Challenge for the IT administrators
Nowadays, it is obvious, that most organizations require an IT system to operate effectively. Even companies which are not related to IT need data storage that is continuously available, highly performed, and cheap. Thus, IT staff is challenged to build innovative, fast, modern and reliable IT systems that fit budget constraints and save some resources for future growth.
To commence, the IT administrator must clarify the purpose of the system and its role in the company. Also, they should consider several core factors when completing an infrastructure project.
Firstly, hardware should be installed in a separate room with air conditioning and fire protection systems. Electric power should be enough to supply all installed hardware. In other words, the IT system should be safe, quiet and invisible.
Secondly, network equipment, Internet connection and networks themselves should be planned according to their purposes and fit the requirements of the project. At the same time, they should be redundant and should not become a bottleneck for the whole setup. In addition, the IT administrator should be familiar with best practices and needs to follow the vendors’ recommendations.
Third, the servers and the storage subsystem should be reliable enough to keep the data safe and make sure that the organization’s processes are not interrupted due to data loss caused by hardware or disk failures. Since data itself is the most valuable “jewel” for the business, it makes sense to choose the storage solution that is reliable, highly available and fast enough to serve the users and applications requests. It means that the servers and their main components should be powerful enough: CPU should be fast enough to handle all requests and operations, the server should contain enough RAM to make sure that OS, applications, and services will operate properly and the disks should be configured with some kind of redundancy which will perform fast.
Another important factor is that data protection should ensure that data security is guaranteed, but, at the same time, data should be available as required for applications and users. Naturally, the IT solution should be cost-effective in terms of maintenance and support, and able to run virtualized environment, to save money on purchasing of server hardware.
One of possible storage solutions that fits all the above requirements is SAN (Storage Area Network). As mentioned earlier, the solution should be redundant on each layer. Any single point of failure with SAN, network switches or server NICs can cause downtime and even data loss, so redundancy is a key element to implementing a SAN.
What is SAN?
SAN is an acronym for Storage Area Network – it is a shared storage box that can be provisioned to servers over a dedicated network. The term SAN is commonly used to refer to dedicated storage servers that use either the Fibre Channel (FC) or the iSCSI protocol. By carrying SCSI commands over IP networks, iSCSI is used to facilitate data transfers over intranets and to manage storage over long distances. iSCSI simplifies the implementation and management of the SAN. Conversely, FC storage requires expensive switches and host cards (HBAs), and by design forces to put the SAN on a completely separate Fibre Channel network. Enterprise SANs typically come with options for single or dual controllers for redundancy and can allocate disks based on various RAID levels as required by the applications and services.
Why iSCSI Storage?
According to Wikipedia, “iSCSI is an acronym for Internet Small Computer Systems Interface, an Internet Protocol-based storage networking standard for linking data storage facilities. It provides block-level access to storage devices by carrying SCSI commands over a TCP/IP network.” Using this protocol is a great opportunity to share and connect storage via LAN from one server to another without breaking a bank.
Let’s try to think about iSCSI storage (based on the iSCSI protocol) benefits in comparison to NAS (Network Attached Storage), DAS (Direct Attached Storage also known as local hard disks), or a FC (Fibre Channel) SAN.
iSCSI storage presents block-based storage just as the internal disk drives, whereas a NAS is just a plain file server, which presents storage as file shares. It offers more flexibility in disks management. Also, iSCSI allows configuring authentication for the servers where storage is going to be connected. This is possible via using Challenge Handshake Authentication Protocol (CHAP) which provides a high level of security.
iSCSI storage is Ethernet-based, so IT Administrators require just basic IP networking knowledge, while FC is more complicated and requires special skills. There is a common misconception that FC offers better performance than iSCSI, but, with modern network cards, it was proved that, in most cases, iSCSI storage performance is equal to the physical disks’ performance. Hence, in most cases, the system’s performance is limited by disks performance storage itself rather than network throughput. Additionally, FC storage is significantly more expensive than iSCSI, assuming that it requires FC HBAs (Host Bus Adapters) on client machines as well as expensive FC network switches.
Also, iSCSI storage is simple to manage because of standard iSCSI initiators, present in most popular modern operating systems and hypervisors like Windows, ESXi, Xen, etc.
What do SANs provide that one cannot get from Direct Attached Storage (DAS)?
Why using a SAN is more effective? It looks like the directly attached disks are working just fine but taking into account that a SAN provides the benefits described below.
A SAN enhances further possibilities and features available for the clustered environments, as well as moving virtual machines between physical servers without any downtime. Such an option for the virtual machines allows for easy management of hardware resources and enables power down of a physical server for repair or maintenance without impacting services and applications. This is a great benefit for the environments which should be continuously available and where downtime is out of question. IT staff is able to perform the equipment maintenance without jeopardising the availability of resources.
Unfortunately, implementing hardware SAN with all additional redundant equipment will significantly increase the cost of the IT solution, making it unaffordable for some businesses.
