November 5, 2013 at 8:09 pm
Comments posted to this topic are about the item Project Hekaton
November 6, 2013 at 4:58 am
Microsoft is continuing to benefit from its open attitude towards research, development and release. Some functionality, products and/or features get dropped along the way but this has always been the case. I can remember that Windows 2000 Advanced Server was due to have a feature called IMDB (In Memory DataBase) which was in the betas but was dropped in the final release. There are always complaints when something is dropped as there is always someone who uses it but there is always the caveat that anything can be dropped at any time (or sometimes support cycles are defined for released products such as editions of Windows or SQL Server).
It will be interesting how much use that this feature will get in production systems.
As a developer I have little interest in this feature.
As a software engineer I have a need to understand what this feature offers.
As a software architect this could be a key feature that alters a proposed solution.
Gaz
-- Stop your grinnin' and drop your linen...they're everywhere!!!
November 6, 2013 at 6:02 am
Everything old is new again. Back in the day, I packed a machine with RAM, created a VDisk and assigned tempdb to the VDisk. The machine's performance jumped!
I tried putting some of the smaller lookup tables there. But that that made rebooting the server a pain. Now they are putting ALL the tables into RAM. How long does a reboot take?
November 6, 2013 at 8:50 am
I saw Dr DeWitt's keynote at Summit 13 and was impressed by how he explained Hekaton in a way that made it easy to understand for an earthling like me.
But then I tried to figure out how I would use it in my work, and can't find a good application for it. But we won't be using this type of feature anytime soon. Most of the OLTP systems we have are still in a mainframe environment. (Yea, we're a little behind the times :).
November 6, 2013 at 10:00 am
Recently I looked at a implementing this feature as I have an environment that is a perfect fit. However because of limitations against xml, lob data as well as FKs I don't seeing this being deployed. I love the idea though.
Russ
November 6, 2013 at 10:28 am
I've seen a number of people who compare the Hekaton features to the old school SQL Server 2000 and earlier OLTP methodology of readers block writers and writers block readers, but I've yet to find a good comparison of Hekaton to the more modern (2005 and up) OLTP methodology of using row versioning isolation levels for concurrency. Also given the restrictions on what things you can't do with them,
http://msdn.microsoft.com/en-us/library/dn246937(v=sql.120).aspx, and that you need memory tables to use compiled stored procedures, I think this is a real fringe feature, and few real world systems will be able to take advantage of them.
November 6, 2013 at 10:30 am
swwg69 (11/6/2013)
Everything old is new again. Back in the day, I packed a machine with RAM, created a VDisk and assigned tempdb to the VDisk. The machine's performance jumped!I tried putting some of the smaller lookup tables there. But that that made rebooting the server a pain. Now they are putting ALL the tables into RAM. How long does a reboot take?
I imagine most databases out there would benefit very little if any from this. But there are some cases that I think I could make some customers a little happier with the performance gain.
Jason...AKA CirqueDeSQLeil
_______________________________________________
I have given a name to my pain...MCM SQL Server, MVP
SQL RNNR
Posting Performance Based Questions - Gail Shaw[/url]
Learn Extended Events
November 6, 2013 at 6:48 pm
I could see this feature to be worth something to someone. But before I would use it I'd find someone to build me a SAN RAID unit based on Solid State Disks with a disk fail system.
I have been burned by one too many UPS/Generator/Switch failure in the past to trust data to plain old volatile RAM.
This is a Mongo DB Is Web Scale solution. As an eBay/Amazon/Google use yes. But they have the money to do triple redundancy failover solutions. Most of us aren't at that scale.
----------------
Jim P.
A little bit of this and a little byte of that can cause bloatware.
November 6, 2013 at 8:48 pm
As far as dealing with RAM and its inherent volatility Hekaton adheres to the durable principal of A.C.I.D., so once committed the data isn't lost. With the limitations and the unknown issues (corruption, impact on RAM, etc) I think I will have to wait until the next release. I do believe that it could have many uses and will start playing with it, just not in production.
Viewing 9 posts - 1 through 8 (of 8 total)
You must be logged in to reply to this topic. Login to reply