What the “T” Means in SNIA Cloud Storage Technologies

The SNIA Cloud Storage Initiative (CSI) has had a rebrand; we’ve added a T for Technologies into our name, and we’re now officially the Cloud Storage Technologies Initiative (CSTI).

That doesn’t seem like a significant change, but there’s a good reason. Our old name reflected the push to getting acceptance of cloud storage, and that specific cloud storage debate has been won, and big time. One relatively small cloud service provider is currently storing 400PB of clients’ data. Twitter alone consumes 300PB of data on Google’s cloud offering. Facebook, Amazon, AliBaba, Tencent – all have huge data storage numbers.

Enterprises of every size are storing data in the cloud. That’s why we added the word “technologies.” The expanded charter and new name reflect the need to support the evolving cloud business models and architectures such as OpenStack, software defined storage, Kubernetes and object storage. It includes data services, orchestration and management, understanding hyperscale requirements and the role standards play.

So what do we do? The CSTI is an active group that publishes articles and white papers, speaks at industry conferences and presents at highly-rated webcasts that have been viewed by thousands. You can learn more about the CSTI and check out the Infographic for highlights on cloud storage trends and CSTI activities.

If you’re interested in cloud storage technologies, I encourage you to consider joining our group. We have multiple membership options for established vendors, startups, educational institutions, even individuals. Learn more about CSTI membership here.

Q&A – OpenStack Mitaka and Data Protection

At our recent SNIA Webcast “Data Protection and OpenStack Mitaka,” Ben Swartzlander, Project Team Lead OpenStack Manila (NetApp), and Dr. Sam Fineberg, Distinguished Technologist (HPE), provided terrific insight into data protection capabilities surrounding OpenStack. If you missed the Webcast, I encourage you to watch it on-demand at your convenience. We did not have time to get to all of out attendees’ questions during the live event, so as promised, here are answers to the questions we received.

Q. Why are there NFS drivers for Cinder?

 A. It’s fairly common in the virtualization world to store virtual disks as files in filesystems. NFS is widely used to connect hypervisors to storage arrays for the purpose of storing virtual disks, which is Cinder’s main purpose.

 Q. What does “crash-consistent” mean?

 A. It means that data on disk is what would be there is the system “crashed” at that point in time. In other words, the data reflects the order of the writes, and if any writes are lost, they are the most recent writes. To avoid losing data with a crash consistent snapshot, one must force all recently written data and metadata to be flushed to disk prior to snapshotting, and prevent further changes during the snapshot operation.

Q. How do you recover from a Cinder replication failover?

 A. The system will continue to function after the failover, however, there is currently no mechanism to “fail-back” or “re-replicate” the volumes. This function is currently in development, and the OpenStack community will have a solution in a future release.

 Q. What is a Cinder volume type?

 A. Volume types are administrator-defined “menu choices” that users can select when creating new volumes. They contain hidden metadata, in the cinder.conf file, which Cinder uses to decide where to place them at creation time, and which drivers to use to configure them when created.

 Q. Can you replicate when multiple Cinder backends are in use?

 A. Yes

 Q. What makes a Cinder “backup” different from a Cinder “snapshot”?

 A. Snapshots are used for preserving the state of a volume from changes, allowing recovery from software or user errors, and also allowing a volume to remain stable long enough for it to be backed up. Snapshots are also very efficient to create, since many devices can create them without copying any data. However, snapshots are local to the primary data and typically have no additional protection from hardware failures. In other words, the snapshot is stored on the same storage devices and typically shares disk blocks with the original volume.

Backups are stored in a neutral format which can be restored anywhere and typically on separate (possibly remote) hardware, making them ideal for recovery from hardware failures.

 Q. Can you explain what “share types” are and how they work?

 A. They are Manila’s version of Cinder’s volume types. One key difference is that some of the metadata about them is not hidden and visible to end users. Certain APIs work with shares of types that have specific capabilities.

