Kingston’s NVMe Line-up Is The Life Of The Party

Disclaimer: I recently attended VMworld 2017 – US.  My flights were paid for by ActualTech Media, VMware provided me with a free pass to the conference and various bits of swag, and Tech Field Day picked up my hotel costs. There is no requirement for me to blog about any of the content presented and I am not compensated in any way for my time at the event.  Some materials presented were discussed under NDA and don’t form part of my blog posts, but could influence future discussions.

You can view the video of Kingston‘s presentation at Tech Field Day Extra VMworld US 2017 here, and download a PDF copy of my rough notes from here.

 

It’s A Protocol, Not Media

NVMe has been around for a few years now, and some people get it confused for a new kind of media that they plug into their servers. But it’s not really, it’s just a standard specification for accessing Flash media via the PCI Express bus. There’re a bunch of reasons why you might choose to use NVMe instead of SAS, including lower latency and less CPU overhead. My favourite thing about it though is the plethora of form factors available to use. Kingston touched on these in their presentation at Tech Field Day Extra recently. You can get them in half-height, half-length (HHHL) add-in cards (AIC), U.2 (2.5″) and M.2 sizes. To give you an idea of the use cases for each of these, Kingston suggested the following applications:

  • HHHL (AIC) card
    • Server / DC applications
    • High-end workstations
  • U.2 (2.5″)
    • Direct-attached, server backplane, just a bunch of flash (JBOF)
    • White box and OEM-branded
  • M.2
    • Client applications
    • Notebooks, desktops, workstations
    • Specialised systems

 

It’s Pretty Fast

NVMe has proven to be pretty fast, and a number of companies are starting to develop products that leverage the protocol in an extremely efficient manner. Coupled with the rise of NVMe/F solutions and you’ve got some pretty cool stuff coming to market. The price is also becoming a lot more reasonable, with Kingston telling us that their DCP1000 NVMe HHHL comes in at around “$0.85 – $0.90 per GB at the moment”. It’s obviously not as cheap as things that spin at 7200RPM but the speed is mighty fine. Kingston also noted that the 2.5″ form factor would be hanging around for some time yet, as customers appreciated the serviceability of the form factor.

 

[Kingston DCU1000 – Image courtesy of Kingston]

 

This Stuff’s Everywhere

Flash media has been slowly but surely taking over the world for a little while now. The cost per GB is reducing (slowly, but surely), and the range of form factors means there’s something for everyone’s needs. Protocol advancements such as NVMe make things even easier, particularly at the high end of town. It’s also been interesting to see these “high end” solutions trickle down to affordable form factors such as PCIe add-in cards. With the relative ubiquity of operating system driver support, NVMe has become super accessible. The interesting thing to watch now is how we effectively leverage these advancements in protocol technologies. Will we use them to make interesting advances in platforms and data access? Or will we keep using the same software architectures we fell in love with 15 years ago (albeit with dramatically improved performance specifications)?

 

Conclusion and Further Reading

I’ll admit it took me a little while to come up with something to write about after the Kingston presentation. Not because I don’t like them or didn’t find their content interesting. Rather, I felt like I was heading down the path of delivering another corporate backgrounder coupled with speeds and feeds and I know they have better qualified people to deliver that messaging to you (if that’s what you’re into). Kingston do a whole range of memory-related products across a variety of focus areas. That’s all well and good but you probably already knew that. Instead, I thought I could focus a little on the magic behind the magic. The Flash era of storage has been absolutely fascinating to witness, and I think it’s only going to get more interesting over the next few years. If you’re into this kind of thing but need a more comprehensive primer on NVMe, I recommend you check out J Metz’s article on the Cisco blog. It’s a cracking yarn and enlightening to boot. Data Centre Journal also provide a thorough overview here.

Dell EMC Announces Isilon All-Flash

You get a flash, you get a flash, you all get a flash

Last week at Dell EMC World it was announced that the Isilon All-Flash NAS (formerly “Project Nitro“) offering was available for pre-order (and GA in early 2017). You can check out the specs here, but basically each chassis is comprised of 4 nodes in 4RU. Dell EMC says this provides “[e]xtreme density, modular and incredibly scalable all-flash tier” with the ability to have up to 100 systems with 400 nodes, storing 92.4PB of capacity, 25M IOPS and up to 1.5TB/s of total aggregate bandwidth—all within a single file system and single volume. All OneFS features are supported, and a OneFS update will be required to add these to existing clusters.

isilon_all-flash_001

[image via Dell EMC]

 

Why?

