Posts Tagged ‘performance’

Ceph is free if your time is worth nothing!

Jyothi Swaroop

Director, Product Marketing at EMC

Latest posts by Jyothi Swaroop (see all)

Ironic how we grow up listening to our parent’s tell us “Nothing in life is free” yet the moment someone claims they have or do something for “free”, we forget that simple truth.

Is anything ever really free?
There are offerings out there today that claim they are open source, free and always will be. However, if we remember what Mom and Dad said – then we need to look deeper into this. Time, overhead and hardware requirements to run these open source solutions are not free.

For this discussion, let us take a look at Ceph and Ceph “Enterprise” editions. Ceph is an open source distributed object store and file system which claims to be free. As an open source platform the code is indeed free but if Ceph is free, why would any company pay to acquire a “commercial wrapper” for it, such as Inktank? When it comes to open source, large companies make their money by selling “enhanced” versions of the software along with professional consulting, services, and support.

Enterprise versions are not free and often expensive. Customers pay more in server hardware, server OS licenses & disk drives. Licensing and support can run as much as $4K per server.

Now, some will say, “I will go with the free version of the open source solution and not the “enhanced” or “enterprise” edition offered – it’s more cost effective and I can support it myself”. It is definitely an option, and in some instances may make sense, but before you make that commitment ask yourself:
• Can I get 24×7 worldwide support – the support I need, when I need it?
• Do I have to wait for a community to help solve my problem and even if a fix is suggested, will it work for me, in my environment?
• Will my customers wait till tomorrow or the next week for a fix?
• Does ‘ongoing support by committee’ really work?
• What am I willing to give up?

If it is free – do you get what you pay for?Car

When it comes to software-defined scale out block storage for high performing applications and / or delivering Infrastructure-as-a-Service (IaaS), “free” may not be better. Will you simply be getting what you pay for?

Starting with installation, as a software-defined offering, Ceph does not constrain or confine you like current hyper-converged appliances do. However, installation is actually extremely complex. Architecting and running Ceph as a storage platform requires deep Linux and Ceph expertise and experience. It requires a multi module, multi-step, deployment process (with different processes for each OS) which complicates management and incurs a larger performance hit to the underlying hardware. Ceph also takes a ‘non-native’ layered approach to delivering block storage, where block is on top of object. David Noy, VP Product Management at EMC pointed out in his blog last month, that with a layered approach “problems come when you have a system that is designed to optimize around the underlying abstraction and not the service layered on top”. This is evident in Ceph’s approach to block (RADOS Block Device – RBD) which has extreme overhead resulting in high latency and an inability to exploit Flash media.

OK, so you know there will be a great deal of work to set up and manage Ceph. You still feel you are ready to deal with this cryptic approach including: compile/decompile/manually edit crush maps; limited system visibility using a command line interface (CLI); and even the manual placement group (PG) planning and repair. Yes, the free approach, even with all of this, will meet your needs. Maybe but let’s not forget what really matters. When delivering IaaS or high performance applications, delays in response are simply not acceptable to your customers or users. How does Ceph measure up where it really counts: Performance and Scalability!

The Proof is in the Numbers
We recently tested Ceph against EMC ScaleIO and the findings were clear as day. Both were tested on the same hardware and configuration with the following assumptions:
• Test “SSD only” using a small volume that will fit the SSD capacity
• Test “SSD+HDD” using a small+large volume spanning HDD capacity
• Test SSD as Cache for HDD using a small+large volume spanning HDD cap
• Test a Mixed workload of 70% Reads, 30% Writes, 8KB IOs

• ScaleIO achieved ~7X better performance than the best Ceph IOPs value for a drive limited configuration
• ScaleIO achieved ~15X better performance than Ceph, when the drives are not the limit
• ScaleIO has ~24X better Response Time with an SSD only configuration
• ScaleIO can support the IOPs at 1/3rd the latency of Ceph, as a result there is no need to second guess performance for applications you run on ScaleIO.

Similar to Ceph, EMC ScaleIO is a software-only solution that uses existing commodity hardware servers’ local disks and LAN to realize a software-defined, scale-out SAN. However, EMC ScaleIO delivers elastic and scalable performance and capacity on demand and beyond what Ceph is a capable of for enterprise deployments. ScaleIO also does not require additional servers for providing storage and supports multiple platforms including OpenStack via a Cinder Plugin. It requires 1/5th to 1/10th the number of drives that Ceph needs to deliver the same performance. This actually results in significant floor and power savings.

The evidence speaks for itself when it comes to performance, scale and enterprise grade capabilities – sometimes you just get what you pay for. But don’t just take our word for it. Here is a perfect example of the kind of issues a company can face, including the potential loss of date, when delivering services with Ceph. Also, Enterprise Strategy Group (ESG) recently published a Lab Spotlight demonstrating extremely high IOPS performance and near-linear scaling capabilities of ScaleIO software on commodity hardware.