 Q. What’s the difference between Cinder’s multi-attached and Manila’s shared file system?

A. Multi-attached Cinder volumes require cluster-aware filesystems or similar technology to be used on top of them. Ordinary file systems cannot handle multi-attachment and will corrupt data quickly if attached more than one system. Therefore cinder’s multi-attach mechanism is only intended for fiesystems or database software that is specifically designed to use it.

Manilla’s shared filesystems use industry standard network protocols, like NFS and SMB, to provide filesystems to arbitrary numbers of clients where shared access is a fundamental part of the design.

 Q. Is it true that failover is automatic?

 A. No. Failover is not automatic, for Cinder or Manila

 Q. Follow-up on failure, my question was for the array-loss scenario described in the Block discussion. Once the admin decides the array has failed, does it need to perform failover on a “VM-by-VM basis’? How does the VM know to re-attach to another Fabric, etc.?

A. Failover is all at once, but VMs do need to be reattached one at a time.

 Q. What about Cinder? Is unified object storage on SHV server the future of storage?

 A. This is a matter of opinion. We can’t give an unbiased response.

 Q. What about a “global file share/file system view” of a lot of Manila “file shares” (i.e. a scalable global name space…)

 A. Shares have disjoint namespaces intentionally. This allows Manila to provide a simple interface which works with lots of implementations. A single large namespace could be more valuable but would preclude many implementations.

 

 

Open Source Software-Only Storage – Really.

Virtually any storage solution is more parts software than hardware. Having said this, users don’t care as much about the percentage of hardware vs. software. They want their consumption experience to be easy and fast to start up, with a pay-as-you-grow model and with the ability to scale without limits. So, it should not be a shock that real IT organizations are using software-only on standard servers to deliver storage to their customers. What’s more, this type of storage can be powered by open source.

At the upcoming SNIA Data Storage Innovation Conference, we are looking forward to discussing software-defined storage (SDS) from a user experience perspective with examples of OpenStack Swift providing an engine for building SDS clusters with any mixed combination of standard server and HDD hardware in a way that is simple enough for any enterprise to dynamically scale.

Swift is a highly available, distributed, scalable object store available as open source.  It is designed to handle non-relational (that is, not just simple row-column data) or unstructured data at large scale with high availability and durability.  For example, it can be used to store files, videos, documents, analytics results, Web content, drawings, voice recordings, images, maps, musical scores, pictures, or multimedia. Organizations can use Swift to store large amounts of data efficiently, safely, and cheaply. It scales horizontally without any single point of failure.  It offers a single multi-tenant storage system for all applications, the ability to use low-cost industry-standard servers and drives, and a rich ecosystem of tools and libraries.  It can serve the needs of any service provider or enterprise working in a cloud environment, regardless of whether the installation is using other OpenStack components.

I know what you are thinking, storage is too critical, so it will never work this way. But the same was said >25 years go when using RAID was seen as too risky given solutions would acknowledge writes while the data was in cache prior to being written to disk. The same was also said >15 years ago when VMware was seen as not robust enough to run any manner of demanding or critical application. Replicas and Erasure Codes are analogous to RAID 1 and RAID 5 respectively, and the uniquely as possible distribution of data behind a single namespace abstracts standard hardware like server virtualization.

Interested in hearing more? Come check out my DSI session, “Swift Use Cases with SwiftStack,” where we look forward to sharing how this new type of storage can work, and to suspend your disbelief that this storage can be enterprise-grade.

 

Data Protection and OpenStack Mitaka

Interested in data protection and storage-related features of OpenStack? Then please join us for a live SNIA Webcast “Data Protection and OpenStack Mitaka” on June 22nd. We’ve pulled together an expert team to discuss the data protection capabilities of the OpenStack Mitaka release, which includes multiple new resiliency features. Join Dr. Sam Fineberg, Distinguished Technologist (HPE), and Ben Swartzlander, Project Team Lead OpenStack Manila (NetApp), as they dive into:

  • Storage-related features of Mitaka
  • Data protection capabilities – Snapshots and Backup
  • Manila share replication
  • Live migration
  • Rolling upgrades
  • HA replication

Sam and Ben will be on-hand for a candid Q&A near the end of the Webcast, so please start thinking about your questions and register today. We hope to see you there!