Dell EMC are saying this solution provides 6x greater IOPS per RU over existing Isilon nodes. It also helps in areas where Isilon hasn’t been as competitive, providing:

  • High throughput for large datasets of large files for parallel processing;
  • IOPS intensive: You can now work on billions of small files and large datasets for parallel processing;
  • Predictable latency and performance for mixed workloads; and
  • Improved cost of ownership, with higher density flash providing some level of relief in terms of infrastructure and energy efficiency.

 

Use Cases?

Dell EMC covered the usual suspects – but with greater performance:

  • Media and entertainment;
  • Life sciences;
  • Geoscience;
  • IoT; and
  • High Performance Computing.

 

Thoughts and Further Reading

If you followed along with the announcements from Dell EMC last week you would have noticed that there have been some incremental improvements in the current storage portfolio, but no drastic changes. While it might make for an exciting article when Dell EMC decide to kill off a product, these changes make a lot more sense (FluidFS for XtremIOenhanced support for Compellent, and the addition of a PowerEdge offering for VxRail). The addition of an all-flash offering for Isilon has been in the works for some time, and gives the platform a little extra boost in areas where it may have previously struggled. I’ve been a fan of the Isilon platform since I first heard about it, and while I don’t have details of pricing, if you’re already an Isilon shop the all-flash offering should make for interesting news.

Vipin V.K did a great write-up on the announcement that you can read here. The press release from Dell EMC can be found here. There’s also a decent overview from ESG here. Along with the above links to El Reg, there’s a nice article on Nitro here.

Tintri Announces New Scale-Out Storage Platform

I’ve had a few briefings with Tintri now, and talked about Tintri’s T5040 here. Today they announced a few enhancements to their product line, including:

  • Nine new Tintri VMstore T5000 all flash models with capacity expansion capabilities;
  • VM Scale-out software;
  • Tintri Analytics for predictive capacity and performance planning; and
  • Two new Tintri Cloud offerings.

 

Scale-out Storage Platform

You might be familiar with the T5040, T5060 and T5080 models, with the Tintri VMstore T5000 all-flash series being introduced in August 2015. All three models have been updated with new capacity options ranging from 17 TB to 308 TB. These systems use the latest in 3D NAND technology and high density drives to offer organizations both higher capacity and lower $/GB.

Tintri03_NewModels

The new models have the following characteristics:

  • Federated pool of storage. You can now treat multiple Tintri VMstores—both all-flash and hybrid-flash nodes—as a pool of storage. This makes management, planning and resource allocation a lot simpler. You can have up to 32 VMstores in a pool.
  • Scalability and performance. The storage platform is designed to scale to more than one million VMs. Tintri tell me that the  “[s]eparation of control flow from data flow ensures low latency and scalability to a very large number of storage nodes”.
  • This allows you to scale from small to very large with new and existing, all flash and hybrid, partially or fully populated systems.
  • The VM Scale-out software works across any standard high performance Ethernet network, eliminating the need for proprietary interconnects. The VM Scale-out software automatically provides best placement recommendation for VMs.
  • Scale compute and storage independently. Loose coupling of storage and compute provides customers with maximum flexibility to scale these elements independently. I think this is Tintri’s way of saying they’re not (yet) heading down the hyperconverged path.

 

VM Scale-out Software

Tintri’s new VM Scale-out Software (*included with Tintri Global Center Advanced license) provides the following capabilities:

  • Predictive analytics derived from one million statistics collected every 10 minutes from 30 days of history, accounting for peak loads instead of average loads, providing (according to Tintri) for the most accurate predictions. Deep workload analysis identifies VMs that are growing rapidly and applies sophisticated algorithms to model the growth ahead and avoid resource constraints.
  • Least-cost optimization based on multi-dimensional modelling. Control algorithm constantly optimizes across the thousands of VMs in each pool of VMstores, taking into account space savings, resources required by each VM, and the cost in time and data to move VMs, and makes the least-cost recommendation for VM migration that optimizes the pool.
  • Retain VM policy settings and stats. When a VM is moved, not only are the snapshots moved with the VM, the stastistics,  protection and QoS policies migrate as well using efficient compressed and deduplicated replication protocol.
  • Supports all major hypervisors.

