NFVi (Network Function Virtualization Infrastructure) is the foundation for carrying VNF (Network Network Element Function). Its core is to support multi-tenancy by using standardized virtualization technology to provide resource support for different types of virtual network elements on demand. Through research, it is found that a disk array with an SSD (Solid State Drive) cache can achieve better delay and IO performance, and is suitable for service scenarios in which data forwarding is mainly used in NFV, and data capacity expansion is not high, such as vCPE. vEPC, vBRAS, vIMS, etc. The IP SAN-based disk array is more compatible and maintainable, and can be used for the entire chassis server that carries the NFVi service. For business scenarios with large data volumes and high data growth requirements, Server SAN can achieve better scalability and cost. Among them, Ceph is very suitable for scenarios where performance requirements are not high and versatility is higher.
Disk Array
A disk array is a redundant array of independent disks that provides a single, large storage device solution. Disk arrays combine multiple physical disks into one larger virtual device. While increasing the capacity of the hard disk, it can also increase the speed of the hard disk, make the data more secure, and make disk management easier. SAN (Storage Area Network) uses Fibre Channel technology to connect storage arrays and server hosts through Fibre Channel switches to establish a regional network dedicated to data storage. After more than a decade of development, SAN has matured and become the de facto standard in the industry. The bandwidth used for SAN storage has grown from 100MB/s and 200MB/s to the current 1Gbit/s, 2Gbit/s, 8Gbit/s, and 10Gbit/s.
SAN structure
NFVi's disk array SAN architecture mainly includes FC SAN and IP SAN. FC SAN is a disk array (RAID) connected to other devices through the FC channel protocol. The FC protocol actually solves the underlying transport protocol, and its high-level protocols still use SCSI, and data processing is "block level." IP SAN utilizes ISCSI (Internet Small Computer Port) technology based on the High Speed ​​Ethernet Protocol (TCP/IP protocol) to implement SCSI protocol over IP networks, enabling fast data access on high-speed Gigabit Ethernet. Backup operation. The advantages and disadvantages of FC SAN and IP SAN are as follows.
1. In terms of cost, FC SAN requires dedicated FC network card, FC switch and other special equipment, and the purchase cost is high; IP SAN cost is low, the layout is simple, the management cost is low, and the purchased network cable and switch are all Ethernet, or even Use existing networks to build SANs.
2. In terms of transmission distance, the FC SAN is limited by the fiber distance, and the IP SAN is based on the inherent advantage of the IP network, so that it has no distance limitation, and it is easy to implement technologies such as off-site storage and remote disaster recovery through the WAN.
3. In terms of transmission speed, FC SAN can achieve high-performance data storage, high bandwidth utilization, small transmission delay, high transmission efficiency, and the actual bandwidth can reach 85% to 90% of the theoretical bandwidth. The IP SAN needs to share the data channel, the transmission delay is high, and the network environment is greatly affected. The actual bandwidth can reach 20% to 30% of the theoretical bandwidth.
4. In terms of security, FC SAN adopts separate Fibre Channel independent networking and has high security. The IP SAN accesses the IP network with relatively low security.
5. Compatibility, although FC SAN has the relevant standard FC protocol, FC devices from different vendors are not compatible. The IP SAN is based on the TCP/IP protocol and has good compatibility.
Based on the above comparison, FC SAN products are often used in enterprise applications such as core databases with high performance and stability requirements. However, IP SAN products can be considered when performance requirements are not high, especially in remote disaster recovery scenarios. For scenarios where the entire cabinet server is used as the NFVi bearer, it is more convenient to use the IP switch for service and data storage exchange.
Storage medium
With the development of CPU and memory performance, traditional mechanical hard disk (HDD) has become the main bottleneck of IT systems. The advent of SSDs has increased the IO performance of hard drives by orders of magnitude. The application of SSD has further developed disk arrays.
1. Traditional disk array: In the traditional disk array, the SAS disk is mainly used as a high-speed storage medium. The IO performance of the disk array mainly depends on the number of SAS disks, the processing capacity of the backplane, and the cache. Parallel read and write of a large number of SAS disks can increase the speed of the disk array.