This Webcast is co-sponsored by two groups within the Storage Networking Industry Association (SNIA): the Cloud Storage Initiative (CSI), and the Data Protection & Capacity Optimization Committee (DPCO).

 

Come See SNIA at the Software-Defined Infrastructure Summit

Demand for software-defined infrastructure (SDI) is on the rise, and with good reason. SDI helps data centers meet the challenges of cloud computing, big data/analytics, mobility and social media, in an agile and cost-effective way.  I’m pleased to announce that SNIA will be an active participant at next week’s Software-Defined Infrastructure Summit in Santa Clara, CA, December 1-3.

My colleagues and I at the SNIA Cloud Storage Initiative have organized a “Working with OpenStack” Seminar that kicks off the Summit on Tuesday, December 1.

I will keynote an OpenStack fireside chat along with Chris DePuy, VP, at Dell’Oro Group. We’ll be discussing the SNIA Cloud Data Management Interface (CDMI) and its interface with OpenStack, OpenStack implementations, how standards play, and the future of open source in the 21st century.

My keynote will be accompanied by additional SNIA talks in the Introduction to OpenStack session and the Application Management session:

  • Sam Fineberg, PhD, SNIA Cloud Storage Initiative member and Distinguished Technologist at Hewlett Packard Enterprise Storage, will provide an overview of the storage aspects of OpenStack including the core projects for block storage (Cinder) and object storage (Swift), and the new shared file service (Manila). He’ll cover some common configurations and use cases for these technologies, and discuss how they interact with the other parts of OpenStack.
  • Richelle Ahlvers, SNIA Open Source Task Force member and Principal Storage Management Architect at Avago Technologies, will discuss application integration in OpenStack and how SNIA-developed standards enable cross-vendor management interoperability and help open source projects interoperate with more industry solutions.

Tuesday’s Seminar day will include additional sessions from leaders in OpenStack, Ceph, and Software Defined Storage. SDI Summit days 2 and 3 will provide information on hardware, software, and data center technology and applications of software-defined infrastructure featuring keynotes from IBM, Intel, Red Hat, and VMware, all SNIA member companies.  It’s a must attend event.

SNIA will also be exhibiting at the Summit. Please stop by booth #408 to learn how SNIA standards are used in open source projects including cloud data management, non-volatile memory, self-contained information retention, and storage management. We will also have information on SNIA programs such as membership, certification, conformance testing, and conferences.

SNIA members and colleagues can use the code SPGP to receive a $100 discount on any level of SDI Summit registration. I hope to see you in Santa Clara!

See SNIA at OpenStack Summit Tokyo

Are you headed to the OpenStack Summit in Tokyo later this month? If so, I encourage you to stop by two “Birds of a Feather” (BoF) sessions I’ll be hosting on behalf of SNIA. Here’s the info on both of them:

Extending OpenStack Swift with S3 and CDMI Interfaces – Tues. Oct. 27th 11:15 a.m.

Cloud application developers using the OpenStack infrastructure are demanding implementations of not just the Swift API, but also the S3 defacto and CDMI standard APIs. Each of these APIs not only offers features in common, but also offers what appear to be unique and incompatible facilities. At this BoF, we’ll discuss how to: Implement a multi-API strategy simply and effectively, sensibly manage the differences between each of the APIs, map common features to each other, take advantage of each of the APIs’ strengths, avoid lowest common denominator implementations

Object Drive Integration with Swift – Thurs. Oct. 29th 9:00 a.m.

