Joseph; One reason you haven't seen benchmark comparisions between the different relational databases (Oracle, Sybase, DBase (you pick the release), Informix), the vendors contractually inhibits their users from publishing the results. Why would they do that? Because they know the outcome. It ain't good for them. The Koreans did publish the results of one benchmark they did in one of the Last MUG Quarterlies. The results were astounding. Same hardware, same load, same task took the relationals 6 hours or more to complete. MUMPS took something less than one hour to complete.
Oracle used to be a true relational database and their performance really sucked. They hired Irene Chen from SAIC and the CHCS project. They learned some of why MUMPS is faster. Soon after Oracle became a relational database that mapped to a heirachical database internally. This gave them a big boost in performance and some improvement in scalability. They just didn't learn the rest of the lesson, how to make it scalable or learn how to build effective memory cacheing. The result is that MUMPS allows for some very important performance enhancement that other databases haven't learned yet. One such enhancement is that most read requests are satisfied in memory cache and these requests don't have to go out to disk. So only about 15% of the reads on a loaded system actually result in a physical read. This is a phenomenal increase in performance. A MUMPS system will speed up with more people on the system (to a max determined by the available memory and the CPU performance), but these limits are much higher numbers of users than Oracle or Sybase could support on the same hardware. The bottom line is that there have been attempts to replace MUMPS systems in the past and the CHCS project for the DoD has been no exception. They have been trying to bring up CHCS II to replace the CHCS I system which was patterned after DHCP, the direct predicessor to VistA. After 15 years and many millions of dollars, CHCS II has finally been withdrawn for the last time and CHCS I still runs the hospitals. If Oracle or Sybase, or Informix could do the job, they would be doing it. Where are they?? Want an idea of the complexity of the VistA model? Look up the Entity Relationship Diagrams. Then show one of the nearly 100 pdf files to your favorite Relational Database Guru and watch him blanch at the numbers of data elements and relationships represented there. On CHCS there were over 22,000 different data elements in the data dictionary. In Northern California, nearly 500,000 patient records are stored in less than 120 gigabytes of disk space. It would be interesting to see how much space the same information would take up in the relational model, then pack a lunch, cause it will take a good long time to traverse that data as a relational database. ----- Original Message ----- From: <[EMAIL PROTECTED]> To: <hardhats-members@lists.sourceforge.net> Sent: Saturday, June 11, 2005 5:22 PM Subject: RE: [Hardhats-members] == VistaWeb Missing Apps == > Yes, it will be interesting to see what can be done with Cache. We're > looking into the patient screening issue I mentioned previously, for > example. > > It's interesting, I never heard of this M/VistA being faster than SQL > relational until I started reading these messages. I'm keen to see if > that's true. I just got done writing an HL7 data access object to talk to > the HDR which is Oracle. (Don't ask why I'm talking HL7 to an Oracle > database, I'm not sure I know myself.) I will certainly be looking at the > speed. > > Yes, Oracle is expensive, perhaps prohibitively so. Is that the only game > in town, I wonder? Surely there's something between SQL Server and Oracle? > > And yes, speed is essential. It's what makes VistAWeb so popular with > providers, along with its simple interface. I'm not gonna be happy if the > HDR slows it down. > > -----Original Message----- > From: [EMAIL PROTECTED] > [mailto:[EMAIL PROTECTED] On Behalf Of Nancy > Anthracite > Sent: Saturday, June 11, 2005 3:00 PM > To: hardhats-members@lists.sourceforge.net > Subject: Re: [Hardhats-members] == VistaWeb Missing Apps == > > It seem to me that you do want is an M database that is reengineered for > VistA > because of its speed, easier mantainance and reliability and additionally, > the ability to do SQL queries on that database. Seems like Cache delivers > that and with the right additional software, GT.M can do that as well. > > The Intersystems folks knew what they were doing when they bought up all of > those flavors of M. The Epic folks know what they are doing as well. > > Relational databases are slower and that has long been recognized, and they > require a lot more work to maintain and design. Oracle is probably the > gorilla in that field and is very expensive and slow. Yea, maybe you get > some pretty reports from it, but not likely in real time. > > Tell all of the busy healh care worker why they have to wait when they > should > be taking care of patients so that you can get reports easily. The patients > > and those who care for them are not likely to be very understanding. > > On Saturday 11 June 2005 01:39 pm, Gillon, Joseph wrote: > > Listen, I'm not sure what this says, but I somehow feel the need to make > it > > clear I have nothing but respect for VistA and its developers. It was > > great, in fact, still is, but the old gray mare ain't what she used to be > > and is aging by the day. All systems that are worth a damn someday reach > > obsolescence. Which ain't to say VistA's going away any time soon. It's > > not hard to imagine a few years from now when several