Tintri04_ScaleOut

You can check out a YouTube video on Tintri VM Scale-out (covering optimal VM distribution) here.

 

Tintri Analytics
Tintri has always offered real-time, VM-level analytics as part of its Tintri Operating System and Tintri Global Center management system. This has now been expanded to include a SaaS offering of predictive analytics that provides organizations with the ability to model both capacity and performance requirements. Powered by big data engines such as Apache Spark and Elastic Search, Tintri Analytics is capable of analyzing stats from 500,000 VMs over several years in one second.  By mining the rich VM-level metadata, Tintri Analytics provides customers with information about their environment to help them make better decisions about applications’ behaviours and storage needs.

Tintri Analytics is a SaaS tool that allows you to model storage needs up to 6 months into the future based on up to 3 years of historical data.

Tintri01_Analytics

Here is a shot of the dashboard. You can see a few things here, including:

  • Your live resource usage for your entire footprint up to 32 VMstores;
  • Average consumption per VM (bottom left); and
  • The types of applications that are your largest consumers of Capacity, Performance and Working Set (bottom center).

Tintri02_Analytics

Here you can see exactly how your usage of capacity, performance and working set have been trending over time. You can see also when you can expect to run out of these resources (and which is on the critical path). It also provides the ability to change the timeframe to alter the projections, or drill into specific application types to understand their impact on your footprint.

There are a number of videos covering Tintri Analytics that I think are worth checking out:

 

Tintri Cloud Suites

Tintri have also come up with a new packaging model called “Tintri Cloud”. Aimed at folks still keen on private cloud deployments, Tintri Cloud combines the Tintri Scale-out platform and the all-flash VMstores.

Customers can start with a single Tintri VMstore T5040 with 17 TB of effective capacity and scale out to the Tintri Foundation Cloud with 1.2 PB in as few as 8 rack units. Or they can grow all the way to the Tintri Ultimate Cloud, which delivers a 10 PB cloud-ready storage infrastructure for up to 160,000 VMs, delivering over 6.4 million IOPS in 64 RU for less than $1/GB effective. Both the Foundation Cloud and Ultimate Cloud include Tintri’s complete set of software offerings for storage management, VM-level analytics, VM Scale-out, replication, QoS, and lifecycle management.

 

Further Reading and Thoughts

There’s another video covering setting policies on groups of VMs in Tintri Global Center here. You might also like to check out the Tintri Product Launch webinar.

Tintri have made quite a big deal about their “VM-aware” storage in the past, and haven’t been afraid to call out the bigger players on their approach to VM-centric storage. While I think they’ve missed the mark with some of their comments, I’ve enjoyed the approach they’ve taken with their own products. I’ve also certainly been impressed with the demonstrations I’ve been given on the capability built into the arrays and available via Global Center. Deploying workload to the public cloud isn’t for everyone, and Tintri are doing a bang-up job of going for those who still want to run their VM storage decoupled from their compute and in their own data centre. I love the analytics capability, and the UI looks to be fairly straightforward and informative. Trending still seems to be a thing that companies are struggling with, so if a dashboard can help them with further insight then it can’t be a bad thing.

New eBook from Dell

I recently had the opportunity to contribute to an eBook from Dell (just quietly it feels more like a pamphlet) called “10 Ways to Flash Forward: Future-Ready Storage Insights from the Experts”. Besides the fact that I need to get a headshot that isn’t the same as my work ID card, I think it’s worth checking out if only for the insights that other people have provided. You can grab a PDF copy here. It’s also available via SlideShare.

Violin Memory Announces Additions To FSP Range

I got a chance to speak to Violin Memory at Storage Field Day 8 and was impressed by the company’s “new” approach to all-flash arrays. They recently announced the addition of the FSP 7600 and the FSP 7250 to the Flash Storage Platform. I’ve been told these will be GA in December 2015. Please note that I’ve not used either of these products in the wild, and recommend that you test them in your own environment prior to making any purchasing decisions.

Violin positions FSP as a competitive differentiator with Concerto OS 7 offering the following features:

  • Comprehensive Data Protection Services (including Syncronous, Asynchronous and CDP);
  • Stretch Cluster for Zero Down time and zero data loss;
  • granular deduplication and compression;
  • sustained low latency with Flash Fabric Architecture;
  • simple and single pane of glass management; and
  • integrated data migration and ecosytem integration.