With the emergence of disk drives and perhaps solid state drives with Key/Value and other object interfaces, what are the implications on solution architectures and systems built around OpenStack Swift. One approach is termed “PACO” where the Object Node speaks Key/Value to the drive and is hosted with other Swift Services. Are there other approaches to this? Are you developing products or solutions based on Object Drives? Come to this BoF to discuss these issues with fellow developers.

I expect both of these BoFs will be full of lively discussions around standards, emerging technologies, challenges, best practices and more. If you have any questions about these sessions or about work that SNIA is doing, do not hesitate to contact me. I hope to see you in Tokyo!

 

 

 

OpenStack File Services for HPC Q&A

We got some great questions during our Webcast on how OpenStack can consume and control file services appropriate for High Performance Computing (HPC) in a cloud and multi-tenanted environment. Here are answers to all of them. If you missed the Webcast, it’s now available on-demand. I encourage you to check it out and please feel free to leave any additional questions at this blog.

Q. Presumably we can use other than ZFS for the underlying filesystems in Lustre?

A. Yes, there a plenty of other filesystems that can be used other than ZFS. ZFS was given as an example of a scale up and modern filesystem that has recently been integrated, but essentially you can use most filesystem types with some having more advantages than others. What you are looking for is a filesystem that addresses the weaknesses of Lustre in terms of self-healing and scale up. So any filesystem that allows you to easily grow capacity whilst also being capable of protecting itself would be a reasonable choice. Remember, Lustre doesn’t do anything to protect the data itself. It simply places objects in a distributed fashion of the Object Storage Targets.

Q. Are there any other HPC filesystems besides Lustre?

A. Yes there are and depending on your exact requirements Lustre might not be appropriate. Gluster is an alternative that some have found slightly easier to manage and provides some additional functionality. IBM has GPFS which has been implemented as an HPC filesystem and other vendors have their scale-out filesystems too. An HPC filesystem is simply a scale-out filesystem capable of very good throughput with low latency. So under that definition a flash array could be considered a High Performance storage platform, or a scale out NAS appliance with some fast disks. It’s important to understand you’re workloads characteristics and demands before making the choice as each system has pro’s and con’s.

Q. Does “embarrassingly parallel” require bandwidth or latency from the storage system?

A. Depending on the workload characteristics it could require both. Bandwidth is usually the first demand though as data is shipped to the nodes for processing. Obviously the lower the latency the fast though jobs can start and run, but its not critical as there is limited communication between nodes that normally drives the low latency demand.

Q. Would you suggest to use Object Storage for NFV, i.e Telco applications?

A. I would for some applications. The problem with NFV is it actually captures a surprising breadth of applications so of which have very limited data storage needs. For example there is little need for storage in a packet switching environment beyond the OS and binaries needed to stand up the VM’s. In this case, object is a very good fit as it can be easily, geographically distributed ensuring the same networking function is delivered in the same manner. Other applications that require access to filtered data (so maybe billing based applications or content distribution) would also be good candidates.

Q. I missed something in the middle; please clarify, your suggestion is to use ZFS (on Linux) for the local file system on OSTs?

A. Yes, this was one example and where some work has recently been done in the Lustre community. This affords the OSS’s the capability of scaling the capacity upwards as well as offering the RAID-like protection and self-healing that comes with ZFS. Other filesystems can offer those some things so I am not suggesting it is the only choice.

Q. Why would someone want/need scale-up, when they can scale-out?

A. This can often come down to funding. A lot of HPC environments exist in academic institutions that rely on grant funding and sponsorship to expand their infrastructure. Sometimes it simply isn’t feasible to buy extra servers in order to add capacity, particularly if there is already performance headroom. It might also be the case that rack space, power and cooling could be factors in which case adding drives to cope with bigger workloads might be the only option. You do need to consider if the additional capacity would also provoke the need for better performance so we can’t just assume that adding disk is enough, but it’s certainly a good option and a requirement I have seen a number of times.

