* Posts by TomLeyden

9 publicly visible posts • joined 26 Mar 2012

Data Direct Networks prepares to board Amazon S3. It's cutlasses now, lads


Hi KX1B,

It's all about scale (and performance, and cost, and reliability). The most important thing we learn about Nirvanix is not about how they failed to take marketshare from Amazon (that was due to bad messaging/marketing/positioning/strategy) but how customers got themselves into the trouble they are having now getting their data off the Nirvanix cloud.

Their is great benefit in public cloud solutions - especially for compute, application delivery etc. But it's not a safe place for your data, and it is expensive. The DDN customers choose for WOS because of that: once their data sets exceed the 500TB limit we can offer them a serious cost benefit, plus they will never face the problem of suddenly having to take their data off a public cloud service.

Here is one blog post I wrote about the topic http://www.ddn.com/blog/?p=494

A Nirvanix-specific post is queued to be posted this week.


Re: Another Proprietary Box

* Because WOS is in the first place a software platform, but our customers happen to like the super fast hardware we make too. We are happy to have a software-only conversation, "Chaosfreak".

* Because you may be up for a management nightmare if you are running a petabytes infrastructure on commodity hardware & Swift.

* Because OpenStack isn't necessarily open either: it's as much influenced by vendors as WOS is.

* Because DDN WOS offers a lot more choices in respect to API's than Swift, so as a matter of fact it is more open, odd as this may sound.

* Because when you include the integration cost, the management cost and the cost of running a storage cloud on a less than optimal architecture, your TCO will be much higher than that of WOS.

* Because WOS provides the choice between replication (sync & async), erasure coding and replicated erasure coding.

* Because WOS is faster: http://www.ddn.com/blog/?p=532

Is Object storage really appropriate for 100+ PB stores?


time will tell but


Of course there is a lot of buzz and marketing around Object Storage and Big Data, but isn't that normal? Back in 2009, more than one person claimed Cloud Computing would never break through. One of them runs a database company that claims to be "the" Cloud company today...

The market drivers are there, the technology follows. And there are a good number of use cases out there as well. Amazon S3 has been mentioned, but there is also Facebook, Google and others.

The Cern reference, which I also saw in your Netapp article is not the greatest for a number of reasons:

* Their choice for tape is two years old

* TDuskin's comment is correct

* It's not the easiest use case; a better one would be Facebook: are they storing the massive numbers of pictures on tape? Or youtube its videos?

So, Chris, object storage may need some time to become a standard, and there might be something else coming up in a couple of decades, but for now it is a much more scalable alternative than file based storage when storage large volumes of unstructured data (the better use case). Also, when designed with the proper Erasure Coding, it's a lot more efficient and reliable than traditional technologies.

Disclaimer: I work for Amplidata, an erasure coding based object storage vendor

Google Drive Issues on Mac


I really really want to love this product, but so far I can't ...

Logged in with the wrong account; found no way to change that account but reinstalling the app.

So far it seems there is no multiple-account support. Kind of lame.

I do wonder though whether they will fix my main Dropbox issues:

*everyone (i.e. all shares) pays for the same file shared by a person

* you can't share a top level folder with one person and a lower level folder with another person (everyone gets access to top level without any notification)

Don't want unified file, block and object storage


1 trillion objects in S3, guys

We will indeed have to motivate application providers to use object interfaces and get rid of legacy protocols. File systems will never really disappear, but applications will hopefully evolve and use more direct interfaces.


Correct, GPFS has good performance (throughput) and was designed to scale out. But object storage has many more benefits: REST (no file system involved) has become very popular with developers in a short term. But it's not only about the interface; the better object storage platforms will provide much better storage efficiency than systems that rely on data copies, etc. I haven't seen an IBM TCO for a 50PB system that comes close to the cost of tape. Systems like Amplidata's Amplistor will provide ten 9's with less than 60% overhead. And S3, one trillion objects can't be wrong? Right?

Object storage and tape


Disk archives at the cost of tape


Providing actual cost per GB numbers would probably be the easiest way to demonstrate how close disk archives can come to the cost per GB for tape archives. Unfortunately we cannot just throw our list prices in the open. So although we do not believe tape will disappear and we are making a case for different active archive architectures, here is an attempt to explain why and how disk archives can be more attractive than tape archives from a TCO point of view.

An accurate TCO analysis of a complete HSM tape solution should take into account the servers (MetaData HA servers, Data Movers, NFS/CIFS gateways), the Fibre Channel redundant switches, FC tier 1 disk cache, etc. Add to this, the cost of HSM software licenses at $200-500/TB, and you eliminate most of the cost benefit of tape when compared to an AmpliStor solution. For an accurate non-HSM TCO, one needs to include the cost of backup or archive software, backup servers, etc., which again are overall as or more expensive than tape libraries. Also, tape library costs are only part of the whole, and the Active Archive Alliance’s (AAA) position that "disk is 15 times more costly than tape and uses 200 times more energy" for active archive is based on old and incomplete information: the AAA uses 5+ year old papers (Clipper Group), out of date disk pricing, and projects tape costs using the largest libraries only.

It is true that for the largest tape libraries, say over 10PB, the cost for two copies on tape is in the $200-300/TB range, uncompressed. If the data compresses, they are still untouchable. And, if they have already paid for the large libraries, then they will just point to the media cost, which is more like $50/TB. But for the majority of installed libraries in both backup and active archive, these libraries will cost from $450 to $800/TB for two copies across two libraries in separate datacenters. AmpliStor can go well below that cost. Any enterprise facing replacement of old tape libraries and renewing costly archive or HSM software licenses, will find the AmpliStor value proposition compelling.

At Amplidata we are merging the disk-based object store and eliminating at least a majority of the cost for libraries and software licensing for the tape tier, assuming a geo-spread configuration. Other less durable storage can't say that, as users need to keep the full two copies on tape (if they have a DR strategy). This is exactly what resonates with our customers: eliminating the need for tape or replicated high-end disk systems.

So while we don’t advocate moving away from tape altogether, our customers could do so. Our exact messaging on this is: “The AmpliStor system delivers such high levels of data durability (ten 9s, fifteen 9s – or beyond) that it can actually consolidate multiple existing tiers of storage. Typically customers deploy these multiple tiers to provide assurance for data durability: tier 1 (disk) + tier 2 (disk), or tier 2 (disk) + tier 3 (tape), or multiple copies on tape. With AmpliStor, a single instance of data provides much higher durability, thereby leading to tremendous savings on both capital and operational costs. AmpliStor can even come close to the TCO of two copies on tape, while providing the convenience, management ease and accessibility of disk, along with scalable, multi-gigabyte throughput.”


Re: Amplidata costs

I don't have anything at hand, but let me work on something and get it to you before the end of the week!


Tape will never disappear ...

(disclaimer: I work for Amplidata, I might be biassed)

... but the use cases continue to become thinner:

While tape was traditionally the medium to go for when archiving data, more and more companies are seeking to archive to disk.

Tape is probably the cheapest medium to archive data too, but also the slowest. As a result, tape archives are usually dead archives. By building disk-based archives, with much better data access speeds, companies can re-activate their archives into rich repositories of resources. (turn your archive into a profit center)

For this purpose, object storage is becoming a popular paradigm: object storage enables companies to deploy applications directly on top of the storage, with no file systems in between.

Some of the latest object storage systems have a cost per GB (including infrastructure, management & power) that comes close to the cost of tape, which makes tape archives obsolete. That said, there several products on the market (e.g. QStar) that enable companies to move data off their object storage disk archive to tape.