What the heck is tail latency?

Mark May elaborates on the problem of tail latency, as presented on by SNIA at Storage Field Day. According to SNIA, these spikes in latency can be up to ten times as bad as average latency. Mark says that this problem could be fixed by building smarter software that can tag an IO with a retry policy hint, allowing it to favor response time over data recovery. According to SNIA, over 2% of IO may be suffering due to tail latency. Mark knows this is a real problem and hopes it can be fixed in the near future.


Excelero NVMesh: lightning fast software-defined storage using commodity servers & NVMe drives

Jon Klaus came away from March’s Storage Field Day impressed with Excelero. The company came out of stealth at the event with their software defined block scale-out storage, NVMesh. This uses commodity servers and NVMe drives to deliver impressive performance. Jon gives a review of their overall architecture, their value proposition, and general impressions from their presentation.


The Year of Cloud Extension

The idea of incorporating cloud storage into the data center has been around for a while. But Stephen Foskett thinks we’re seeing philosophically different approaches to it recently, with many companies embracing the premise of data non-locality. Stephen sees this change in the assumption from data being tied to a data center to the cloud as a transformative shift allowing for true data center transformation.


Excelero’s NVMesh Magic

Excelero made a big splash with delegates from their debut at Storage Field Day in March. Rich Stroffolino has been mulling over their presentation and wrote up his thoughts. The company’s scale out storage architecture really seemed to excite Rich at the possibility, as it was able to achieve millions of IOPS on relatively cheap hardware. This combination, despite a dearth of data services, seemed to open up a lot of possibilities.


Datera Elastic Data Fabric

Rich Stroffolino takes a look at Datera’s Elastic Data Fabric, as presented at Storage Field Day in March. With this solution, Datera hopes to provide a way to make on-site storage as flexible and automated as public cloud offerings, and to provide easy ways for organizations to adopt a hybrid cloud strategy.


Back to the Storage Future with Intel’s SPDK

Rich Stroffolino gives an over of Intel’s SPDK, which includes a storage driver that bypasses the kernel for lower latency and better scaling with NVMe storage. Intel’s software approach to performance optimization perhaps signals a performance plateau based on hardware advances alone.


Moving to and between clouds made simple with Elastifile Cloud File System – FastStorage

Moving data to the cloud and moving between clouds are some of the primary challenges facing organizations today. Often cloud provider lock-in is a very real concern. From what Jon Klaus saw from Elastifile, their Elastifile Cloud File System is the secret to make both easier. ECFS is designed for web scale clients and users, and can be extended into the public cloud. Designed with true distribution in mind, the file system should allow you to easy and reliably move data between on-premises and the public clouds.


The Future of On-Prem in a Cloud World

Rich Stroffolino shared an interview from ZDNet, considering what the roll of the cloud would be in the data center. Rich found it part of an interesting trend in the enterprise, seeing cloud computing as less of a zero-sum game with on-prem, and more of a spectrum. In the interview, Michael Howard and Monty Widenius both actually predict a move back to on-prem infrastructure, as cost and performance are better presented vs the seeming ease of the cloud. It’s an interesting debate, we’ll see how the winds of change impact the role of the public cloud going forward.


Open19 Brings a new build paradigm to HyperScale Buildouts

At Storage Field Day, Excelero gave some of their presentation time to Open19, an project from LinkedIn to standardize and speed data center deployment. Using a Lego like approach to speed plugin time, their technical specifications allow for delivery of a rack of components in as little as two hours, down from an estimated eight. While this approach might not fit data centers with a lot of legacy equipment, Matt seems to like it for new build outs. Make sure to check out Matt’s piece and their video from the event for more specifics.


Excelero achieves amazing stats at #SFD12

Matt Leib got to see Excelero’s debut at Storage Field Day last month. It’s a presentation that’s gotten a lot of the delegates excited, and Matt is certainly no exception. Their storage solution allows you to access disks in a storage array via their own RDDA, essentially NVME over Fabric. This makes the overhead for disk access negligible. Using commodity networking, this essentially allows the NVMe drives to effortlessly scale. The company demoed getting millions of Iops with minimal latency on hardware costing less than $15,000. Matt summed the presentation up by calling it “[a]stounding technology”.


SNIA: Avoiding tail latency by failing IO operations on purpose

At Storage Field Day last month, Jon Klaus saw a presentation from SNIA, where they talked about tail latency. These are tiny spikes in latency, 2-10x longer than higher. These can have big impacts, effecting up to 2% of all IO, which is magnified when seen at a hyperscale level. Smaller organizations might simply over provision to lower overall latency so that these spikes still fall within acceptable ranges, but at scale this approach isn’t practical. SNIA presented several ways to overcome tail latency, and Jon shares his thoughts on what he saw.


There’s a new cluster filesystem on the block, Elastifile

Elastifile debuted their new file system at Storage Field Day last month, and Ray Lucchesi wrote up his thoughts. It’s designed to support thousands of nodes, exabytes of capacity, and infinite numbers of files, in an effort to make a better cluster file system. It was in development since 2013, and offers some impressive features, including compression, deduplication, and cloud storage tiering. It only caches metadata, and maintains consistency thanks to key-value consensus based algorithm called Bizur. Ray’s not sure how it’ll perform in terms of marketshare going forward, but thinks it shows a lot of great backend engineering to offer a competitive file system right out of the gate.


Nimble’s InfoSight – An #SFD12 Follow-up

Matt Leib reviews what he saw from Nimble Storage at last month’s Storage Field Day. He goes into what he really liked about their InfoSight solution. This allows collect information across environments and send it to a secure collective database. This allows you to quickly cross reference anomalies across environments to better diagnose problems. Overall, Matt came away really impressed.


Impact from Public Cloud on the storage industry – An insight from SNIA at #SFD12

SNIA, the Storage Networking Industry Association, presented at Storage Field Day last month, discussing the impact of the public cloud on the overall storage industry. These hyperscale companies operate differently, they buy storage in bulk, but not from traditional enterprise storage companies like Dell EMC or NetApp. Instead, they buy disk directly from manufacturer, often at extremely large scale. Chan Ekanayake thinks that while this will prove to be disruptive for traditional storage companies, overall it will push innovation and be better in the end for customers.


Elastifile launches cross-cloud data fabric

Elastifile launched its Cloud File System at Storage Field Day last month. Andrea Mauro gives the details a run down in this blog post. They’ve designed it as a software-only, scale-out, elastic and flash-native solution, which can bring the efficiency normally found in hyperscale to the enterprise and cloud service providers.


What Next for the new Tier 0 Storage?

Chris Evans takes a look at the state of tier 0 storage, the custom hardware platforms that can deliver millions of IOPS. This was given a shake up recently with Dell EMC’s announcement that they aren’t moving forward with DSSD. Chris thinks the problem with these custom platforms is they can’t keep up with the rapid pace of flash storage development. He sees the future of tier 0 high performance storage with software defined solutions on commodity hardware, like what we’ve seen recently out of Excelero and E8 Storage.


Visiting Intel with SFD 12

Howard Marks was at Storage Field Day and got to see the latest and greatest from Intel. As the current de facto CPU and chipset vendor in the data center, Howard sees them as having an immense responsibility in terms of dictating computer architecture. Luckily from what Howard saw at Intel’s presentation, they are living up to this. As an example, he goes into the applications and implication of their Storage Performance Developer Kit, as well as their recent release of the first commercial 3D XPoint.


Storage Field Day 12 – Wrap-up and Link-o-rama

Dan Frith has done extensive writing about all the presenters and happenings at Storage Field Day earlier this month. He’s compiled a list of all of the posts for easy reference. Make sure to check them all out for Dan’s usual excellent analysis and commentary.


Netapp – an #SFD12 Update

When a company with NetApp’s extensive history and portfolio presents at Storage Field Day, sometimes you don’t know what to expect. Matt Leib certainly found a lot to like from their presentation earlier this month. He got a glimpse into how the company is proceeding in light of the SolidFire acquisition, and remained impressed by their approach. The company has managed to integrate the company without quashing its consistent innovation, always a challenge for a large company like NetApp. Matt saw this innovative spirit actually flowing back into some of NetApp’s new and updated solutions.


SNIA Know What Time It Is

Dan Frith looks at what SNIA presented at Storage Field Day earlier this month. He found it an intriguing presentations, detailing the world of storage in the hyperscaler world. He makes the case for why SNIA has carved themselves out as a valuable asset and standards body in the enterprise storage world.