2. Hybrid array: SSD price is higher than HDD, and hybrid array with SSD+HDD is a compromise solution, which can achieve superior performance compared with traditional disk array without greatly increasing cost. Meet growing capacity needs. According to the role played by SSDs, hybrid arrays can be divided into automatic tiering and cache acceleration.
3. All-flash array: All SSDs are used as storage media in all flash disk arrays. They are suitable for applications with low capacity requirements and very demanding performance. The all-flash array can be divided into two types according to the array control software, one is AFA (All Flash Array), which is optimized on the basis of the original disk array, for example, the function of prohibiting disk defragmentation; For SSA (Solid-State Array), it is completely rewritten according to the characteristics of SSD.
Server SAN
Server SAN is a storage resource pool consisting of multiple independent server storages with good cost performance and scalability. The traditional storage network is mainly based on FC, and the transmission bandwidth and delay are far superior to Gigabit Ethernet. However, with the introduction of 10Gbit/s, 40Gbit/s Ethernet and Infiniband RDMA network technology, storage has become a system performance bottleneck. Server SAN can be used to achieve unified computing and storage networks. SSDs with high I/O performance, extremely low latency and high bandwidth make Server SAN performance comparable to high-end storage. Server SAN storage is currently mainly used by Internet companies for research and development, namely Hyperscale Server SAN, and has gradually entered the IT field of the communications industry.
1.VSAN
VMware VSAN (Virtual SAN) is a scalable distributed storage architecture developed by VMware based on VSphere. VSAN builds a VSAN on top of SSDs and HDDs directly connected to the VSphere cluster. These storage devices are uniformly controlled and managed by VSAN, using flash SSD acceleration, which can provide large storage space with ordinary mechanical hard disks and high performance through SSD, thus forming a unified shared storage layer with low cost and excellent performance.
2.Ceph
As the most open source storage solution for the OpenStack community, Ceph is also the most widely distributed global open source software definition storage project. Ceph builds the underlying general-purpose storage system by implementing an object storage interface of the POSIX-like interface. The upper layer implements different storage scenarios such as object storage, block storage, and file system storage by stacking different storage standards (such as POSIX, block interface, and S3/Swift). Applications. This approach is also the technical basis for Ceph's unified storage goals. Ceph's advanced architecture combined with SSD SSDs, especially the high performance of high-speed PCIe SSDs, will undoubtedly become a typical scenario for Ceph deployment.
The main goal of NFV is to achieve the separation of software and hardware of network equipment. The overall trend is that the underlying hardware implementation is relatively general. Server SAN technology based on Ceph open source distributed storage is widely supported by NFVi vendors and is more versatile. The use of commercial distributed storage systems can achieve higher performance, but it will affect the versatility of the underlying hardware, subject to the constraints of manufacturers.
3. Other commercial distributed storage systems
Mainstream commercial distributed storage systems include EMC ScaleIO, Huawei FusionStorage, HP StoreVirtual VSA, and Dell Fluid Cache. These mainstream commercial distributed storage systems claim to achieve higher performance than Ceph and provide better after-sales support.
Comparison of disk array and server SAN
After the SSD is introduced into the disk array and the server SAN, the performance has been greatly improved. The following compares the characteristics of both the disk and the server SAN from four aspects.
reliability
Reliability is the primary problem for storage. Traditional disk arrays use dual controllers, multipathing, and disk redundancy to ensure that there is no single point of failure for the disk array. The disk redundancy of the disk array can generally be Raid5, Raid10, etc., to ensure that the disk data can be recovered by verifying the data and other disk data in the case of disk corruption. The Server SAN uses software to ensure reliability. It requires 3 copies to ensure that business data is not affected in the event of disk corruption. Disk arrays can be used for disaster recovery across data centers through mature remote backup and dual-active technologies such as remote mirroring, which is better than Server SAN. In addition, disk arrays are optimized for SSDs in firmware, and SSDs have a longer life. The Server SAN lacks a related mechanism, which may affect the life of the SSD after prolonged use.
Capacity and scalability
Due to the small capacity of the SSD, the price of the single disk is expensive, and the capacity of the all-flash array is small. With Server SAN products, you can achieve PB-level storage space, and as the number of servers grows, its performance and capacity can grow linearly, especially for applications with large data volumes and growing applications. To achieve PB-class capacity, you need to use expensive high-end disk arrays, and you can only purchase disk cabinets of this model, which has high expansion cost and long cycle. In terms of capacity and scalability, the best to the worst are Server SAN, high-end disk arrays, mid-range disk arrays, and all-flash arrays. Therefore, Server SAN has advantages for scenarios where business data grows faster and capacity requirements are larger.