The FSP 7250 is being positioned as an entry-level, sub-$100K US AFA that is:

  • Data Reduction Optimized (Always on Dedupe);
  • Integrated 3U Platform;
  • 8-26TB Raw; and
  • Up to 92TB Effective capacity.

The FSP 7600 sits just below the FSP 7700, and offers:

  • “Extreme” Performance
  • An integrated 3U Platform
  • 35-140TB Raw
  • 1.1 M IOPS < 500 μsecs

Unfortunately I don’t currently have links to useful things like data sheets, but you can read a nice summary article at El Reg here, and a link to the Violin Memory press release can be found here.

Pure Storage Announces FlashArray//m, Evergreen Storage and Pure1

That’s one of the wordier titles I’ve used for a blog post in recent times, but I think it captures the essence of Pure Storage‘s recent announcements. Firstly, I’m notoriously poor at covering product announcements, so if you want a really good insight into what is going on, check out Dave Henry’s post here. There were three key announcements made today:

  • FlashArray//m;
  • Evergreen Storage; and
  • Pure1 Cloud-Based Management and Support.

 

FlashArray//m

Besides having some slightly weird spelling, the FlashArray//m (mini because it fits in 3RU and modular because, well, you can swap modules in it) is Pure’s next-generation storage appliance. Here’s a picture.

Pure_hardware1

There are three models, the //m20, //m50, and //m70. Each of these has various capabilities. I’ve included an overview from the datasheet, but note that this is subject to change before GA of the tin.

Pure_hardware2

The key takeaway for me is that, after some time using other people’s designs, this is Pure’s crack at using their own hardware design, and it will be interesting to see how this plays out over the expected life of the gear.

 

Evergreen Storage

Pure_evergreen

In the olden days, when I was a storage customer, I would have been pretty excited about a program like Evergreen Storage. Far too often I found myself purchasing storage only to have the latest version released a month later, sometimes before the previous generation had hit the loading dock. I was rarely given a heads up from the vendor that something new was coming, and often had the feeling I was just using up their old stock. Pure don’t want you to have that feeling with them. Instead, for as long as the array is under maintenance, Pure will help customers upgrade the controllers, storage, and software in a non-disruptive fashion. The impression I got was that these arrays would keep on keeping on for around 7 – 10 years, with the modular design enabling easy upgrades of key technologies as well as capacity.

 

Pure1 Cloud-Based Management and Support

I’ve never been a Pure Storage customer, so I can’t comment as to how easy or difficult it currently is to get support. Nonetheless, I imagine the Pure1 announcement might be a bit exciting for the average punter slogging through storage ops. Basically, Pure1 gets you in touch with improved analytics and management of your storage infrastructure, all of which can be performed via a web browser. And, if you’re so inclined, you can turn on a call home feature and have Pure collect info from your arrays every 30 seconds. This provides both the customer and Pure with a wealth of information to make decisions about performance, resilience and upgrades. You can get the datasheet here.

 

Final Thoughts

I like Pure Storage. I was lucky enough to visit them during Storage Field Day 6 and was impressed by their clarity of vision and different approach to flash storage architecture. I like the look of the new hardware, although the proof will be in field performance. The Evergreen Storage announcement is fantastic from the customer’s perspective, although I’ll be interested to see just how long they can keep something like that going.

 

Brisbane VMUG – May 2015

hero_vmug_express_2011

Buckle up, the May Brisbane VMUG will be held on Thursday 21st May at the Dimension Data office in the city (Level 22, 141 Queen St) from 4 – 6 pm. It’s sponsored by Pure Storage.

Craig Waters will be talking about how to “Start a Flash Revolution in your Data Centre” while covering the following questions:

  1. What is Flash and why is it different to traditional storage?
  2. What are the benefits of this medium?
  3. How is flash being deployed in the data centre today?
  4. Demonstration on how to make storage simple

You’ll also have the opportunity to network with peers over refreshments and discover technology solutions and strategies. You can find out more information and register for the event here. I hope to see you there.

Storage Field Day 7 – Day 1 – Kaminario

Disclaimer: I recently attended Storage Field Day 7.  My flights, accommodation and other expenses were paid for by Tech Field Day. There is no requirement for me to blog about any of the content presented and I am not compensated in any way for my time at the event.  Some materials presented were discussed under NDA and don’t form part of my blog posts, but could influence future discussions.

