Avere to tart up FXT with cloud storage gateway, mutterings foretell
'Talk to the clouds, man, they'll answer you'
SaaS data loss: The problem you didn’t know you had
Avere is working on a strategy to front-end cloud object storage products with FXT, its filer-accelerating NVRAM, flash and disk tiered system technology, say storage insiders.
Conversations with people close to the situation suggest that customers using Avere's clustered filer acceleration products - and there are about 100 customers now - are keen to push Avere into making its FXT products talk to an cloud object-based storage service such as Amazon's S3, the de facto standard, and RackSpace.
There is background information about Avere's filer accelerating FXT technology here.
The idea is to use the FXT's caching and scalability and NAS support to talk to the cloud object stores and provide fast and local access to data for (and from) the cloud in the FXT caches. It means adding a cloud access layer to the FXT and our understanding is that this technology has been developed. Further we believe that Avere is close to the point at which customers might be invited to take part in alpha testing of a technology demonstrator or proof of concept exercise.
Avere has not been focusing on object storage product suppliers although it has, we understand, had discussions with several of them. It occurs to us that having a single logical accelerating resource talking to a multi-node object storage system through a single access point would not be a good idea.
Why would customers want an Avere cloud storage gateway when they could get one from Nasuni? Our belief is that the reason is performance. Nasuni products are good for accessing nearline and archive data in the cloud whereas Avere's FXT would enable fast access to primary data in the cloud. Panzura and Riverbed's Granite also provide cloud storage gateway functionality and how they would relate to any Avere product is not known.
Fujitsu and Avere
Separately Fujitsu and Avere have signed a deal to have a converged system which bundles a Fujitsu Core filer, with dual active-active UDS NAS controllers, an ETERNUS DX80 S2 Disk array and a two-node Avere FXT 3100 Edge filer cluster. It should provide customers with 100TB of capacity - roughly - and 100,000 IOPs for $100,000, hence its 100/100/100 moniker.
It should scale up to 2.5 million IOPs and 2PB capacity. The Fujitsu storage provides the base capacity - one control enclosure with dual RAID controller and a dozen 3TB 7.2K SATA drives plus three storage trays with a dozen similar drives each - and data management with the Avere accelerators turbo-charging the base performance. Affordability comes from the low-cost SATA drives used and performance from the FXT's caching, namespace and I/O handling capability.
Avere is making steady progress building out from its initial base of a multi-tiered caching filer-accelerator product and its message of replacing expensive filers with lower cost ones is resonating with customers faced with a mismatch between growing storage needs and non-growing IT budgets. Now they want to take the next step and move the data, which is still growing, off the filers completely and into the cloud, thus sidestepping the other problem many face which is limited data centre floorspace.
That's the object of Avere's object storage access exercise. ®
COMMENTS
no such thing
as fast access to primary data in the cloud, unless the thing accessing that data lives in the same cloud. Even then your not likely to get it fast (relative to something like Avere's technology)
In fact the whole concept is totally stupid, you better hope for 100% cache hit rate because the moment you have to go to the cloud to get some of that data your stuff is going to come to a grinding halt. Latencies will go from microseconds(I assume) on the Avere SSD cache to minutes or even hours when data is pulled in from the cloud. Meanwhile users will be confused why everything just came to a grinding halt, it's not as if your NFS or CIFS system can present a message to the client(that I know of) saying "please wait while we get the data for you ...................................."
It's not Avere's fault for this latency of course, the latency comes from the distance between the client (Avere) and the cloud, having a fancy tiering system doesn't make network packets go any faster.
Though you could slash efficiency and blow up your bandwidth costs by chunking your data out in lots of pieces and accessing it in parallel, but unless your limiting yourself to very infrequent accesses I suspect the cost of bandwidth will outweigh the cost savings of cloud storage very quickly.
does Avere accelerate writes or only reads? Looks like they do writes as well. Most of those ssd acceleration thingies only accelerate reads.
IT infrastructure monitoring strategies
Agentless Backup is Not a Myth
Top 10 SIEM implementer’s checklist
Steps to Take Before Choosing a Business Continuity Partner
Enabling efficient data center monitoring