Performance Due to the use of dedicated storage hardware and optimized design for SSDs, all-flash array performance is much higher than other types of storage. In the case of extensive use of SSD cache, mid-range disk array performance has been greatly improved. Commercial distributed storage systems such as EMC ScaleIO and Huawei FusionStorage perform near the midrange disk array using SSD cache. According to published test data, the performance of commercial Server SANs such as FusionStorage is much higher than that of Ceph. In addition, Ceph has a large performance gap in different optimization modes.
cost
Server SAN costs more than 50% of the disk array in terms of hard disk purchase cost. In the case of high disk drive prices, the cost of the Server SAN is relatively high. The disk array needs to purchase special equipment such as control cabinets and disk cabinets that match the original equipment, and the expansion cost is high. The Server SAN can use a general-purpose server, and the expansion cost is low. Therefore, the cost of using a Server SAN is lower when the storage scale reaches the PB level or higher and is still growing.
Operational difficulty
The disk array is mature and stable, and its maintenance is simple, and the daily maintenance workload is small. Ceph is an open source project. Its configuration and performance tuning are complex. The operation and maintenance personnel need to deal with system bugs, have high technical capabilities, and have a large maintenance workload. Commercial distributed storage systems have been commercialized by manufacturers, and they are highly protected by vendors. The operation and maintenance difficulty is less than Ceph, but it is also larger than disk arrays.
NFVi's demand for storage
The application scenarios that VNF ​​is the first to implement include vBRAS, vCPE, vEPC, vIMS, and vSR. Possible future application scenarios include vCDN. According to the requirements of data types, it is mainly divided into the following three categories.
1. There is no need to store service data, such as vCPE, vEPC, vBRAS, vIMS, and vSR, covering the main application scenarios of NFV. The virtual machines in this kind of scenario mainly play the functions of data processing and forwarding. The data does not fall into the disk, and the growth of storage space is not high. Business needs can be met by using SSDs as cached disk arrays. This type can be used with traditional SAN storage.
2. Store structured data, such as vHSS in vEPC. The main storage of structured data in the database has certain data growth requirements. This scenario can be achieved with both disk arrays and Server SANs.
3. Store unstructured data, such as vCDN. This type of data grows faster and requires greater scalability for storage space. This scenario uses Server SAN to have the advantage of cost and scalability.
Overall, for NFVi, disk arrays can achieve better latency performance.
In modern clinical medicine, ventilator, as an effective means to artificially replace the function of spontaneous ventilation, has been widely used in respiratory failure caused by various reasons, anesthesia and respiratory management during major surgery, respiratory support treatment and emergency resuscitation. It occupies a very important position in the field of modern medicine. A ventilator is a vital medical device that can prevent and treat respiratory failure, reduce complications, and save and prolong patients' lives.
With excellent quality and preferential prices, Kable-X (Su Zhou) Technology Co., Ltd. provides services for the wiring harness industry in various fields such as industrial control, medical care, automobiles, and new energy. It has obtained IATF16949\ISO9001\ISO14000\UL\ISO13485\IPC620, etc. Industry certification. At the same time, the company also passed the certification of high-tech enterprises in 2019 and obtained a number of industry patents. It has the ability to integrate wiring harnesses and electronic components and hardware injection molding parts. It actively participates in customers' initial product design selection, and Maintain the greatest flexibility to meet the various needs of customers. Punctual delivery to meet customer needs. Suzhou Kable-X Wire Harness Processing Factory adheres to the business philosophy of "Integrity, Responsibility, Innovation", and perfect after-sales service to create together with customers. A better tomorrow!"
We produce a large number of Medical Cable Assembly, including Ventilator Wire Harness, Medical Defibrillator Wire Harness and Medical Aid Equipment Wire Harness.
Ventilator Wire Harness,Wire Harness Manufacturing,Wiring Harness Price,Electronic Wiring Harness Assembly
Kable-X Technology (Suzhou) Co., Ltd , https://www.kable-x-tech.com