[Fwd: FYI:When will it all stop]

Rick Rose rrose@dsw.net
Mon, 23 Mar 1998 10:53:33 -0800


This is a multi-part message in MIME format.
--------------3FF5D53A4C04E609EEA7077A
Content-Type: multipart/alternative; boundary="------------9F09300835D70EFF722184CF"


--------------9F09300835D70EFF722184CF
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit

For all SSA Managers,
From: Rick Rose (rrose@dsw.net) Data Systems West:
FYI:  After reading the forwared letter this may interest you:

I've been working for a third party systems integrator and am tired of
hearing Microsoft
put downs of some of SUN's best products to date namely the A5000 &
A7000 Storage arrays. I've tried to keep myself impartial because our
company (and I) support both
SUN and MicroSoft platforms. We typically install about 10 servers a
month.
These new storrage arrays kick ass, our Oracle db
programmer with over 8 years experience is seeing right now (not on
paper) I/O
through-put that he's very impressed with. Its true the 100, 200 series
throughput was'nt
that great but now with SUN's new offerings Microsoft should put up or
shutup because
for real big sites MS servers still can't compete with UNIX power.
True coding multithreading MS isn't  there yet!  SUN's CPU through-
put to CPU curve is very linear upto 64 processors were PC servers still
have CPU fall off known as blue-droop around 4 processors. Even with NT
5.0 which is suppost
to support 8 processors there CPU throughput is not a very linear curve.
Because SUN
has optimized there platforms (not only the CPU is optimized for high
efficency but all
the supporting CPU circuitry) for there OS they are able to achive over
87% linearity
(and close to 96% in the SUN StarFire) in regards to units of work. MS
will get there
but there not there yet currently Windows only has a 60% CPU to "Units
Of Work"
rating.

R.R.




--------------9F09300835D70EFF722184CF
Content-Type: text/html; charset=us-ascii
Content-Transfer-Encoding: 7bit

For all SSA Managers,

From: Rick Rose (rrose@dsw.net) Data Systems West:
FYI:  After reading the forwared letter this may interest you:

I've been working for a third party systems integrator and am tired of hearing Microsoft
put downs of some of SUN's best products to date namely the A5000 & A7000 Storage arrays. I've tried to keep myself impartial because our company (and I) support both
SUN and MicroSoft platforms. We typically install about 10 servers a month.
These new storrage arrays kick ass, our Oracle db
programmer with over 8 years experience is seeing right now (not on paper) I/O
through-put that he's very impressed with. Its true the 100, 200 series throughput was'nt
that great but now with SUN's new offerings Microsoft should put up or shutup because
for real big sites MS servers still can't compete with UNIX power.
True coding multithreading MS isn't  there yet!  SUN's CPU through-
put to CPU curve is very linear upto 64 processors were PC servers still have CPU fall off known as blue-droop around 4 processors. Even with NT 5.0 which is suppost
to support 8 processors there CPU throughput is not a very linear curve. Because SUN
has optimized there platforms (not only the CPU is optimized for high efficency but all
the supporting CPU circuitry) for there OS they are able to achive over 87% linearity
(and close to 96% in the SUN StarFire) in regards to units of work. MS will get there
but there not there yet currently Windows only has a 60% CPU to "Units Of Work"
rating.

R.R.