non-VA systems will > > be using it and the VA won't be. But the stuff the VA produces to replace > > it will be public domain too... > > > > -----Original Message----- > > From: [EMAIL PROTECTED] > > [mailto:[EMAIL PROTECTED] On Behalf Of Richard > > G. DAVIS > > Sent: Saturday, June 11, 2005 12:18 AM > > To: hardhats-members@lists.sourceforge.net > > Subject: Re: [Hardhats-members] == VistaWeb Missing Apps == > > > > I understand the "dichotomy" Gregory mentions, and I agree with the views > > he has expressed. > > > > However, I believe the fundamental issue is less a technical matter and > > more an natural 'tension' between concerns for operational > > effectiveness--delivery of quality health care, and the interest in > > administrative IT that arises from management. These two groups in most > > any enterprise you may choose to study are chronically in a state of > > 'conflict' due their different priorities and requirements for IT, and > > information management architecture. > > > > Where these competing forces arrive at a state of least tension, you > > usually find that the two groups have separated their IT, both software > and > > hardware into two largely independent systems. > > > > For those who have been in university/college settings, you may remember > > that those institutions tend to have "academic" computer centers and > > associated resources under the governance of some faculty body, and > > separate "administrative" computer systems used to run the university > > controlled the the 'CEO' of the institution. > > > > DHCP has also been chronically plagued with an interesting form of this > > problem, that waxes and wanes over time. > > > > At present within the VA we see the administrative folks throwing their > > territorial imperatives around and not giving due consideration to the > > operational requirements of the enterprise. These 'conflicts' are often > > deflected into debates about hardware issues, and software system issues, > > all of which are usually not at all relevant to the bed-rock forces at > work > > here. The person who says "MUMPS is a dead language, get rid of VistA" > > simply doesnąt understand the problem. It is easier to indict a > technology > > that can't defend itself, that to tackle the real core problem(s). > > > > Thus, the matter is not really about decentralized or centralized IT > > resources. It is all about meeting the needs of health care delivery > staff > > and at the same time satisfying the needs of management. > > > > I am frequently convinced that the best outcomes are achieved with > > organizational structures, and IT systems that permit two relatively > > independent systems to coexist, one optimized for health care delivery, > and > > one optimized for organizational administration. (The obvious need for > > effective interaction between these two systems is a subordinate or > > collateral matter to the core issue.) > > > > At the outset, DHCP was essentially free of intrusion of administrative > > demands. We built what seemed best for patient care. Only as Congress > has > > insisted that the DVA generate some revenue to supplement appropriated > > funds has the 'administrative' fist come down in a heavy handed way on top > > of the health care interests. (Yes, there are other forces in play here > as > > well.) This trend has promoted increased 'tension' between the operational > > and administrative sides of the house. > > > > The process needs to be focused on the organizational behavior issues > where > > the natural conflicts between the two sides of the house can be resolved > > appropriately. These processes are not technology based or driven. > > Instead, they are very much about organization and management. (It is in > > the spirit of the best of breed bureaucrats who prefer diffusion of > > responsibility and scape-goating of technology that we so often see the > cry > > to throw out the 'old' stuff.) > > > > Clearly, the VAH health care delivery process is mission critical and > > focused on the patient. The IT system need there is for a highly > > 'decentralized' framework that is centered on the patient/caregiver > > partnership. > > > > On the other hand, the DVA administration process must be able to > > effectively manage the resources of the DVA to maintain its two main lines > > of business--entitlements and sick veterans. That requires a highly > > 'centralized' IT framework that is centered on the problem of enterprise > > management. > > > > Whether these two groups are served by a single monolithic hardware system > > buried in a Colorado mountain, or by two slightly smaller computer systems > > situated at opposite ends of the Continental US, or by a massive number of > > desktop computers is really not too important once the balance between > > these two competing groups has been achieved. After that, all us > > 'techno-nerds' can go off and play with our cool software and hardware. > > :-) > > > > Regards, > > > > Richard. > > > > > From: Joseph Dal Molin <[EMAIL PROTECTED]> > > > Reply-To: hardhats-members@lists.sourceforge.net > > > Date: Fri, 10 Jun 2005 19:36:54 -0400 > > > To: hardhats-members@lists.sourceforge.net > > > Subject: Re: [Hardhats-members] == VistaWeb Missing Apps == > > > > > > The significant opportunity we have outside of the VA is to avoid > > > repeating the "data problem". How can that best be done? I think Greg's > > > option is on target....not sure if this would be called a "federated > > > model"... > > > > > > Joseph > > > > > >> There is, of course, another option, one that > > >> has been explored in a limited way, but