First impressions included a marked absence of tape compared to just two years ago - remember Exabyte and VXA, and Breece Hill? Then of course there is that green thing that everyone is already tired of. It has gone from bright spot on the horizon to boring topic in a matter of months. Do I want to read another green survey suggesting that some small niche of the world's looming environmental calamity would be mitigated by buying the sponsoring vendor's product.? Why of course I do. Let's have another survey.
In alphabetic order here are some highlights for me of the show
The big man, David Scott, the CEO, said that 3PAR was in its quiet period but he could talk about trends. He thought that tier two storage needed 'hardening'. Interesting as 3PAR doesn't produce tier two storage; it's a tier one (utility) storage provider with performance for transaction-type processing as a primary driver of its technology.
He also thought that storage access control security was not standardised. Each vendor had its own access controls which meant that multi-vendor shops had to understand different was of controlling access, and had to work this into any existing Active Directory (AD) and LDAP (Lightweight Directory Access Protocol) scheme. More on this here.
David Scott also said that disaster recovery (DR) using a WAN was hindered because you typically lost the last one or two transactions when a primary site crashed. Synchronous protocols don't suffer this loss but they don't handle distance - thousands of kilometres - well because of latency issues. Async protocols do handle the distance but expose you to the risk of losing transactions immediately preceding a primary site outage.
Scott says that sorting this out is not a network plumbing responsibility. That's interesting because he is hinting that 3PAR has a solution on this problem in mind. Perhaps he is thinking of some flash memory storage being added to a 3PAR InServe array into which transactions are continually journalled, or perhaps there is a flash store and, if a power outage occurs then a battery or capacitor is used to write the in-mmory transaction data to the flash and thus preserve it?
Scott also said that 3PAR hadn't joined The Green Grid because to do so might put 3PAR at risk. He's hinting that The Green Grid folks could, if 3PAR IP was identified and chosen as necessary for greener data centres, make it publicly available. Instead 3PAR has joined the SNIA green initiative and, through it, has a watching brief type connection to The Green Grid.
The new F5 business unit's presence at Storage Expo is discussed here.
From sources close to Brocade comes confirmation that Neptune, the big, big new director, will have some details revealed at the Brocade event in Las Vegas on Monday. It will not be a pair of 48000s cobbled together because the bandwidth needed inside the big new box with 8Gbit/s ports is far higher than a pair of 48Ks could cope with.
Ideas about being over-dependent on Emulex HBAs for the server access were shrugged off. Brocade has it own HBAs and, anyway, works well with Emulex. The overall idea of the 48K storage fabric director being built on for the new box to head towards a data centre 'director' is right and server connectivity to it via Ethernet - iSCSI and FCOE - is also right.
The Ethernet is not common-or-garden Ethernet though. It is a new and lossless form of Ethernet suitable for data centre use.
Fujitsu Siemens Computers
FSC has a new FibreCAT array, the SX88. It has 12 drive slots in a 2U rack shelf with RAID 0, 1, 5 and 6 provided. I suppose we might think of it as an RBOD, a RAID'ed Bunch of Disks, it being clearly more than a JBOD. You can have four of the chassis daisy-chained together and the chassis can have serial-attached SCSI (SAS) or serial ATA (SATA) drives, with a maximum of 42 drives. One terabyte SATA drives are coming soon and the maximum capacity will then be 42TB.
FSC says it is competitively priced; not the most expensive and not the cheapest. The RAID controllers are said to be very powerful and the SX88 outperforms anything else in its class with benchmarks being released soon.
There was also a new autoloader, the TX24, with two 12-slot magazines laid out fore-aft in a rack unit and either side of a central robot. This moved quite slowly. Why was that? FSC's storage product manager, Sean Haffey, said that there were two concerns driving robot design. One as long-life and dependability. You use an autoloader because you want to load it with tapes and forget about it. A failing robot is absolutely not what you want.
Secondly, tape cartridges have to be handled with kid gloves. They are susceptible to reduced tape performance, even failure, if they are handled roughly. The TX24 robot handles them carefully and moves them smoothly.
The X24 can have on or two half-height LTO2 and LTO3 drives or one full-height LTO4 drive. When half-height LTO4 drives are available then two of them will be fitted. An LTO4 TX24 can hold 19.2TB of data.
Its also got a capacitor (FibreCAP) instead of a battery to power the writing of cache data to non-volatile memory in there is a power-faolure. This is better than batteries, taking less time to charge up amongst other things.
A brief word with Steve Mackey, Quantum's UK sales director, resulted in this assertion: nobody today realistically is going to move away from tape for long-term data retention. (Are you listening Copan and Plasmon?) Quantum is unique in being able to talk to customers about physical tape, de-duplication and virtual tape.
Um, what about Sun/StorageTek IBM and HP, who each have offerings in those three areas?
The reply was that Sun is a systems company; it is about too many products. Quantum is a specialist and only does one thing. That non-specialist tag applies to BM and HP and any other systems vendor too.
IBM has just been declared number one in branded tape systems revenue by IDC. Sun's StorageTek is a close second.
You can't store de-duped data on tape because it is a sequential access medium and can't have a file system on it and de-duping software assumes a disk is the target. But what about removable random-access storage media?
There are two kinds: Plasmon UDO optical disks with 30 and 60GB capacities; and removable hard drives such as the ProStor ones sold by Tandberg and Dell, and the IdealStor ones such as the new Teralyte. These could store de-duped data and thus have their raw capacities increased by a factor of fifteen or twenty.
This is just speculation and nobody, as far as I know, is actually thinking about putting de-duped data onto removable random-access media.
TapeWise and DPS
TapeWise is DPS software that checks out a tape and tape drive's reliability, producing a 3D graphic of errors during a run along a tape for reading and writing. It is the only application commercially available to do this and is in use at CERN to help check out its massive tape library. DPS MD Matthew Augier hinted that Tapewise will have physical tape library support extended from StorageTek with its ACLS interface to IBM tape libraries which don't use ACLS.
Then the tapes in an IBM library could be proactively checked for reliability and error-prone cartridges put into quarantine so that data for long-term retention isn't written to them and put at risk.