Ok so further to my previous blog entry http://grumpystorage.blogspot.com/2009/10/rfe-for-rfes.html , and in the spirit of sharing, here's my draft list of 'non-NDA' RFEs :-
· Ability to terminate multiple VSANs onto a single array FE port directly without needing to use IVR and/or dedicated array FE ports per VSAN
· XML interface for customer retrieval EOL/EOSL information, eg XML to complement :-
NetApp http://now.netapp.com/NOW/products/eoa/
EMC
"PowerLink -> Home > Support > Interoperability and Product Lifecycle Information > Release and End of Life Dates"
"PowerLink -> Home > Support > Interoperability and Product Lifecycle Information > Documentum and Former Legato Product Information"
· XML interface & client for 3rd party interop matrix status and updates, eg to complement :-
NetApp http://now.netapp.com/matrix/ http://now.netapp.com/NOW/products/interoperability/
EMC
"PowerLink -> Home > Support > Interoperability and Product Lifecycle Information > Interoperability Matrices"
HDS http://www.hds.com/corporate/resources.html
· API / XML driven interface for licenses :-
a) To be able to remotely programmatically determine each & every licensed featured installed / not installed within the device
b) The be able to remotely programmatically determine the license metric and it’s usage for each & every licensed featured installed / not installed within the device
· API / XML driven interface for environmental impact
a) To be able to remotely programmatically determine the real time energy consumption by the asset (power etc)
b) The be able to remotely programmatically determine the high, low & average energy consumption by the asset (power etc) over a given period
IBM
· Would be very interesting to use the SVC to perform Raid5/6 over multiple arrays - to remove the array enclosure as an SPOF in a data-centre
Microsoft
· SQL Server formally & fully supporting NAS for it's database storage
· Formally & fully supporting 3rd party backup/recovery tools for it's database products
Zimbra
· Support of object storage (Castor, S3, Atmos etc)
· zBackup to operate multi-threaded and support for backup / restore of 10TB database messagestores
Caringo
· Full support for running Castor as a VMWare guest image
· Castor AWS S3 compatible API option
· Castor support for Zimbra, Sharepoint & Exchange+Enterprise Vault
· TCG SSC Oasis support
EMC
· Centera usability & readability of config reports
· SRDF & Mirrorview replication inbuilt cross compatibility
· Ability for SRDF sync replication with delayed updated at remote site
· Atmos support for Zimbra, Sharepoint & Exchange+Enterprise Vault
· Atmos support for an AWS S3 compatible API option
· Networker full support of Atmos as target
· TCG SSC Oasis support
· Greater qty of Ethernet ports on NAS devices
· EMC Powerlink to present at least the same deployment statistics and uptime information as NetApp NOW site
a) Enhancing the current information presented at :-
"PowerLink -> Home > Support > Interoperability and Product Lifecycle Information > Storage Target Revisions and Adoption Rates"
Release Date
Number of customer systems currently running this release
Number of customer sites currently running this release
Average run-time days per system
Qty of issues, by severity
c) To make the following additional information also available via an XML interface
Product major & minor (eg DMX & 4), Code Name, Code Rev, Release Date, # or % product running this release, # or % sites running this release, Average up-time, Target / Recommended Release
· To have an equivalent of EMC PowerLink or NetApp Now for self-support and information access
· API reporting of hot-spare qtys correctly on USP range
· API alert reporting for service processor utilisation on AMS range
· Ability for sync replication with delayed updated at remote site
· TCG SSC Oasis support
· Support for Thin-Provisioing in USP-1100
NetApp
· TCG SSC Oasis support
· Release the software only variant of OnTap as a commercial product (capacity limited if needs be)
· Greater qty of Ethernet ports on NAS devices
· Much larger aggregate & flexvol capacity sizes without having to upgrade to v8
· Support for OnTap v7.3 & v8 on older generation equipment and also for newer products (eg 2020 & 2050)
· Support for native OnTap 'in box' tiering of a file-system over multiple disk types (eg to support FAN)
Cisco
· 'terminal release' concept for SANos (check current name) - where support partners must converge upon a certainly release variant within X months (to aid interop)
Twidroid
· To have a "eMail tweet" (inc URL to tweet) that integrates with your google mail account
· To have a "Reply All" option against each tweet
· TwidroidPro posts to be 'sent from' TwidroidPro rather than just 'Twidroid'
· To correctly support Androids 'select & hold' ability to correct default spell-check suggestions
TweetDeck
· eMail tweet to include URL to tweet
· Option to save state of currently 'in memory' tweets upon shutdown, and reload upon restart (I'd pay for this feature alone)
· UK spellchecker
· Ability to set frequency of refresh of searches (in similar fashion to DMs, Mentions etc)
Snarfer
· An update to code?