not yet fully realized, that is > > >> to build loosely coupled systems that are, at once, decentralized an > > >> integrated (at the functional level). > > > > > > Gregory Woodhouse wrote: > > >> This is why I've never been very enamored of the original name of > VistA, > > >> Decentralized Hospital Computer System (DHCP). It wasn't decentralized > > >> at all, but rather each DHCP system was one of a number of independent, > > >> facility level systems. An ant colony is an example of a decentralized > > >> system: there is no centralized control, yet the colony is able to work > > >> together to accomplish a common task. > > >> > > >> Unfortunately, this false dichotomy has continued to plague VistA and > > >> the continued development of VistA. Far too often, centralized > solutions > > >> (either data or control, or both) are thought to be the only > alternative > > >> to completely independent application instances having no ability to > > >> work together effectively. There is, of course, another option, one > that > > >> has been explored in a limited way, but not yet fully realized, that is > > >> to build loosely coupled systems that are, at once, decentralized an > > >> integrated (at the functional level). > > >> > > >> === > > >> Gregory Woodhouse > > >> [EMAIL PROTECTED] > <mailto:[EMAIL PROTECTED]> > > >> > > >> "The whole of science is nothing more than a refinement > > >> > > >> of everyday thinking." -- Albert Einstein > > >> > > >> On Jun 10, 2005, at 2:30 PM, Sowinski, Richard J. wrote: > > >>> oe, your note title example can be explained in this way. For years > > >>> Vista was used in a standalone manner. Each site, was essentially an > > >>> island, that could create it's own note titles, lab test names, etc. > > >>> > > >>> Then, software like RDV's and Vistaweb came along, and exposed > > >>> (everyday) the fact that people used different terms at different > > >>> sites, for essentially the same data. > > >>> > > >>> This is not a software problem, it's a data problem. It's an artifact > > >>> of making what was essentially a standalone system, into a networked > > >>> system. > > >>> > > >>> It's cure, is standardization work. Unfortunately that work can be > > >>> tedious and not as glorious as other work, which is why it has been a > > >>> slow starter. But sooner or later that work will get done, and Vista > > >>> will be around for awhile. > > > > > > ------------------------------------------------------- > > > This SF.Net email is sponsored by: NEC IT Guy Games. How far can you > > > > shotput > > > > > a projector? How fast can you ride your desk chair down the office luge > > > > track? > > > > > If you want to score the big prize, get to know the little guy. > > > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20 > > > _______________________________________________ > > > Hardhats-members mailing list > > > Hardhats-members@lists.sourceforge.net > > > https://lists.sourceforge.net/lists/listinfo/hardhats-members > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: NEC IT Guy Games. How far can you > > shotput > > a projector? How fast can you ride your desk chair down the office luge > > track? > > If you want to score the big prize, get to know the little guy. > > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r > > _______________________________________________ > > Hardhats-members mailing list > > Hardhats-members@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/hardhats-members > > > > > > ------------------------------------------------------- > > This SF.Net email is sponsored by: NEC IT Guy Games. How far can you > > shotput a projector? How fast can you ride your desk chair down the office > > luge track? If you want to score the big prize, get to know the little > guy. > > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r > > _______________________________________________ > > Hardhats-members mailing list > > Hardhats-members@lists.sourceforge.net > > https://lists.sourceforge.net/lists/listinfo/hardhats-members > > -- > Nancy Anthracite > > > ------------------------------------------------------- > This SF.Net email is sponsored by: NEC IT Guy Games. How far can you > shotput > a projector? How fast can you ride your desk chair down the office luge > track? > If you want to score the big prize, get to know the little guy. > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r > _______________________________________________ > Hardhats-members mailing list > Hardhats-members@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/hardhats-members > > > ------------------------------------------------------- > This SF.Net email is sponsored by: NEC IT Guy Games. How far can you shotput > a projector? How fast can you ride your desk chair down the office luge track? > If you want to score the big prize, get to know the little guy. > Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20 > _______________________________________________ > Hardhats-members mailing list > Hardhats-members@lists.sourceforge.net > https://lists.sourceforge.net/lists/listinfo/hardhats-members > > ------------------------------------------------------- This SF.Net email is sponsored by: NEC IT Guy Games. How far can you shotput a projector? How fast can you ride your desk chair down the office luge track? If you want to score the big prize, get to know the little guy. Play to win an NEC 61" plasma display: http://www.necitguy.com/?r=20 _______________________________________________ Hardhats-members mailing list Hardhats-members@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/hardhats-members
