Oracle Solaris and SPARC (and x86) Performance, Part 4
By Larry Wake -Oracle on Apr 08, 2013
East or West?
I admit it. Being a bit removed from my days as a system administrator, and not necessarily steeped in the day-to-day existence of network design, I was slightly perplexed when I first saw the original of the illustration on the right.
This was shortly after we'd acquired Xsigo, a company whose name also did not immediately reveal anything about the cool stuff they made, and what problems it solved.
So here's this slide I need to present, with this illustration, and no speaker notes. Well, most of it makes sense, but "W" and "E"? Ah, possibly "Web" and "Enterprise"! It sure couldn't be anything like, I don't know... "West" and "East!" What would that have to do with networking?
And many of you are already throwing things at your screen, yelling, of course it's "West" and "East," dumkopf! (Especially if you're German.) It turns out that if you're smarter than, oh, say, me, you know that the "East/West problem" is a thing, and Oracle Virtual Networking, the products the Xsigo acquisition adds to our portfolio, are the things that solves that thing.
Home is Best
(If by "home" we mean "simplified data center deployments with a wire-once solution and simple software defined network configurations." And it's my blog post, so that is indeed what it means.)
Simply speaking, east/west traffic is the traffic that rather than going in and out of the data center (which in the world of this metaphor is called north/south traffic), goes between servers in the same data center -- or even within the same physical server.
Clearly, the more devices you have interposed in this traffic, or the more congested your connections are, the slower, more complicated, and expensive things are. Equally clearly, east/west traffic is becoming a really big deal in the age of virtualization and networked storage.
Oracle Virtual Networking addresses this in a big way, with an open architecture data center fabric delivering 80 Gb/sec bandwidth. Not only does this make for faster services (customers have seen fourfold application performance improvement), it can also cut capital expenditures by significant amounts and simplify administration, allowing new services to be set up in minutes instead of days. We have one customer who was able to take the 98 network cables they would run into a single blade chassis down to 6, and drop 6 switches in such a configuration down to 2.
And, as the title of this post indicates, this is now available for Oracle Solaris deployments. It's supported with Oracle Solaris on both x86 and SPARC systems, including the new SPARC T5 and M5 servers. It also supports Oracle Linux, Oracle VM, VMware, Microsoft Windows and Microsoft Hyper-V, so we've pretty much got you covered.
Where things can get very interesting with Oracle Solaris, of course, is how this can be used in conjunction with a feature like Oracle Solaris Zones, and how that in turn leverages the power of ZFS. Basically, you're putting everything in your data center on frictionless bearings, so that instead of recabling your systems, or having to physically provision new storage, you can instead stand up new application environments at a moment's notice by reallocating existing resources.
Now, instead of throwing things at the screen, I imagine you're now cleaning up the coffee spray and saying, "My gosh, you guys should do a Webcast about this!" Excellent idea.
Webcast: How to Manage Your Data Center with Oracle Virtual Networking
Tuesday April 30, 2013