 

OpenStack File Services Options

How can OpenStack consume and control file services appropriate to High Performance Compute (HPC) in a cloud and multi-tenanted environment? Find out on September 22nd when SNIA Cloud hosts a live Webcast and examines two approaches to integration.

One approach is to have OpenStack manage the storage infrastructure services using Cinder, Nova and Neutron to provide HPC Filesystem as a Service.

A second option is to use Manila file services for OpenStack to control the HPC File system deployment and manage the exports etc. This part also looks at the creation (in progress) of the Lustre Manila driver and its current progress.

I hope you’ll join Alex McDonald and me as we discuss the pros and cons of each approach. Register today and

Cloud Storage Development Challenges – An SDC Preview

This year’s Storage Developer Conference (SDC) is expected to draw over 400 storage developers and professionals. On August 4th, you can get a sneak preview of key cloud topics that will be covered at SDC in this live Webcast where David Slik and Mark Carlson Co-Chairs of the SNIA Cloud Technical Work Group, together with Yong Chen, Assistant Professor at Texas Tech University will discuss:

  • Mobile and Secure – Cloud Encrypted Objects using CDMI
  • Object Drives: A new Architectural Partitioning
  • Unistore: A Unified Storage Architecture for Cloud Computing
  • Using CDMI to Manage Swift, S3, and Ceph Object Repositories

You’ll learn how encrypted objects can be stored, retrieved, and transferred between clouds, how Object Drives allow storage to scale up and down by single drive increments, end-user and vendor use cases of the Cloud Data Management Interface (CDMI), and we’ll introduce Unistore – an innovative unified storage architecture that efficiently integrates heterogeneous HDD and SCM devices for Cloud storage systems.

I’ll be moderating the discussion among this expert panel. It should be an enlightening and lively hour. I hope you’ll register now to join us.

 

Swift, S3 or CDMI – Your Questions Answered

Last week’s live SNIA Cloud Webcast “Swift, S3 or CDMI – Why Choose?” is now available on demand. Thanks to all the folks who attended the live event. We had some great questions from attendees, in case you missed it, here is a complete Q&A.

Q. How do you tag the data? Is that a manual operation?

A. The data is tagged as part of the CDMI API by supplying key value pairs in the JSON Object. Since it is an API you can put a User Interface in front of it to manually tag the data. But you can also develop software to automatically tag the data. We envision an entire ecosystem of software that would use this interface to better manage data in the future

Q. Which vendors support CDMI today?

A. We have a page that lists all the publically announced CDMI implementations here. We also plan to start testing implementations with standardized tests to certify them as conformant. This will be a separate list.

Q. FC3 Common Services layer vs. SWIFT, S3, & CDMI – Will it fully integrate with encryption at rest vendors?

A. Amazon does offer encryption at rest for example, but does not (yet) allow you choose the algorithm. CDMI allows vendors to show a list of algorithms and pick the one they want.

Q. You’d mentioned NFS, other interfaces for compatibility – but often “native” NFS deployments can be pretty high performance. Object storage doesn’t really focus on performance, does it? How is it addressed for customers moving to the object model?

A. CDMI implementations are responsible for the performance not the standard itself, but there is nothing in an object interface that would make it inherently slower. But if the NFS interface implementation is faster, customers can use that interface for apps with those performance needs. The compatibility means they can use whatever interface makes sense for each application type.

Q. Is it possible to query the user-metadata on a container level for listing all the data objects that have that user-metadata set?

A. Yes. Metadata query is key and it can be scoped however you like. Data system metadata is also hierarchical and inherited – meaning that you can override the parent container settings.

Q. So would it be reasonable to say that any current object storage should be expected to implement one or more of these metadata models? What if the object store wasn’t necessarily meant to play in a cloud? Would it be at a disadvantage if its metadata model was proprietary?

A. Yes, but as an add-on that would not interfere with the existing API/access method. Eventually as CDMI becomes ubiquitous, products would be at a disadvantage if they did not add this type of interface.