One of the alternatives, which has acceptable TCO (Total Cost of Ownership), fits modern requirements, such as the ability to run virtualized highly-available environment with storage replication, guarantee data consistency, great performance, no downtime and 24×7 ProActive support, could be StarWind HyperConverged Appliance. StarWind HyperConverged Appliance comprises at least two servers that contain hypervisor to run a cluster with virtual machines on a highly-available storage, replicated between these servers by StarWind VSAN. StarWind HyperConverged Appliance can be managed from a single console and provides options to perform storage tasks easier and with more reliability and flexibility, by using the iSCSI protocol.
Virtual SAN implementation recommendations
With SAN in mind, the key words are “Storage” and “Network”. Working closely with IT administrators, I observed that, occasionally, people using SAN failed to plan and take into account some basic and obvious things. Therefore, I would like to outline some general mistakes in planning and ways to avoid them.
Storage recommendations
As mentioned earlier, the storage subsystem should be reliable enough to keep the data safe and fast enough to serve the users and the applications. Before making decisions, please estimate the desirable performance of the storage subsystem. To determine the performance numbers, evaluate the workload generated by applications, services, and the OS, taking into account the future 20-40% growth of the production. So, the storage configuration should be able to provide the performance numbers, estimated above and it also should be redundant. The easiest way to get the redundant storage is to build RAID (Redundant Array of Independent Disks). The most popular arrays are RAID 1 for the operating systems and RAID 5 and RAID 10 for the data. Here is the first pitfall. RAID 10 is fast and redundant but gives 2-times less space in comparison to the total disks space, which also could be a very expensive solution. RAID 5 (6) looks promising in terms of the space and a lot of people are choosing it, forgetting that the write performance of RAID 5 (6) is limited by parity disk performance which is about 160 IOPS for HDD and, in most cases, is not enough.
Luckily, there is a gentle solution here – deploying RAID 5, based on SSDs. Since we are interested in the performance, it is better to choose mix-used SSDs. The efficiency of the space usage and the performance of such configuration is significantly higher while the price could be lower than spindled disks in RAID 10. Assuming that storage is fundamental to an IT system, it makes sense to invest into a good solution.
There are many other options for how to build a storage, thus if the storage performance estimation is done correctly, there will be no issues in the future.
Network recommendations
Next very important component is the network configuration. In case of iSCSI SAN, a network performance is vital given that it completely determines the SAN performance. The network should be physically redundant and should have a throughput that should not be a bottleneck for the storage and for the whole setup.
With the iSCSI protocol, it is entirely possible to connect iSCSI storage directly into the existing corporate network, also known as LAN. However, the corporate LAN should not be designated for the SAN even though it is possible to use it for that purpose. This design could flood the corporate LAN with storage traffic which may slow everything down. So with iSCSI storage, it is highly recommended to use a separate network to connect to the storage device. Of course, since it will be a dedicated network for storage traffic the data should be secured and not adversely affect the existing network traffic in the corporate LAN. So, in the case of iSCSI storage, it is better to install dedicated interfaces into the server for the SAN traffic.
Back to the network redundancy and performance, some people believe that using NIC teaming will resolve both of the above mentioned questions. Unfortunately, with iSCSI traffic in mind, it is not the case as NIC teaming causes significant performance issues for the SAN. To avoid such issues, it is better using MPIO for ISCSI targets connection – a simple and functional solution.
Protecting Data with SAN
Once a centralized/shared storage is in place, one of the first questions the administrator should ask himself is how to back up the data that resides on the storage server.
For more detail, please read carefully the Best Practices document about backing up VSAN from StarWind Environment (and other SAN environments) :
Getting Your Own SAN and Cluster
The simplest way is to visit StarWind HyperConverged Appliance product page and request session:https://www.starwindsoftware.com/starwind-hyperconverged-appliance
Before working with StarWind VSAN, please read carefully StarWind Best Practices here: https://www.starwindsoftware.com/resource-library/starwind-virtual-san-best-practices
Also, please take into account StarWind VSAN system requirements published here: https://www.starwindsoftware.com/system-requirements
The technical guides about cluster configuration are available in StarWind Resource Library: https://www.starwindsoftware.com/resource-library/resource-type/technical_papers
Summary
Although the recommendations and best practices above seem to require additional expenses while building iSCSI SAN, but, taking into account the hardware solution cost, it clearly makes sense to follow these recommendations and get a robust IT configuration rather than deal with an on-going stream of user complaints caused by the IT system malfunction. IT administrators gain obvious advantages when using an iSCSI Virtual SAN and, at the same time, stay within the constraints of a limited budget.
The documents should help with selecting the suitable hardware for the future cluster, although it is still important to note that the CPU speed and amount of RAM, as well as hard disk speed, the disks quantity in the RAID and RAID type itself, affect the system’s performance.
So, please note the importance of the above mentioned points for the construction of a new IT infrastructure.