 
 
  --------------9F09300835D70EFF722184CF-- --------------3FF5D53A4C04E609EEA7077A Content-Type: message/rfc822 Content-Transfer-Encoding: 7bit Content-Disposition: inline Return-Path: Received: from dswint (fw.dsw.net [206.214.66.1]) by nexxus.dsw.net (Netscape Messaging Server 3.01) with SMTP id AAA3532; Fri, 20 Mar 1998 18:01:46 -0800 Received: from mercury.Sun.COM ([192.9.25.1]) by dswint. via smtpd (for nexxus.dsw.net [206.214.66.38]) with SMTP; 21 Mar 1998 01:38:56 UT Received: from West.Sun.COM ([129.153.100.31]) by mercury.Sun.COM (SMI-8.6/mail.byaddr) with SMTP id RAA23377; Fri, 20 Mar 1998 17:58:47 -0800 Received: from valley.West.Sun.COM by West.Sun.COM (SMI-8.6/SMI-5.3) id RAA16079; Fri, 20 Mar 1998 17:58:37 -0800 Received: from bruin.West.Sun.COM by valley.West.Sun.COM (SMI-8.6/SMI-SVR4) id RAA21886; Fri, 20 Mar 1998 17:58:29 -0800 Received: from valley.West.Sun.COM by bruin.West.Sun.COM (SMI-8.6/SMI-SVR4) id RAA16271; Fri, 20 Mar 1998 17:58:11 -0800 Received: from bruin by valley.West.Sun.COM (SMI-8.6/SMI-SVR4) id RAA21872; Fri, 20 Mar 1998 17:58:05 -0800 Message-Id: <199803210158.RAA21872@valley.West.Sun.COM> Date: Fri, 20 Mar 1998 17:58:09 -0800 (PST) From: "Stanley Chan [SFValley Reseller SE]" Reply-To: "Stanley Chan [SFValley Reseller SE]" Subject: FYI: Sun Storage Products's Business Unit responds to EMC FUD To: laresellers@bruin.West.Sun.COM Cc: reseller.west@central.Sun.COM MIME-Version: 1.0 Content-Type: MULTIPART/mixed; BOUNDARY=Dule_of_Doves_503_000 X-Mailer: dtmail 1.2.1 CDE Version 1.2.1 SunOS 5.6 sun4m sparc --Dule_of_Doves_503_000 Content-Type: TEXT/plain; charset=us-ascii Content-MD5: 0BTQzeK4FR1CmpDUL5WryQ== Sun has definitely hindered EMC's success with our A5000 and A7000. They've decided to comeback at us with some technical FUD against our solutions. Below are responses to "re-counter" their claims. -STAN ------------- Begin Forwarded Message ------------- Date: Fri, 20 Mar 1998 16:46:29 -0800 (PST) From: Steve Rogers All, EMC is taking some big losses to Sun Storage Edge products. They have responded by attempting to attack both the A5000 and the A7000 products at a technical level. The two attachments contain our detailed rebuttles to their claims, point-for-point... I'd like to acknowledge the key contributors in the BU who worked deligently to bring these responses to you so quickly: A5000 - Dave Fisk, Lynn Whitfield, Michael Nalls. A7000 - Art Licht, Tim Reid, Adam Hawley. Please Forward these on to whomever you think could benefit from the information. BTW - we will also be posting these to our website. Go get-em! Regards, Steve Rogers _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ /_____/\ Steve Rogers /_____\\ \ Manager, Technical Marketing -Storage Products. /_____\ \\ / Sun Microsystems Computer Company /_____/ \/ / / 2550 Garcia Avenue MS: MPK12-203 /_____/ / \//\ Mountain View, California 94043-1100 \_____\//\ / / Phone: (415)786-7828 | FAX (415)786-7816 \_____/ / /\ / Pager: 1-800-759-7243 Pin# 1240531 \_____/ \\ \ Email: steve.rogers@eng.sun.COM \_____\ \\ ===================================== \_____\/ _/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/_/ ------------- End Forwarded Message ------------- ________________________________________________________________ Stanley Chan Sun Microsystems Computer Company Systems Engineer 15821 Ventura Blvd., Suite 270 West Reseller Region Encino, CA 91436 ph 818-379-3420 fax 818-905-5841 --Dule_of_Doves_503_000 Content-Type: TEXT/plain; name=A7000FUD; charset=us-ascii Content-Description: A7000FUD Content-MD5: yMTq2aaBGqbJLhEAPPfC4g== Written by Art Licht - Technical Marketing Tim Reid - Product Marketing Adam Hawley - Product Marketing Date: 3/20/98 Storage Products Business Unit's response to the EMC's claims about the Sun StorEdge, A7000, Intelligent Storage Array. EMC's statements are marked with the '#' (pound sign.) Our response follows. # #SUN ANNOUNCES INTELLIGENT STORAGE SERVER A7000 (FORMERLY THE ENCORE SP40) # #BRIEF DESCRIPTION # #Nearly one year to the day after announcing its intent to enter the # "enterprise storage" market, Sun Microsystems has announced its first # product offering. # # In November 1997, Sun acquired Encore Computer's storage business unit - # becoming the third company (Amdahl and IBM being the others) in less # than three years to attempt to make the Encore SP40 a viable storage # solution The technology inherent within the SP40 (now the A7000) was, and still is, considered by a number of independent analysts amongst the best available in the industry. Well ahead of other competitive offerings. The Encore storage technology is unique in the industry with its powerful dual UNIX SMP storage server nodes providing high value-add features such as datashare. To take this technology to its logical extension, Encore wanted to partner with an experienced systems solution provider so it looked at the rest and settled on the best: SUN - #1 in UNIX systems, #1 in UNIX systems storage. Sun, with its proven experience at creating the industry's most reliable