Everyone wants to go fast. Lewis Hamilton does a pretty good job of it. As does Usain Bolt. One might argue that their need for speed goes beyond their competitive desire to win a race. That they are driven (so to speak) by the desire to be the greatest of all time in their respective sports.
When you say that you want to increase the performance and availability of your Oracle databases, what are you really trying to achieve? Speed time to market for new services and products? Respond rapidly to changing consumer demand? Make sure that operations run smoothly during peak periods of demand? All of the above?
NVMe is a technology created to accelerate the transfer of data between enterprise and client systems and SSDs over a computer’s high-speed bus. NetApp is the first vendor with true end-to-end support of NVMe. There are several exciting aspects to NVMe in Oracle environments:
- Low latency. Faster response times.
- Greater availability. Reduced operational disruption.
- Simplicity. As the next step in the evolution of SAN, NVMe is easier to manage.
- Reduced overhead. NVMe is less taxing on server CPUs than traditional SAN work, freeing your CPUs for real application work.
- Oracle licensing. Why pay for expensive Oracle licenses for server CPUs that are doing storage work?
Literally every Fibre Channel (FC) customer should be planning their migration from SCSI/FC (which is what we usually call SAN) to NVMe/FC. Every one of them. It doesn’t matter if they need the performance, it just makes sense. It’s easier to configure, response times are better, and the storage system works more efficiently.
Here’s the really good news: Virtually nothing else is required. In most cases, the FC SAN you own now can already support NVMe/FC. The usual best practices apply, too. You still make file systems or Oracle ASM disk groups, and you manage them as usual. The only real impact is that you probably don’t need as many LUNs as you used to.
An Oracle database, and really most applications, can’t tell the difference between a LUN and an NVMe namespace, except that NVMe performs better. The application just “sees” a device that looks like a big file. It’s a swath of bytes that the app can read from and write to. If Oracle wants to read block #384,283 from a database, it reads the 384,283rd 8K block off that OS file device.
What NVMe does is replace the ridiculously complicated code that translates a request to read that block into ancient SCSI commands, sends them to the storage system, waits for the response, and translates the SCSI data back into the data requested.
In essence, NVMe does less than SCSI. It’s a subtraction, not an addition or change. It’s streamlined SAN, designed for solid state storage. The applications do the same things they always did, but the OS has to do less to complete an I/O. That’s why you don’t need to do much of anything differently. Just upgrade and start getting more from your SAN.
Of course, there are choices when it comes to how you implement NVMe; for example, some vendors are wedded to RoCE. Here are a few things you should know:
NVMe over RoCE is based on older proprietary standards that require specialized hardware. Most agree that RoCE will be replaced with NVMe over TCP, but the standard was ratified only about a year ago. Products are just now starting to appear, so the technology is not quite mature.
SCSI may be obsolete, but FC is not. FC is a specialized, high-performance, storage network protocol that:
- Has built-in QoS for more reliable storage performance.
- Has built-in discovery capabilities. Anyone who has worked with iSCSI knows what a pain it can be to configure storage without a native discovery capability.
- Has zones, so it’s easier to secure.
- Is everywhere. You paid for it, so why not use it?
Just stop doing that old SCSI thing on FC and start using NVMe on FC.
To see how NetApp is accelerating Oracle, and other, workloads, with NVMe watch the video below.
To learn more about what NetApp is doing to secure and accelerate Oracle workloads, go to: