Virtualization has transformed how server processing power is utilized, but the bottleneck is still storage. Server virtualization poses problems for storage, and storage has to adapt in new ways when desktop virtualization is implemented.
Commonwealth Financial Network, a financial services company based in Waltham, Mass., and San Diego, Calif., supports independent financial advisors. The company provides custom apps, offers general business support on the back end, and hosts clients' email and websites. To be licensed, Commonwealth Financial Network must operate in compliance with Financial Industry Regulatory Authority (FINRA), an independent regulator for all securities firms doing business in the U.S.
Commonwealth has virtualized 90% to 95% of its servers, and considers VMware-based virtualization essential to its business. They have not virtualized some large database servers … yet. Dave Hughes, senior storage engineer with Commonwealth, says the company's data center would quadruple in size and wouldn't be able to provide instant-access, 24/7 app availability if it weren't for virtualization.
When speaking with SearchVMware.com at a New England VMUG meeting centered on storage's role in virtualization, Hughes described his company's decision to move from VMware vSphere 4.1 to vSphere 5.1 and what the upgrade entails. He also explains the company's desktop virtualization trial and how it has structured a backup site.
You said virtualization is key to Commonwealth's ability to operate, but also noted that virtualization has not transformed storage with as much success as it has had in servers. As a storage admin, what do you expect from VMware?
Dave Hughes: VMware has always struggled with storage in the virtualized environment, but has made several enhancements to vSphere storage functionalities from version 4, onward. The issue for virtualization is that multiple servers are clamoring to talk to the same storage. With vSphere, storage is a bottleneck.
VMware has always struggled with storage in the virtualized environment, but has made several enhancements to vSphere storage functionalities from version 4, onward.
Commonwealth uses VMware vSphere 4.1 and is currently spec'ing out a transition to vSphere 5.1. VMware vSphere 5.1 is attractive because it allows servers to share storage resources with more automation, although there is still a lot of manual management work involved.
We'll need to update hardware to use vSphere 5.1, as VMware has become stricter with its supported hardware. It's a good opportunity to improve our infrastructure.
What's next for a company that's almost 100% virtualized? How can they improve operations via virtualization?
Hughes: Commonwealth is in beta with [virtual desktop infrastructure] VDI, experimenting with VMware View 5.1. Our goal is to not buy any more PCs, and VDI is promising.
We're considering the storage implications of VDI, especially boot storms and antivirus storms. Users shouldn't feel a difference when they're on a virtual desktop, so these intensive actions undertaken en masse, like everyone starting up their OS at the same time in the morning, or all of the computers' antivirus scans running at 8 p.m. on Wednesday, for example, require tweaks to the storage architecture. I'm considering adding solid state storage [SSS] to handle boot storms. I already use SSS to manage hot spots, or peak usage, from the servers. Solid state drives are less expensive and more reliable than they were when the technology first came out, but there are still limitations. I also plan to separate VDI and production servers so each can be optimized for the task.
How do you establish an appropriate recovery point objective (RPO)? And what do you do to ensure 24/7 availability of your apps for reps in the field?
Hughes: Commonwealth has a dark-fiber-connected backup location in Marlborough, Mass., for its Waltham and San Diego offices. San Diego originally started out as our disaster recovery (DR) site, and we added Marlborough as the business grew. Marlborough's infrastructure runs synchronously with the production servers for up-to-the-bit images. Commonwealth requires 0 RPO due to the nature of its industry, financial services.
The DR site is connected to Waltham by two dark fiber cable routes that are physically different, to guard against an unpredictable disruption, such as a car accident taking down a power-line pole. We do everything we can with redundancy.
High availability is more of a focal point for VMware. While it adds expense, undisrupted operation is our company reps' service expectation. If a server crashes, they may experience a pause in the app, or at worst have to log back into it. No data is lost.