UNIX operating system, and the most scalable performance servers available today, combined with a well-know background in networking, is better positioned than competitors to deliver on next generation storage solutions beginning today with the A7000. Storage companies without control of operating systems, compute platforms, and networking technology will increasingly fall behind in delivering customers flexible solutions optimized to their application's needs. #proclaims extensive plans for this product, the only changes to date are the #name (Sun Intelligent Storage Server A7000) and the use of 9GB drives, neither #of which is particularly innovative. No significant enhancements are expected #for the remaining lifecycle of the A7000, with Sun focus on transitioning the #SP40 from a proprietary UNIX/RISC product to a Solaris and SPARC based #replacement. Sun has already disclosed a number of committed enhancements for the A7000. Notable amongst these are; Fibre Channel Arbitrated Loop host interfaces (a long standing strategic direction for Sun); ATM-based wide area remote dual copy capability; and DataShare enhancements are some of the highlights to come. The A7000 will enjoy a full and rich life including enhancements just as any other sun product would. #Some key points to remember when competing with Sun's Intelligent Storage #Server A7000: # #- Sun is entering into a new market - the heterogeneous enterprise. Sun has #neither the sales, engineering or support structure for such a market. Sun is an established systems company that has been selling, installing, and supporting enterprise systems for many years. We offer, install and support business solutions from the desktop to the high end data center. These system are used in mission-critical application environments for many of today's most demanding enterprises. Sun is currently supporting high-end data center environments with mission- critical systems such as its Enterprise 10000s in 36 countries throughout the world. # #- Sun has no experience supporting the mission-critical storage market. Sun's #customers have traditionally accepted downtime for maintenance and upgrades. #In contrast, the mainframe and mid-range open systems markets require #mission-critical data storage solutions with online upgrades and maintenance. Sun not only has this experience, but with the recent acquisition of Encore's technologies, Sun has undertaken a program to further its capabilities in service delivery. With the A7000, Sun is now offering advanced phone-home and remote diagnostic capabilities. In the event that the A7000 detects any abnormal condition within the 1000+ monitored points within the system, the A7000 will automatically "phone-home", and a Sun service specialist will be notified. Within 15 minutes, a Sun service specialist will be remotely connected to the A7000 to begin the diagnosis of the detected abnormality. # #- Sun storage products offer no common enterprise storage architecture or #strategy. Between the SPARC Storage Array Family, RSM 214/218 Arrays, #RSM Array 2000, ENA 3000 and now the A7000, Sun offers five disparate #products. Each product has different integration requirements, #maintenance and configuration tools and none support the same features #and function sets. Last, but not least, none of the above products support #Sun's entire server family, each is designated to a specific segment #incapable of supporting Sun's entire enterprise. Sun does not believe that any single storage solution can meet the needs of the wide range of demands that we see from our customers. In responding to our customer's needs and requests, Sun has developed a wide range of storage solutions that are targeted at being "best of breed" for today's dynamically changing and diverse enterprise APPLICATION environment. EMC MUST portray a "one size fits all" product line as an advantage, as if the requirements for each of these areas are the same. If you only have one product, then this must be the best way to respond to your customers requirements. However, if you have more than one product in your kit, why not deliver a solution that is tailored to the customer's requirements rather than compromise them? This is a classic example of the value of a systems company rather than a niche player as a supplier. As a systems company, Sun has investments in all areas of the enterprise and has the resources available to respond to these diverse needs. Additionally, Sun has gone one step further. With its announced Intelligent Storage Network vision and strategy, Sun is committing to the delivery of highly adaptable storage environments that can leverage existing investments. With this strategy, our customers will be able to enjoy all the benefits of simplified management, consolidated storage, and direct information sharing without the need to compromise on their specific application requirements. # #- The Sun A7000 will primarily be sold and supported by former Encore #representatives (East US) and resellers (West US). The international product #announcement, availability dates, sales and support details, are still #undefined. The availability announcement could be limited to the US and #Canada only as Sun has no international capability to support this product. As announced, Sun is providing worldwide sales and support for the former Encore storage products throughout the world. # #PRODUCT ANNOUNCEMENT SPECIFICATIONS # #Other than the introduction of 9GB drives and addition of the Sun logo, the #A7000 remains identical to the original Encore SP40. The A7000 is a dual VME #bus server accommodating 28 total system bus slots for host channel, disk, #processor and memory cards. The A7000 actually represents a new approach to the development and deployment of storage subsystems. Existing, old-style storage architects continue to have problems understanding the advanced architectural concepts of this breakthrough design. Simply stated, the A7000 in its current form consists of two UNIX-based SMP computers acting as storage subsystems through the use of innovative software. This software allows any connecting host to view the systems as if they were compatible storage environments. Mainframe environments see a 3990 compatible controller with any mix of 3390/3380 device images, while other hosts see multiple SCSI devices of the size and type required by the hosts operating environment. # #Drawbacks of the A7000 # #The A7000 is not a fault-tolerant or high-availability storage solution; # it's a server with 28 VME bus slots for disk and host adapters. The actual # foundation for the A7000 is quite simple and has a wide range of flaws # (many are outlined below). As stated, the A7000 is actually two separate computers that work together to provide high availability. In the event of either computer failing, from software to the backplane itself, the surviving computer (remember you always get two) continues to provide data availability to the attached hosts. Each of these computers is within a separate power region, with common components (disks, fans, etc.) each having n+1 power supplies and other high availability features. Competing products offered by EMC do not offer redundant cache, creating a single point of failure with the risk of data loss, making claims of high-availability questionable. Each computer comprising the A7000 actually contains 3 buses, all operating in parallel. One such bus is the I/O bus, derived from an industry standard. There are additional buses for processor to memory access, and for memory-to-memory interconnection. These are in addition to the independent host and disk channel interfaces, which are all capable of concurrent operation. The combined throughput of these 6 buses (each A7000 contains two computers, each with 3 buses) is from 500MB/sec to 920MB/sec, depending on the configuration. As cache is added, so ARE memory channel buses. This promotes scalability. Sun makes extensive use of these multiple buses, operating in parallel, to avoid many of the "hot spot" issues seen within traditional storage architectures. # #- The A7000 is based on a four-way RISC multiprocessor (Encore server) # running a proprietary flavor of UNIX. This Encore flavor of UNIX will not # be easily upgraded or enhanced by Sun. Sun's long-term goal is to # transition Encore's functionality to Solaris and SPARC processors, not to # continue to invest in an old RISC platform with a proprietary operating # system. The advanced software environment of the A7000 is highly portable and has no dependencies on any specific "flavor" of UNIX. There are no hard-coded processor to software linkages like those seen in traditional microcode based storage architectures. The entire environment is written on top of an operating environment that allows for easy portability of any added functionality and/or enhancements. It is well known that UNIX is a supported environment on a large number of hardware platforms. The question you should be asking...How portable is a proprietary, fixed-function storage architecture that is dependent on the processor selected? The use of C based applications in a UNIX environment has clear portability and migration advantages over the use of proprietary, fixed-function microcode environment that is hardware dependent. The A7000 is built on 4way SMP Motorola RISC processors. Although EMC would like you to believe that this is an "old RISC platform", the Symmetrix is also built with Motorola processors. # #- DataShare is a software program that runs on the A7000 server. The program #offers minimal flat file transfer capability from MVS to UNIX only. No #significant upgrades or enhancements will be made to DataShare until migration #to a Solaris/SPARC-based server engine is complete (12-18 months). No DB #sharing functionality exists. DataShare is a unique, advanced function offered by Sun that allows our customer to directly access a single copy of data without the need for any transfers. This is often mistakenly compared to EMC's SMTF, which is nothing more than a proprietary, high specialized transfer utility using