· Better support for database store to prevent / fix corruption
This is Caringo speaking!
ReplyDeleteWe're very much in favor of open, transparent communications re our products and intentions, so we welcome the opportunity of being the first to answer our batch here and now ;-)
1. [Full support for running Castor as a VMWare guest image]
Management summary: we would love to do so and have already devoted a considerable amount of time and energy to the issue, but as far as we understand, it is technically impossible to build a production worthy reliable VM running CAStor, for the following reasons associated with proper keeping of real time - which is essential to a distributed storage cluster (freely quoting Jim Dutton, VP R&D Caringo):
"There is no problem with prospects kicking CAStor's tires or developing client applications by using VMs. Many have done so and experienced no issues.
BUT, VMs are not a supported platform for production clusters. The reason is, there is a well-known, systemic problem with VMs, called "clock skew" that affects embedded software like CAStor (but also game software, industrial control, and hard real-time systems).
We're not the only ones to notice this. Just google "virtual machine clock skew" or "VMWare clock skew" or "VirtualBox clock skew." VMWare has a 25 page whitepaper (http://www.vmware.com/pdf/vmware_timekeeping.pdf) discussing the problem. The second sentence summarizes the situation:
"VMware virtual machines use several techniques to minimize and conceal differences in timing behavior, but the differences can still sometimes cause timekeeping inaccuracies and other problems in software running in a virtual machine."
Some servers support so-called "tickless timekeeping" which ameliorates the clock skew problem. "Green" CPUs support clock stepping (which we take advantage of with our green features in the upcoming CAStor 3.5) and often worsen the problem. Clock skew gets worse as the number of VMs running on the same guest machine increases. So, to make the VM work right, you need to know a lot more about the underlying hardware, which is paradoxical.
We will not stop anyone from making CAStor VMs and distributing them for testing and evaluation. There are good, solid business reasons for not supporting them though, as they *will* fail in strange, unpredictable ways (indeed, they will not reliably make it through the CAStor test suite). "
This being said, if any source of expertise, preferably within VMware, can provide a practical approach to resolving this issue, we would be ecstatic, as many customers are requesting the capability. On the other hand, I would invite those interested to redo their cost comparisons for a production facility using CAStor VMs against physical boxes, including realistic estimates of all TCO elements. I haven't seen a solid business case for the former, ever, so far.
(continued below)
This comment has been removed by a blog administrator.
ReplyDelete(continued from previous post)
ReplyDelete2. [Castor AWS S3 compatible API option]
This is simpler. We do have an existing front-end product for CAStor, called SSP, built initially on request of a large telco. It implements a RESTful HTTP interface providing namespace partitioning, URI mapping, encryption, authentication and billing interfaces. Its API isn't S3 compatible at this time, but it is close enough for us to accept reasonably sized orders against specs for a next version that would be S3 plug compatible indeed. We do believe the approach makes sense as S3 may well become a de facto cloud storage standard.
3. [Castor support for Zimbra, Sharepoint & Exchange+Enterprise Vault]
CAStor support for Zimbra and Sharepoint is currently available from Caringo's partner Stealth Software ( http://bit.ly/BsxKT and http://bit.ly/6seyz ).
Caringo is aware of customer demand for an interface to Enterprise Vault. We are working towards having a solution in the market in the course of 2010.
4. [TCG SSC Oasis support]
Generally, the support of disk encryption schemes in CAStor should not be an issue, with one big caveat however: there is no user dimension whatsoever associated with the use of a CAStor cluster, so a proposed key management scheme cannot be user-dependent. As a consequence, I do not see another possibility than using a single key for all disks of a whole cluster, which is rather non-selective and kind of defeats the purpose. It would be really useful to get a description of the exploit(s) targeted with this proposal. In general, if selective encryption needs to be added to CAStor in the future, then it will have to be selective at the object level - with at least part of the metadata non-encrypted and containing the encryption selector or class which then leads to selecting the right decryption key for the object from a key management authority (after the requesting app provides the right credentials, that is).
-- Paul Carpentier, Founder and CTO, Caringo
Paul,
ReplyDeleteI urge you to check the following links on timekeeping best practices for VMware Linux VM's.
http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1006427
http://www.vmware.com/pdf/vmware_timekeeping.pdf
Longer term I cannot see ISV's like yourself having any other option BUT having to provide Virtual Appliance capability offerings. Large volumes of vendors have used Virtual Appliances as a means to gain market niche, so maybe this could be your opportunity.
Additionally you as an ISV have probably less to worry about when it comes to maintaining a HCL of running the core Caringo app and OS, Virtual Appliances have a standardised Hardware Base which mitigates this requirement.
I could go on but I wont...so hopefully this is of use to both yourself and Ian.
This comment has been removed by a blog administrator.
ReplyDelete