A scalable storage is a data storage system which allows the users to grow their storage resources in line with their business requirements. It addresses the challenges created by legacy systems which present storage limitations.
Scalable storage systems are two types 1.) Scale-up storage system 2.) Scale-out storage system. In scalable storage architecture the functionality and performance are maintained on par with the storage resources expansion. It’s a great advantage for the IT managers who are challenged with the infrastructural growth issues, as they can manage the growth of their data within limitation of the IT budget without having to do a costly “forklift” upgrade.
Difference between Scale-up and Scale-out storage
Many IT managers are in confusion that Scale-up Storage and Scale-out storage are synonymous. But in reality they are two different concepts related to storage expansion
- Scale-Up Storage- Scale-up means taking an existing storage system and adding capacity to it in order to meet the capacity demands. That is adding up storage capacity in the same node and thus making the node bigger. The point to be noted over here is that a scale-up storage also requires additional space, power and cooling requirements, but without the need of an additional controller.
- Scale-out Storage- In this architecture, additional storage called nodes are added horizontally to increase capacity and performance. The only difference in between a scale-out system and adding of additional hardware onto the floor space is that Scale-out is still considered as a single system. In a scale-out system additional power, cooling and space requirements have to be achieved along with the addition of controller functions.
Some systems come up with a combination of scale-up and scale-out.
Generally, start-ups commence with simple storage, but it turns complex when there is data growth and need to upgrade arises. Securing additional capacity will be cited as the most common reason to upgrade a storage system in order to accommodate more users, files, applications and attached servers.
In addition to capacity, the other storage resources which have to be available with the growing demands of an organization are computing power and bandwidth. It is evident that in the absence of enough Input/Output bandwidth, the users will face difficulty in connecting to the servers due to network bottlenecks and thus the overall performance of the entire storage system will be in jeopardy. Coming to computing resources, the need for additional services like snapshots, replication and volume management generates the need for highly qualified storage software.
The disadvantage with legacy scale-up systems is that capacity is added as per requirement, but bandwidth and computing power remains as it is. So, as a result the performance will be high on the very first day of use. But gradually the performance starts to degrade, which prompts the need to overbuy storage compute power and bandwidth, thus it increases the costs.
In scale-up storage, after it reaches the performance peak the compute and bandwidth capabilities of the storage engine have to be replaced or additional stand-alone storage system has to be purchased. The first option of forklift upgrade will surely be expensive; while the second option becomes a management nightmare as additional IT staff has to be hired.
In scale-out storage architecture, there are individual storage components called ‘nodes’. Each node is enclosed with capacity, processing power and storage I/O bandwidth. As the addition of nodes takes place the three resources in the system will also be upgraded simultaneously. Additionally, the nodes are interconnected by a high speed network and so they can communicate with each other. So no matter how much capacity is added in a scale-up architecture, the performance of the system will be increasingly faster. In a scale-up storage system, granular expansion of various storage resources can be achieved. So purchase of only the resources which are needed is possible.
Scale-out storage software plays an important note in making the interconnected nodes accessible as a single storage system to the servers and storage administrators. So, the nodes are made into a single cluster or grid and the storage cluster software administers the writing of data across all the nodes in the scale-out storage infrastructure. So, the data writes and data reads are spread across more processors and I/O connections in the clusters. It is also essential that any node in the scale-up architecture must act as a control node at any time. If in case, the data path has to be routed through a single control node then the cluster storage system will revert to something which has the same limitations.
At the same time, storage cluster software should also be capable to take benefit of the extra RAM that each node carries to the cluster and must be able to leverage it to promote enhanced performance. Another key function of the storage software is the ability to support wide range of workloads types with the use of additional storage compute and I/O resources. Finally, the storage software must support all the data services that IT executives are expecting from a storage system like snapshots, thin provisioning, cloning, replication and automated storage tiering.
Modern data centers use scale-out storage
Modern data center architectures are ideal environments for scale-out storage as there is a mixture of workloads and I/O demands. Since, user files are large in size and numerous in numbers, frequent sequential processing becomes essential. For this reason, data centers are moving towards server virtualization in order to cut down on hardware costs and IT staff.
In order to deal with the function of sequential processing of larger and numerous files, Scale-out storage is uniquely positioned to address this need. As bandwidth and computing resources grow simultaneously with the capacity, the need to fine-tune the system for exceptional cases gets reduced in this architecture. Hence, with the default system configuration, sufficient amount of performance can be achieved and that too without the need for expertise storage managers.
Although scale-up storage systems vendors prop up that their concept will support management interfaces, capabilities and cross-system file systems; these three factors are not satisfied in reality. For instance, if an IT manager needs to add a server to the scale-up storage setting, then he needs to first evaluate and determine which resource is the best candidate to support that server’s workload.
In a scale-out storage system, everything seems like a single system and one volume. So, the IT manager simply needs to connect the server to the storage and start using it. In case the new server work functions spike up the need for more storage, then a new node can be added and it brings in more bandwidth, computing resources into the environment.
So, scale-out storage can be an ideal solution for modern data center where workload and high performance needs cannot be compromised while trying to bring down the management costs.