Take your datacenter into the future with SMB 3.0

THE MARKET HAS SPOKEN: Ethernet is the de-facto IT fabric standard around the world. If market predictions, customer tendencies, or current purchasing patterns hold any weight, Ethernet’s success is set to continue into the future as well. Ethernet’s dominance should come as no surprise; with the ability to support many storage protocols (i.e. iSCSI, FCoE, NFS, SMB/CIFS) over the same wire, it is an easy choice to deploy, and helps ‘future proof’ against changing infrastructure requirements, workloads, and data transports.

The block vs. file debate
Okay, Ethernet is great, but what about block versus file, and which protocol to use when? For starters, take a look at your workloads and data.

Data is like entropy - it always increases. If opening the newspaper and reading countless headlines about datacenter expansion isn’t enough evidence, market research firm IDC predicts 4X data growth by 2017 in the File- and Object-Based Storage market, representing a five year 47.2% CAGR and 173 Exabytes of storage sold in 2017.1 Plus, File- and Object-based storage is outpacing overall storage growth. This poses a significant challenge in today’s environment: administrators are tasked with managing an exploding amount of data while their own headcount dwindles or remains the same.

How are they going to do it?
I often ask customers to describe their ideal next generation datacenter 5 years from now. Sometimes I get the “clean lines, well-lit” answer. Interior design jokes aside, I’m looking for how they want their data, applications, hypervisors, and management planes to work and behave together. From a storage administrator’s standpoint, the answer is a pervasive and consistent trend towards the goals of total automation, self-service, and self-provisioning. This trend is not only a matter of convenience, but also one of survival. It sounds difficult, but these goals are made easier with file-based storage.

Take a moment to reflect on your regular admin tasks. Will you still be mounting LUNs and dealing with HBA’s, WWN’s, or iSCSI targets in 5 years? Many of these tasks cannot be automated, and regular data movement means these operations are a roadblock to the dreamland we so desire. Most folks I speak with no longer have time for the tasks associated with managing traditional SAN arrays, and everyone is looking to streamline storage administration. How do we simplify while managing larger datasets?

Moving SAN workloads to NAS
The answer is simple: take traditional SAN workloads and put them on NAS. Managing files versus LUNs allows for easier data mobility and granular management. You can move data around, restore single VMs and databases (or groups of them), and little technical knowledge is required. Configuration, troubleshooting, and upgrades are all simplified. The market validates this trend, too. If Ethernet usage is trending up as a whole, NAS file protocol usage (SMB/CIFS, NFS) is skyrocketing.

In addition to the data-entropy theory and continued growth in traditional file sharing, the success of file-based storage protocols is also because major applications, databases, and hypervisors moved to support them. Software companies saw the benefits of file storage and wanted their products to be at the forefront. Oracle and VMware have supported NFS for several years, and with the advent of Microsoft SQL Server 2012 and Windows Server 2012 Hyper-V, customers have great options for their Windows environments thanks to the new SMB 3.0 protocol. There are also workloads where structured applications manage unstructured data, and Microsoft SharePoint is the poster child; SharePoint can also run over SMB. Given the direction of file-based storage as well as Microsoft’s investment in their applications and protocol stack, there should be no doubt that SMB 3 is the future of storage protocols for Windows Environments.

A new look at SMB
Some reading this may be thinking, “SMB? As in...(gasp)...CIFS??” Yes. We have come a long way since the days of slow and chatty CIFS/SMB 1, and SMB 3 has both the technology and capability to support today’s demanding and critical workloads. SMB 3 delivers SAN-level availability and resilience, allowing workloads like mission critical VMs and databases to run over it with high availability. It also delivers SAN-level performance with features like SMB Multichannel and SMB Direct (SMB over RDMA), and scales very well when compared to Fibre Channel. SMB Multichannel enables bandwidth aggregation with multiple NICs and automatic session failover, whereas SMB Direct leverages the RDMA transport to reduce latency and CPU overhead. Other key features include Remote VSS, which allows integrated data protection for applications stored on a SMB share; Optimized Data Transfer (ODX), which enables server-side copies; SMB Encryption, which enhances security with encryption over the wire; and Management improvements with PowerShell and SMI-S.

Since SMB 3 was designed for high availability and non-disruptiveness, it makes the most sense to use it with two or more nodes on the server side for failover handling. This way, the continuously available (CA) shares providing data will always be online and available to SMB 3 clients. CA shares are enabled by updating another server node with a client’s state changes. In the event of a failure, the surviving server node has the client’s state information and can recover the connection with no error returned to the application or user. The Cluster Client Failover feature achieves a similar effect on the client side, allowing for failover between two client nodes running the same application.

SMB 3 will have different impacts to your environment depending on how you use it, but Hyper-V is generating the most excitement at the moment. Quick provisioning - either via Hyper-V Manager, System Center Virtual Machine Manager, or PowerShell - is a huge boon to storage administrators who can spend days pushing out golden VM images on LUNs or via client-side copy. Thanks to tie-ins with ODX, with SMB 3, you can simply copy and paste the golden VM (or database, etc.) as many times as you want, with no CPU utilization or data transfer through your laptop.

Additionally, VMs (or databases, etc.) show up as normal files that you can browse in Windows explorer, enabling greater transparency. Beyond the impressive features, ease of management, performance, and resiliency, SMB 3 will also enable a more flexible IT environment, reducing complexity and offering additional design options. Service providers or corporations will have the ability to create different offerings based on storage backend if they prefer. Larger companies will be able to reduce operational costs with heterogeneous datacenters (i.e. one DC is block-based, another is file-based), and data transfer between them via ODX will be seamless.

Although SMB 3 has the technology and capability to support its upward trajectory as a data storage protocol, it is not just theoretical, it is real and ready. SMB 3 offers a new way of thinking for your datacenter and application storage designs. Just as Ethernet has dominated the physical transport layer, SMB 3 is set for success as a first-class file-based storage protocol, and offers many unique and compelling features for Windows environments.

Recently SNIA’s Ethernet Storage Forum delivered a webcast on this same subject. If you are interested in downloading the webcast entitled SMB 3.0 - New Opportunities for Windows Environments, please visit: https://www.brighttalk.com/webcasts?q=SNIA