Virtualization Technology News and Information
Article
RSS
QLogic 2016 Predictions: Open Source, Speeds and Feeds Will Rule in 2016

A Contributed Article by Greg Scherer, Vice President & Chief Technology Officer of QLogic

With the arrival of 2016, I'd like to share my thoughts on four key technology and market trends that will be driving the transaction-heavy data center into the New Year: open source technology, faster speeds and feeds, flash and solid state drive (SSD) solutions, and ease of management tools.

1.  Open Source Technology

What began with Linux in the early 1990s has exploded with OpenStack, Open Compute Project (OCP), Data Plane Development Kit (DPDK) and applications like Hadoop, memcached, etc. The proliferation of open source technology will be driving IT development in 2016 more than ever. We're taking it seriously in our labs, and think you should too.

Here are three of the most notable open trends to keep your eye on in 2016:

  • OpenStack will gain even more momentum in the coming year, particularly in the private cloud market. And with the launch of Kilo, it is more robust than ever and will be a big draw for the enterprise space.
  • OCP form-factors will extend their reach in 2016 to data centers that want to take advantage of an optimized card form-factor in a variety of vendor-independent servers. At one time, far-eastern hyperscale hardware providers were the only vendors to offer OCP servers. But now, they are becoming more mainstream with Tier 1 server vendors beginning to embrace the OCP form-factor. OCP is all about broadening customer choice and allowing a new class of vendor-independent server, I/O, storage and switch hardware to emerge.
  • Network Functions Virtualization (NFV) is an emerging trend that allows customers to replace or augment proprietary switch-based appliances with x86 servers, DPDK-enabled Network Interface Cards (NICs), and software that creates "virtual appliances" and provides huge flexibility to carriers, managed service providers (MSPs) and to the general data center as a whole.

2.  Faster Speeds and Feeds

Today, the majority of data center traffic is "east-west," meaning that servers, storage and appliances all need to be interconnected in a manner that facilitates peer-to-peer conversations, as opposed to pushing traffic "north" to have conversations. This is not unlike the transformation that the telephone system went through from "operator-assisted" calls to "direct-dial" peer conversations. Between greater use of Storage Area Networks (SANs) that couple virtual servers to their shared storage to scale-out storage environments like Hadoop/MapReduce, east-west traffic patterns are driving the need for fatter interconnection (bandwidth) and faster interconnection (lower latency).  

  • 25Gb Ethernet (25GbE) will be a loud topic of conversation in 2016, because it lays the groundwork for 100Gb Ethernet (100GbE) and offers up the bandwidth and performance demanded by next generation and Web-scale environments. I predict that some of the early adopters of 40Gb Ethernet (40GbE) may even backtrack to 25GbE, because of 25GbE being so much more cost effective. When you can have a single lane that offers 2.5x the bandwidth of 10Gb Ethernet (10GbE) and 60 percent the bandwidth of 40GbE at a quarter of the infrastructure cost (40GbE = 4 lanes versus 25GbE's single lane) there's not much more to say. Serial, single lane technology will always be much more cost effective and much easier to deploy than multi-lane technology.
  • Gen 6 (32Gb) Fibre Channel is not just following on the heels of Gen 5 (16Gb) Fibre Channel, but will be the next big wave for data center consideration in as early as 2016. The shift to Gen 6 (32Gb) Fibre Channel is being driven by the demands of next generation workloads, low latency SSD/flash storage and caching, and memory-intensive and transaction-heavy applications including virtualization, online transaction processing (OLTP), data migration and 4K video production.

3.      Flash and SSD storage solutions are transforming the data center, and will be a force to contend with in the immediate future as virtualization proliferates and performance-challenged enterprise applications create bottlenecks within the infrastructure. Welcome to the era of flash/SSD.

My conclusion is that flash is THE emerging trend, pushing aside the hard disk drive (HDD) incumbent; this is supported by a plethora of computing benchmarks and statistics. Because we are space-constrained, I'll give you my top three proof points:

a)  A single 2TB Non-Volatile Memory Express (NVMe)-based SSD can handle 1 million IOPS, more than 150 times the IOPS capacity of an HDD storage array in the same form factor.[1]

b)  SSD/flash offers storage performance that is 10 times that of HDD[2] 

c)  Flash is 1,000 times cheaper than HDD in dollars per IOPS.[3]

The growing prevalence of flash storage is one of the key drivers for the critical requirement for higher speeds and feeds (my number two trend). Memory-intensive, I/O-starved, transaction-heavy applications beg microsecond response times and simply put, are fired up by flash. But flash needs network performance to get the job done right. This leads us into the symbiotic discussion of the latest technology trend towards NVMe. NVMe provides a standards-based approach for PCI Express (PCIe) SSD access that significantly improves performance by reducing latency and streamlining the command set. Flash is fast. NVMe can make it even faster. 

There are two approaches to NVMe that are looming brightly on the 2016 horizon:

  • NVMe over Fibre Channel (FC-NVMe) extends the simplicity and efficiency of NVMe by transferring NVMe commands and structures end-to-end, requiring no translations. Fibre Channel's purpose-built, dedicated storage transport along with its parallelism and battle-hardened reliability make it an ideal transport for NVMe.
  • NVMe over Remote Direct Memory Access (RDMA) allows Ethernet RDMA-capable environments to transport NVMe SSD payloads rather than use the traditional rotating media transport that adds additional translation and stack overhead.

4.  Ease of Management

Mission-critical, private cloud and flash-accelerated workloads are transforming the data center, and because of this you're going to see more demands than ever before for performance, Quality of Service (QoS) and infrastructure reliability. Therefore, centralized management is one of the most critical trends for a technology-mature data center to acknowledge in 2016. "Cool technology" as laid out in this blog is only cool if it is easily managed. It must be "transparently" managed, or at least be managed in the context of commonly used tools that are native to the operating system (OS) or application.

##

About the Author 

Greg Scherer joined QLogic in March of 2014 from Broadcom, where he was the vice president of server and storage strategy for the company's ING group. Greg has more than 35 years of diverse technology experience that spans mainframes to microcomputers across a variety of technologies including storage, networking and virtualization.

Prior to joining Broadcom, Greg was the Chief Technology officer for Neterion, where he helped to drive the industry's first SR-IOV and MR-IOV capable Ethernet adapters. Before Neterion, Greg spent nearly 24 years at Emulex where he held ever increasing roles from individual contributor up thru Chief Technology Officer. Greg was a founder of the Fibre Channel business at Emulex and the inventor of Emulex's driver interface (SLI).

Greg remains very active in identifying solutions to key customer pain-points and working closely with product development teams to bring these solutions to market. He is an active participant in the Storage, Networking and Virtualization communities, holds several patents and is a frequent speaker at industry and partner events.

Greg Scherer 



[1] IT Brand Pulse Industry Brief: Fibre Channel Supercharged for Performance and Scalability

[2] https://www.sandisk.com/content/dam/sandisk-main/en_us/assets/resources/enterprise/white-papers/sandisk-establishes-the-big-data-flash-category-with-its-infiniflash-introduction.pdf

[3] https://powermore.dell.com/technology/flashdance-3-reasons-flash-storage-might-stay/

Published Wednesday, January 13, 2016 6:31 AM by David Marshall
Filed under: ,
Comments
There are no comments for this post.
To post a comment, you must be a registered user. Registration is free and easy! Sign up now!
top25
Calendar
<January 2016>
SuMoTuWeThFrSa
272829303112
3456789
10111213141516
17181920212223
24252627282930
31123456