For each of the presentations I attended at SFD7, there are a few things I want to include in the post. Firstly, you can see video footage of the Kaminario presentation here. You can also download my raw notes from the presentation here. Finally, here’s a link to the Kaminario website that covers some of what they presented.

 

Overview

Dani Golan, CEO of Kaminario, gave us a quick overview of the company. They’ve recently launched the 5th generation of their all-flash array (AFA), with the majority (80%) of customers being in the midrange (rev $100m – $5B).

The entry level for the solution is 20TB, with the average capacity being between 50 and 150TB. The largest implementation runs to 1.5PB.

Use cases are primarily:

  • VDI / Virtualisation;
  • Analytics; and
  • OLTP.

Kaminario state that they’re balanced across all verticals and offer general purpose storage.

 

Architecture

Kaminario state that architecture is key. I think we’re all agreed on that point. Kaminario’s design goals are to:

  • scale easily and cost-efficiently; and
  • provide the lowest overhead on the storage system to fulfil the customer’s needs.

Kaminario want to offer capacity, performance and flexibility. They do this by offering scale up and scale out.

Customers want somewhere in between best $/capacity and best $/performance.

The K2 basic building block (K-blocks, not 2K blocks) is:

  • Off the shelf hardware;
  • 2x K-nodes (1U server);
  • Infiniband;
  • SSD Shelf (24 SSDs – 2RU); and
  • SSD expansion shelf (24 SSDs – 2RU).

Here’s a diagram of the K2 scale up model.

SFD7_Day1_Kaminario_ScaleUp

And here’s what it looks like when you scale out.

SFD7_Day1_Kaminario_Scale_Out

I want to do both! Sure, here’s what scale up and out looks like.

SFD7_Day1_Kaminario_ScaleUpandOut

In the K2 scale-out architecture:

  • Data is spread across all nodes;
  • Metadata is spread across all nodes;
  • Provides the ability to mix and match different generations of servers and SSDs;
  • Offers global deduplication; and
  • Provides resiliency for multiple simultaneous failures.

Data is protected against block (nodes and storage) failure, but the system will go down to secure the data.

As for metadata scalability, modern data reduction means fine grain metadata:

  • Pointer per 4KB of addressable; and
  • Signature per 4KB of unique data.

According to Kaminario, reducing the metadata footprint is crucial.

  • The adaptive block size architecture means less pointers;
  • Deduplication with weak hash reduces signature footprint; and
  • Density per node is critical.

K-RAID

SFD7_Day1_Kaminario_K-RAID

K-RAID is Kaminario’s interpretation of RAID 6, and works thusly:

  • 2P + Q – 2 R5 groups, single Q parity on them;
  • Fully rotating, RAID is fully balanced;
  • Fully automatic, no manual configuration; and
  • High utilisation (87.5%), no dedicated spares.

The K2 architecture also offers the following data reduction technologies:

Deduplication

  • Global and adaptive;
  • Selective – can be turned off per volume; and
  • Weak hash and compare – low MD and CPU footprint, fits well with flash.

Compression

  • Byte-aligned compression;
  • Adaptive block size – large chunks are stored contiguously, each 4k compressed separately;
  • Standard LZ4 algorithm; and
  • Optimized zero elimination.

From a resiliency perspective, K2 supports:

  • Two concurrent SSD failures per shelf;
  • Consistent, predictable and high performance under failure; and
  • Fast SSD firmware upgrades.

The architecture currently scales to 8 K-Blocks, with the sweet spot being around 2 – 4 K-Blocks. I strongly recommend you check out the Kaminario architecture white paper – it’s actually very informative.

 

Final Thoughts and Further Reading

I first came across Kaminario at VMworld last year, and I liked what they had to say. Their presentation at SFD7 backs that up for me, along with the reading I’ve done and the conversations I’ve had with people from the company. I like the approach, but I think they have a bit of an uphill battle to crack what seems to be a fairly congested AFA market. With a little bit more marketing, they might yet get there. Yes, I said more marketing. While we all like to criticise the marketing of products by IT vendors, I think it’s still a fairly critical piece of the overall solution puzzle, particularly when it comes to getting in front of customers who want to spend money. But that’s just my view. In any case, Enrico did a great write-up on Kaminario – you can read it here. I also recommend checking out Keith’s preview blog of Kaminario.