host channel interconnects. With DataShare, there are no proprietary requirements. You read and write to a single disk image using your existing operating system interfaces. Today, the most commonly understood file format is sequential, fixed record. Since DataShare allows our customers to directly read disk information, this is one format that is immediately and readily understood for use in existing application environments. Added function will be implemented based on this standard, highly portable capability. The intelligence required for this implementation is contained within the A7000. Unlike other offerings which are actually "middleware" offerings for host environments that require individual licenses for each platform supported. In addition, these often have operating system dependencies and/or perform non-standard I/O operations across the host channels without the host being aware of these actions. # #- Each customer must develop tools and associated software to use the # DataShare feature of the A7000. Without such integrated software, the # customer's applications will not function coherently with the A7000. As # each application environment has a different set of expectations, customer # satisfaction has not traditionally been a strong point for this solution. Our actual experience indicates just the opposite. Since we allow existing applications to directly access a single disk image, without the need for specialized operating system code on each host, our customers can take immediate advantage of our DataShare facility. However, we are always willing to assist our customers in implementing innovative, new approaches to solving their most demanding business challenges that, until the existence of DataShare, were not even possible to address. # #- RAID functionality is enabled by single processor commodity VME host bus #adapters. Any RAID recovery or rebuild process is managed by the host # adapter, causing significant performance degradation to production # applications. EMC does not do RAID-5, they do RAID-S. Its been compared to a Raid 4+ implementation. Our RAID-5 is implemented via RAID-5 hardware to handle the XOR logic. If this hardware should fail the raid set would failover to the surviving node and processing continue. Raid-5 protection remains intact, since no disk failure occurred. Raid-5 reconstruction automatically occurs, if raid sparing has been configured, upon a disk failure occurring. The actual rebuild rate is controllable and throttlable, in 10% increments. # #- Multiplatform support limited to MVS, HP-UX, IBM, AIX, Sun Solaris and # Windows NT. No mention of support for HP MC/ServiceGuard, IBM HACMP or # NT clusters. Sun has announced support for a number of systems, with more under review. # #- Connectivity is limited to the three-year old FW SCSI host adapters for the #proprietary VME system bus. No support for Ultra-SCSI or FC-AL connectivity #exists. Sun is committed to providing Ultra-SCSI and FC-AL connectivity options for the A7000. # #- No communication between disparate operating environments. Correct, we do not need specialized software for communicating between environments. The A7000 acts as a compatible storage subsystem for each attached host. Any need for communications between host applications can be done through DataShare, which allows simple read/writes to disks to be used as a mechanism for cross application communications. # At this point, the writer begins to repeat themselves...perhaps to make the list appear longer? The following points are discussed within the information above. #- No point-in-time copy software (TimeFinder) # #- Lack of data-migration capability # #- No form of internal workload balancing to midigate "hot-spots" The A7000 with its internal data striping is not as prone to hots stops as the competition. # #- No proven track record of production usage # #- No upgrade path from existing Sun Subsystems # #- Single points of failure in each disk pack include the power distribution # unit and the single processor RAID controller (commodity host adapter) Not true. The A7000 has 2 separate PDU and power regions. A failure of a raid accelerator does not prevent access to the data. # #An example of a maximum A7000 server configuration of 2.9TB requires the #following configuration: # #A total of 13 disk adapters (13 slots) # #Dual four-way CPU boards (two slots) # #Four 1GB memory cards (four slots) # #Six PW SCSI host channel adapters (six slots) # #Three ESCON channel cards (three slots) # #As the system has a maximum internal bandwidth of 50 MB/sec (vme bus speed), #no capability exists to support sufficient performance across more than 4-5 #host channel connections. Clearly, EMC does not understand the system. They have incorrectly assumed it to be a single monolithic implementation of 28 slots. In actuality it is a dual redundant configuration consisting of dual chassis. # #A7000 Product Details # #- Multihost connectivity supports only MVS, HP-UX, IBM AIX, Sun Solaris and #Windows NT platforms. The above plus various