If you STILL want to be able use software for free BEFORE you make a long term, strategic commitment, EMC provides you the same opportunity with ScaleIO. May 2015, EMC is offering a free download of ScaleIO, for non-production use, for as much capacity and time as you want. You can experience all of the features and capabilities and see for yourself why enterprise grade software-defined scale out SAN with EMC ScaleIO is better than “free” with Ceph. Virtual Geek is ready! Are you?

Service Outage Hits Home for Cloud Provider

Are there blind spots in your service assurance approach?

Netflix, a provider of online streaming media, made news over the holidays when customers experienced a service outage on Christmas Eve.  Imagine taking the wrapping off of your new mobile device and deciding to try it out to stream a movie.   For those located in North America, you probably found that the Netflix movie streaming service was down. 

This outage was caused by issues within Amazon Web Services that Netflix employs to support movie streaming.  Initially, the Amazon support team pursued API errors before learning that the root cause of the outage was actually a configuration issue caused by human error.  This misstep ultimately delayed the restoration of service to Netflix customers.  Over the course of that day, the configuration error first manifested itself as performance degradation, and then cascaded to a full service outage for many customers. One way of avoiding a situation like this one could have been to take a more system-wide approach to service assurance. (more…)

Service Assurance In The Cloud

How do you provide service assurance in your virtualized data center or cloud?

Read any blog, online article, or hard copy publication about the IT industry and you will likely see much about the future of IT. But, it is the every-day challenges and victories that often determine success in the long run.

While the industry considers more logical abstractions to further loosen the ties between business processes and physical infrastructure, IT management needs to assure ongoing application availability and performance in virtual and cloud deployments with a high rate of data growth. IT needs solutions that provide a complete picture of data center health across compute, network, and storage.

The just-announced EMC Storage Resource Management Suite provides service assurance from application-to-storage by monitoring and reporting on availability, performance, and compliance. It combines EMC storage management and analysis with technologies from the recent Watch4net acquisition to ensure application consistency in dynamic data centers. (more…)

Performance Management In A Cross-Domain World

How do you ensure performance across your virtualized data center or cloud?

There is much emphasis on enterprise management tools being able to assure the availability of IT delivered services from a centralized event console.  You can search the Web and find many solutions that claim they can consolidate events into one single dashboard. However, availability is only part of the story when it comes to building an application-aware infrastructure to deliver always-on business services with better-than-expected service levels. A central management point for performance across all domains is essential in assuring delivery of business services to customers. 

Recent posts here have addressed how Web design models such as REST are setting the stage for standing up new services more quickly and providing some semblance of portability across private, public, and hybrid clouds. Standing on the brink of delivering on the promise of cloud computing, organizations need to ensure both the availability—and performance—of business-critical applications as they move to non-traditional deployments that still support demanding customers mostly disinterested in how computing gets done. (more…)

Carriers And Clouds

How do the tier-1 telecommunications carriers fit into cloud?

The large telecommunication service providers or carriers might seem well-positioned to provide and benefit from the cloud computing model. The tier-1 carriers (e.g. AT&T, France Telecom, NTT Docomo, Vodafone) have the infrastructure, enterprise customer base, and the geographic reach to provide cloud services over great distances. Additionally, they are well-versed in multi-tenancy; all I have to do is look at my monthly communications bill to get a glimpse in the detailed reporting of my mobile roaming and texting use.

However, though in a good position, these carriers need to add new proficiencies to their core competencies to meet the demands of the different size organizations that comprise their potential customer base for cloud computing services.  Much like the late 1990s when deregulation in the telecommunications industry and the shift of voice services to the Internet and the emergence of new data services, carriers need to navigate the changes to offset declining revenues in some of their traditional services with new cloud-based offerings. (more…)

Cloud Architectures Are Not New

Well, if cloud architectures are not new, exactly how old are they? The title almost begs the question.

It may be the latest, hottest buzzword, but architecturally, the cloud is more than 20 years old.  And, you are intimately familiar with it. It’s the Web. What makes it “the cloud” are the enterprise applications getting moved onto the Web.  But, this move isn’t just a matter of deploying some application servers and offering HTTP-based interfaces to existing software. It is a marked change with the architecture of these software systems built on and for the Web, rather than for traditional enterprise infrastructure.

A key tenet that makes enterprise application deployment in the cloud possible over the Web is REST. Having emerged over the past few years as a predominant Web service design model, REST has increasingly displaced other design models such as SOAP because of its simpler style. Joining me in the development of this blog post to be published when I’m on vacation is EMC senior technologist Cornelia Davis. (more…)



Connect with us on Twitter