Vietnam International Commercial Joint Stock Bank

The Client

Vietnam International Commercial Joint Stock Bank VIB) is one of the leading banks in Vietnam, with total assets of more than VND 100,000 billion. Currently, the bank has 4,300 people working at nearly 160 branches and transaction offices in 27 key provinces/cities across the country.

The Partner

1-V System JSC (1VS) is one of the most successful and fast growing 1C certified partner. 1VS offers a wide range of business automation solutions based on 1C:Enterprise platform. There are over 3000 companies in Vietnam using 1VS solutions in their everyday business operation.

The Goal

In 2013 bank management planned and performed a company’s business-processes reorganization, focusing especially on IT infrastructure. Advanced enterprise content management system deployment was one of the priorities. Studies of the options presented on the marked, conducted by VIB IT management, resulted in choosing 1C:DocFlow as by far the most cost-efficient solution meeting all the requirements of the bank. 1VS Company was chosen as a main contractor for the project being the most experienced and professional 1C partner in Vietnam.

The System

Now there are over 100 concurrent users working with 1C:DocFlow system in the head office of the bank. The system provides users with rapid access to consolidated bank documents, automates and speeds-up processes of approval, engagement and confirmation of documents, secures the tasks fulfillment control and more.

Down the road the bank plans to extend the functionality of the system and deploy it across the branches.

The Tasks

Below is the list of the tasks performed with the participation of 1C experts as a part of the Extended TechSupport project.

Choose server-side hardware

Before the system could be deployed we needed to choose the hardware characteristics for its server-side. In order to perform the task the official 1C recommendation were used. Due to time pressure there was no chance to build full-scale multi-user test system but there was a comparable system, being live in another company, which could be used as a Model.

In order to make a prediction about the bank system hardware workload we needed to get the Model System hardware utilization data. Two different sets of data – 1C:Enterprise server utilization and DBMS server utilization – were collected using Windows Performance Monitor.

The data was manually transferred into the Excel worksheet and the following requirements for the bank system hardware were calculated.

Choosing CPU

1C:Enterprise server

Calculated number of nominal CPU cores was 1.14, which basically meant we needed at least 2 CPU cores for 1C server of the target system. Bank had available IBM PureFlex with Intel Xeon E5-2690 (8 cores, 2.9 HGz) CPUs environment, whereas the Model system had Intel Xeon E5640, 2.60 GHz CPUs. According to the following independent benchmark, E5-2690 outperforms E5640 2.5 times, which complied with the requirement, that the Target CPU had to have the same or greater one-thread performance that the Model CPU.

Therefore the following recommendation was given: allocate 2 virtual CPU cores on IBM PureFlex for 1C:Enterprise server.

DBMS server

Calculated number of nominal CPU cores was 0.66 so one core was supposed to suffice for the Target system DBMS server needs. Model DBMS server had Intel Xeon E5-2630, 2.3 HGz CPUs, which was 36% slower than E5-2690 according to the same benchmark. Just to be on the safe side the recommendation was to allocate 2 virtual CPU cores on IBM PureFlex for the Target DBMS server.

Choosing disk array

Calculated relative disk array performance was about 16% of the Model disk array performance for both 1C and DBMS servers. It meant that the Target system needed about one tenth of the Model system disk array performance.

Model disk array model was IBM DC3500 – entry-level disk array targeted on small business systems. The Client had IBM Storwize V7000 (midrange array) available for the task. Therefore we could use the Client’s array for both servers, being sure its performance was more than enough.

Choosing RAM size

1C:Enterprise server

Overall amount of memory allocated by all 1C:Enterprice processes (ragent, rmngr and rphost) in the Model system was 6,887 Mb. It gave us calculated amount of 3,150 Mb of memory for 1C:Enterprise server of the Target system. Therefore the recommendation was to allocate 4 Gb of RAM.

DBMS server

First of all we needed to assure that there was no memory pressure on DBMS server. Collected cache hit ratio was close to 100% which means DBMS had enough memory. The overall amount of memory allocated by sqlservr process was 66,502 Mb which gave us calculated amount of 12,688 Mb for DBMS server of the Target system. Therefore the recommendation was to allocate 14 Gb of RAM for DBMS server.

Fix major performance problem

During the user acceptance testing (which was basically a beta testing in production environment) everyone involved in the process were registering persistent and permanent performance problems. Literally every user action (like opening a form or refreshing a list) used to take a few seconds while the more massive operations (like posting a heavy document) could take up to 30 seconds.

Basing on the symptoms we assumed there is one factor impacting on the overall system performance. First of all we needed to determine what layer of the system was the source of the issue. There were a number of suspects: applied solution (1C:DocFlow), 1C:Enterprise platform, middleware, operating system, hardware, network and so on. To narrow the search we started excluding one layer after another.

First of all we excluded the applied solution and 1C:Enterprise. We ran and tested the exact copy of Client’s solution in two different environments: on Partner’s hardware and on 1C:ETS test servers. Both tests gave the same result: performance was fine and solid.

After that we checked the Client’s hardware utilization, which happened to be extremely low on both physical and logical levels.

This left us, basically, only two possibilities: operating system (MS Windows) and middleware (VMWare). To ensure that the problem source was here we performed a series of tests on Clients environment using various benchmarks including MS SQLIO tool. Tests showed that memory and disk response times were extremely poor comparing with any other computer we tested. Further investigation pointed at VMWare, which, apparently, was incorrectly tuned.

At this point Client made a decision to workaround the problem and move on with the system deployment. The solution was transferred to another environment and performance met current Client’s requirements.

Fix incorrectly displayed Vietnamese font in Internet Explorer

Testing the solution users noticed a font readability issue. In some text spaces between letters seemed missed (so-called «caked type») and the words were hard to read.

First of all we try to reproduce the issue on few different environments and found out that if was reproducible only when you use 1C WEB-client under Internet Explorer 8. It worked OK under any other browsers (including IE 6, 7, 9, 10).

Further investigation showed that there was a problem in Vietnamese MS Sans Serif font coming along with IE. If was recommended to client to upgrade the IE version, which solved the issue.

Other

There were a few other minor problems solved within the framework of the 1C:ETS project.

The Feedback

«Vietnam International Bank (VIB) has successfully achieved the goals of the first stage of enterprise content management system implementation project»

Please refer to the full text of the Client’s feedback.

Be the first to know tips & tricks on business application development!

A confirmation e-mail has been sent to the e-mail address you provided .

Click the link in the e-mail to confirm and activate the subscription.