to persistent memory-optimized instances? What should I note?
Prioritize business stability and data reliability during migration. Start with a single persistent memory-optimized instance, run a small number of services, and direct traffic to it for testing. Verify that the performance and capacity model meets your expectations, then gradually increase the number of persistent memory-optimized instances until migration is complete.
If I use persistent memory as memory, can I directly deploy my parameter server (PS) on a persistent memory-optimized instance?
PS server nodes store all training parameters for a cluster. Traditionally, storing all parameters in memory requires large amounts of memory and is expensive. With a persistent memory-optimized instance configured to use persistent memory as memory, you can store all parameters in persistent memory and keep only the hash table in standard memory, significantly reducing TCO.
After I purchase a persistent memory-optimized instance, how do I configure the persistent memory to be used as a local disk?
Which of my applications require higher-performance local disks?
Persistent memory-optimized instances can resolve common performance and cost issues for I/O-intensive applications, such as:
-
High latency for a single SQL query or the need for more stable response time (RT).
-
Slow resource loading in application scenarios such as game frontends, high-load databases, and high-load web applications.
-
Unnecessary costs from purchasing large-capacity cloud disks or local disks for higher IOPS or bandwidth when the extra capacity remains idle.
Typical I/O-intensive scenarios where persistent memory-optimized instances balance performance and cost include:
Redis and other NoSQL databases such as Cassandra and MongoDB
Structured databases such as MySQL
I/O-intensive applications such as e-commerce, online games, and media applications
Search scenarios that use solutions such as Elasticsearch
Live video streaming, instant messaging, and room-based online games that require persistent connections
High-performance relational databases and OLTP systems
If I use persistent memory as a local disk, can I directly deploy Redis or MySQL applications? Do I need to modify the applications as I would if I used persistent memory as memory?
No, application modification is not required when you use persistent memory as a local disk. Redis and MySQL applications detect the persistent memory as a standard SSD.
When persistent memory is used as a local disk, how does its performance compare to local NVMe SSDs and cloud disks?
To run a test:
-
Configure persistent memory as a local disk and mount it.
-
Test disk performance with a benchmarking tool.
For fio-based Linux disk performance testing, see the test commands in Test the performance of a block storage device.
How reliable is persistent memory?
Data reliability in persistent memory depends on the physical server and the persistent memory device, creating a risk of a single point of failure. Implement data redundancy at the application layer and store data that requires long-term persistence on cloud disks.
When you release a persistent memory-optimized instance, data in persistent memory is automatically purged. Back up your data before releasing the instance. Because data purging is time-consuming, releasing a persistent memory-optimized instance takes longer than releasing a standard instance.
What are the target users and business scenarios for u1 instances?
u1 instances target small and medium-sized enterprises and suit most general-purpose scenarios without high CPU performance requirements, such as web applications, websites, enterprise office applications, offline data analytics, and small and medium-sized databases.
Which business scenarios are not suitable for u1 instances?
-
Scenarios that require high computing power and performance, such as games and high-frequency trading. Use c7, g7, or r7 instances instead.
-
u1 instances run on multiple server platforms with potential performance differences between platforms. For scenarios that require consistent performance, such as HA setups where primary and standby servers must perform identically, use c7, g7, or r7 instances instead.
Which processor models do u1 instances support?
u1 instances support Intel® Xeon® Platinum 6149 (Skylake), 8163 (Skylake), 8269CY (Cascade Lake), 8260 (Cascade Lake), 8369HB (Cooper Lake), and 8369HC (Cooper Lake) processors.
Do u1 instances support specifying a processor model?
No, they do not.
What are the base and turbo frequencies of the processors for u1 instances?
u1 instances run on server platforms with varying computing power. CPU base and turbo frequencies depend on the host's CPU model. Because an instance may be migrated to different CPU platforms during its lifecycle, the actual frequencies may change.
What does it mean that a u1 instance may be migrated to different server platforms during its lifecycle?
During its lifecycle, a u1 instance may be migrated to a new host due to host upgrades or O&M. The new host may have a different processor. All instance types can experience host migration. See General-purpose compute-optimized instances (u1) for details on potential processor changes.
Which business scenarios are not suitable for economy e instances?
Enterprise-level applications that require high performance stability and a committed performance SLA.
What are the CPU models for economy e instances? Do they support specifying a processor model?
e instances use Intel® Xeon® Platinum scalable processors and do not support specifying a processor model.
e instances run on mainstream, cost-effective CPU platforms and may be migrated to different platforms during their lifecycle. Technical measures ensure business compatibility between platforms.
What are the base and turbo frequencies of the processors for economy e instances?
The base frequency is 2.5 GHz. Because e instances may be migrated to different CPU platforms during their lifecycle, the turbo frequency is not guaranteed.
What does it mean that an economy e instance may be migrated to different server platforms during its lifecycle?
During its lifecycle, an economy e instance may be migrated to a new host due to host upgrades or O&M. The new host may have a different processor.
Alibaba Cloud's resource pooling technology and intelligent scheduling algorithms handle dynamic resource management, providing continuous computing power, stability, supply, and elasticity for your applications.
What are the fundamental differences among ECS Bare Metal Instances, traditional cloud servers (virtual machines), and traditional physical servers?
What is the network performance of ECS Bare Metal Instances?
Network performance depends on the instance type — larger specifications provide higher performance. See Instance families for details on each instance type.
Which disk types do ECS Bare Metal Instances support? How many data disks can be attached?
ECS Bare Metal Instances support enhanced SSDs (ESSDs), standard SSDs, and ultra disks. You can attach a maximum of 16 data disks.
Do ECS Bare Metal Instances support specification upgrades/downgrades and failover migration?
ECS Bare Metal Instances do not support specification upgrades or downgrades but support failover on hardware failure. All data is preserved on cloud disks.
In the Windows 10 Professional Edition OS on an ECS Bare Metal Instance, why is the number of logical processors displayed less than the number of vCPUs in the instance type?
Some ECS Bare Metal Instance types have more than two physical CPU chips, but Windows 10 Professional Edition supports only two. Switch to a Windows Server public image from Alibaba Cloud to use all CPU chips.
How do I use the RDMA feature of SCC?
When creating an SCC instance, select an SCC-specific custom OS image that supports the RDMA over Converged Ethernet (RoCE) driver and the OFED stack. Use RDMA through IB verbs programming or MPI.