flavors of VM and VSE. # #- A7000 server is based on a dual-VME system bus (80 MB/sec internal bandwidth #of the VME bus is the servers bottleneck). As stated previously stated there are multiple buses that operate in parallel. There are 2 L-Buses, each rated at 100MB/sec, 2 IO Buses, each rated at 80MB/sec and multiple memory channel buses, each rated at 140MB/sec. Each 1GB of cache adds an additional Memory Channel bus. Therefore, an A7000 with 1GB of cache has available 500MB/sec of bandwith, while an A7000 with 4GB of cache has 920MB/sec of available bandwith. # #- Maximum of 8GB system memory (4GB per four-way CPU). Maximum memory #configuration will require decreasing the disk or host channel configurations. #Available system memory also will be impacted by mirrored writes. The A7000 supports up to 4GB of mirrored cache. Sun believes that non mirrored cache is a serious single point of failure. Lose the cache, lose the data. # #- Sun positions the server memory as "4GB mirrored non-volatile cache". # #- Remote subsystem mirroring for campus environments only. # #- Although 2.9TB may be configured on a single A7000 server, it represents a #major bottleneck as opposed to a storage subsystem architected from the #ground up to allow maximum bandwidth and performance scalability. As the #A7000 has a #fixed limited front-end and back-end channel bandwidth, #performance will be inhibited as additional capacity is added. As disks and cache are added to the A7000, so are additional SCSI buses and Memory Channel buses. Each additional 6-pack adds a SCSI bus. Each additional 1GB of cache adds a Memory Channel bus. # #- Basic phone home monitoring/diagnostic support (limited remote capability). Not true. Basically, anything that one can do sitting at the console, SunService can do remotely, with the exception of replacing hardware. The machine can be monitored, failures diagnosed, software upgraded and reconfigurations can all occur remotely. # #- DataShare software capability limited to flat file extract from MVS # mainframe to UNIX only. This server-based software program offers # minimal support and configuration tools. # #- Capacity range from 220GB-2.9TB. Scales via 54GB drive packs (6x9GB # drives). Total of 54 drive packs enables maximum capacity of 2.9TB # (unprotected). # #- RAID levels 0, 1 and 5 supported (host adapter based), no disk # adapter-based read/write cache exists. Read-ahead logic occurs within the general cache pool. Once sequential mode is entered, data begins to be prestaged into the A7000's cache. # #- List price for maximum configuration approximately $2/MB (expect 50% #discount). # #BOTTOM LINE # #EMC's enterprise hardware and software solutions, including Symmetrix, SRDF, #TimeFinder, InfoMover, DataReach, ESP, EDM and PowerPath represent the #industry standard for enterprise storage. While some vendors may claim #support for segments of the enterprise and its wide range of diverse #applications, their ability to execute may be aggressively challenged. #The historical lack of acceptance of the A7000 demonstrates its limited #effectiveness in competing with EMC Enterprise Storage. # EMC has done a good job marketing a vintage 1990 architecture by marketing specialized software and middleware and at the same time locking customers into a proprietary solution. --Dule_of_Doves_503_000 Content-Type: TEXT/plain; name=A5000FUD; charset=us-ascii Content-Description: A5000FUD Content-MD5: 1ofrLEdHAd7S6Z72BqWAUw== Written by Dave Fisk - Technical Marketing. Lynn Whitfield - Product Engineering. Michael Nalls - Product Marketing. Date: 3/20/98 Storage Products Business Unit's response to the EMC's claims about the StorEdge, A5000, FC-AL storage array. EMC's statements are marked with the '#' (pound sign.) Our response follows. The over all message was to the effect that the A5000 is "5 years behind in technology". This is far from true. Overall, The A5000 represents the most advanced, efficient, high performance array in the industry. It delivers 95+ MB/sec on a single FC-AL channel, with a single 14 disk enclosure. What does Symmetrix deliver on one channel, one channel director? A 40 MB/sec Ultra channel on a Symmetrix has been o bserved to deliver only 14 MB/sec with 100% cache hits. Presumably a 100 MB/sec FC-AL on a single channel director will also deliver around 14 MB/sec. Let's talk about old design... The evolution of technology is in general the pursuit of efficiency. And, in terms of efficiency Symmetrix can not compare to A5000. For a light work load of small writes, Symmetrix is a good choice, if you don't mind paying the premium. For high bandwidth sequential read or write, or sustained random read or write, Symmetrix is an inefficient use of server I/O slot and channel bandwidth. Key points in rebuttal to EMCs' assertions follow: # #IB boards ( interface boards in the Photon) are not hot pluggable... Incorrect. Both the GBICs and IBs are hot pluggable. #The IBs are unstable and failed drives can bring down a loop... #and by-passing failed drives is an issue? Incorrect. IBs are STABLE. There are no outstanding engineering problems against IB boards. It is possible that a failed drive can hang a loop ie. broken h/w interface to drive. However, there is a procedure for by-passing failed drives, albeit, this is not automatic today. In addition, because the A5000 is using FC dual ported drives "ON MULTIPLE LOOPS". Access to the customer data is maintained even in the event of a failed/hung loop. The two loops have independent hardware interfaces to the drive thus a loop hang on one port does not imply lack of access through the data. Obviously, a mirrored configuration will further protect against the very rare possibility of multiple double loop failures. # #The Vixel hubs and GBICs are unstable and have problems initializing #loops and dealing with extended distances? Incorrect. There are NO outstanding problems with Vixel hubs or GBICs and there has not been since FCS. The hubs/GBICs work at distances up to spec.,500 meters (using 50/125 fibre), and this has not been a problem. We are in the process of qualifying new GBICs (long wave laser) as an option that will support up to 10KM (using single mode cable)... # #Microcode on the Sbus FC host adapter is unstable...many #patches expected. Issues around multi-hosting, OS rev. integration, Oracle #integration, fc arbitration and initialization are major hurdles...ie. #unresolved. The Seagate FCAL drives will require new microcode and chipsets to #match in the field late in Q198...once again "forklift" upgrade... Incorrect. Sbus host adapter Ucode has been the most stable part of the entire A5000 development. Most development integrations issues (even prior to FCS) were with host s/w drivers, host commands (luxadm), IB and disk drive f/w. The FCS A5000 product supports OS levels 2.5.1 and 2.6 ie.the OS levels that most Sun customers are using today. There has never been any Oracle integration problems to our knowledge. Multi-hosting was not supported at FCS due to mostly Seagate drive f/w issues although the Photon had fixes as well to support multi-host configs. These have been resolved recently with patches that were made available last week per the engineering plan. In this release there were several patches ie. s/w driver, disk f/w, and a Veritas patch. This was the first patch suite for the A5000 since FCS. We believe our product has the most robust RAS features in this area. This is what we gain as we are delivering the 2nd generation of the FC products. Over time we will continue to improve the RAS capability of the system, and S/W releases will be made to deliver them. Regarding the Seagate drives. We are not planning or expecting any "forklift" upgrades to FCAL drives. The FC-AL interface is new maturing technology. Seagate happens to be the only supplier of FC-AL drives at this time. Our recent patch release does include an update to the Seagate Drive's firmware to fix their Multi-initiator support. Any manufacturer using these drives would be subjected to subsequent firmware enhancements/upgrades. # #The A5000 fails to meet EMI standards in the rack configurations and #will require a "forklift" upgrade in the field. Incorrect. The A5000 was limited to 2 Photons in rack configs for the FCS product. The Max Rack config.(4 Photons) is now shipping. A simple rack kit ( FCO # FCO129) is required when additional photons are added in the field to the original 2 photon configs. This rack kit contains a screen door for the rack and a kick plate to cover an opening in the bottom of the rack. This is provided at no cost to the customer. In addition the original Photons do not have to be upgraded in any way ie. no truth to a "forklift" upgrade rumor. Photon and Photon rack configs are FULLY COMPLIANT TO CLASS A STANDARDS. The FCO will not require any interruption of service. # #Raid capability is via. Veritas host s/w...leads to low Raid-5 #performance... EMC does not implement true RAID-5, their implementation is RAID-s. The Photon uses the Veritas volume manager. Please note that host based RAID-5 on the A5000 out performs Symmetrix RAID-s Implementation in high bandwidth operations by nearly an order of magnitude, and has efficiency near 100%. For high bandwidth sequential work loads, the EMC cache is in the way. Furthermore, host based RAID provides a higher level of availability with mirrors on separate array devices, which can be physically dispersed. For small sequential writes to Raid-5, Sun offer the A3000 which out performs Symmetrix RAID-s implementation for this work load. # #No A5000 support for the Sun cluster. Incorrect. The A5000 WILL BE SUPPORTED with the next Sun Cluster release slated for April. It will require a patch upgrade for the s/w drivers and possibly drive f/w if additional problems are found beyond the fixes in the recent patch release. # #Bottom line A5000 FCAL is a step backwards... This ridicules and desperate statement shows that EMC is terrified over the leadership Sun StorEdge has in this technology. --Dule_of_Doves_503_000-- --------------3FF5D53A4C04E609EEA7077A--