将云计算用于EDA工作可能仍遥不可及,但在三到五年内,云计算将在主要EDA供应商和客户之间的设计事务中占据高达20%比例。
这一结论是由IBM创新设计副总裁在设计自动化大会为云计算定下的基调。Bernie Meyerson在其不久前的主题演讲中预测,在未来5~10年的时间里,云计算将是芯片设计人员使用的主要媒介,“这是由于IT开支不可持续的增长速度决定的。”
“一直以来,对IT资源的需求在加速且似乎永远得不到满足,由“网络事物”的出现而推动,” Meyerson表示。这在强迫即便EDA也要去探索新兴的计算模式,例如云计算。
关于这个问题的小组讨论,以SaaS(作为一种服务的软件)模式为题,来自计算机企业空间渐渐渗入设计自动化团体,以便将其设计的某些部分转移到远程服务器群。
实际上,IC设计兜了一圈:从先前的设计是采用大型主机来计算设计参数和设计出相应的芯片,而如今是大量的服务器在那里耗费公司的电费——一种老套的情节。下一步是将设计带到远程IT服务器群和他人共享服务器群,如同大型主机时代,只不过现在是在远程使用。
Synopsys公司前雇员,现为一名咨询顾问的Raul Camposano表示,云计算是由大量的、可扩展的计算资源所具有的成本效益的按需供应所推动。 “云计算为许多企业和消费者应用建立了典范。然而,要在IC设计领域获得成功的可能性有限”。
Synopsys公司营销/战略发展部高级副总裁John Chilton表示,“很显然,EDA产业必须确定如何最好地利用这一资源可能存在的优势。”
Chilton表示,在复杂性无情地增长而预算不断被审查和削减的情况下,Synopsys公司将云计算视为一种管理和控制其设计总成本的途径。“获得实时可用和弹性的计算资源是解决这一挑战的关键,和所有优化问题类似,仔细而彻底地限制成本将至关重要”。
美国加州大学伯克利分校Rean Griffith教授表示,在云计算中,某些类别的应用程序变得更加引人注目:“并行批处理作业可以受益于计算成本。例如,它将1台机器使用1,000个小时和在1个小时内同时使用1,000台机器,所花费的成本是相同的,但可以更快地得到结果。”
{pagination}
Amazon.com业务拓展经理Deepak Singh自2006年以来吹捧该公司的亚马逊网络服务(AWS)作为云计算中基础网络服务平台。 “Synopsys和其它EDA供货商正在使用AWS来提供关于其工具的培训,”Singh表示。 “有了AWS,您可以根据需要通过网络服务API迅速地进行存储和运算,只需要为你所占用的资源付费。”
Xuropa公司创始人兼首席执行官James Colgan已成功游说Synopsys和Cadence公司将Xuropa平台用于云计算。 “我们刚刚在Synopsys签署了合同,并且有更多机会。”
Colgan坚定地相信,云计算是IC设计的未来:“我们处在技术采纳推广曲线上,而转移是缓慢的,项目是受限制的。”在打造Xuropa平台时,Colgan意识到,要让IC设计过渡到云计算,预期的环境必须是“零IT开销、许可协议保持不变,并且能提供高效而安全的数据传输。”
Colgan表示,“这项业务非常吸人注目,向这项技术转移是肯定的”。
“针对网络兼容的应用程序而言,云毫无疑问取得了成功,并且一眼看去是适用于电子设计的一个完美解决方案,” Cadence公司设计服务之系统芯片设计主管Samuel George表示。
当前面临很多挑战,如可用性、性能与价格等,但 George认为,“如果公司愿意作出折衷,他们可以充分利用云。” Cadence公司在2008年推出了其用于半导体设计的SaaS产品。服务规模从一家只有几名设计人员的小公司发展到在世界各地拥有多个大型设计团队。
Altera公司设计自动化主管Paul Leventis预期将存在两个重大的挑战:一个是将Altera、第三方和代工IP放置到不在处所的计算群的安全性和法律的复杂性。 “另一个挑战是授权”。 “即使大的EDA公司提供了云计算友好的授权,但估计会要对此收费,这样我们总的EDA花费不会减少,甚至有可能会增加。
"Despite these challenges, I will be surprised if we are not leveraging the cloud within a few years," said Leventis.
Leventis表示,“尽管面临着这些挑战,如果几年之内我们没有利用和推动云,那我反倒会觉得很惊讶”。
下一页:参考原文(IC design bound for clouds, By Nicolas Mokhoff)
《电子工程专辑》网站版权所有,谢绝转载
{pagination}
IC design bound for clouds
By Nicolas Mokhoff
Cloud computing for EDA work may still be in the clouds but in three to five years it will occupy up to 20 percent of design transactions between major EDA vendors and their customers.
That conclusion was reached at the Design Automation Conference here where the tone for cloud computing was set by IBM's vice president of innovation. In his Wednesday June 16 keynote Bernie Meyerson predicted that in 5 to 10 years time it will be the dominant medium used by chip designers, "due to IT spending growing at unstainable rates."
"There has been an accelerating and seemingly insatiable need for IT resources, driven by the emergence of the ‘Internet of Things’," said Meyerson. This is forcing even the EDA world to explore emerging compute paradigms, such as cloud computing.
A panel discussion on the subject pegged the SaaS (software as a service) model from the computer enterprise space encroaching on the design automation community for relegating certain portions of their designs to remote server farms.
In effect, IC design is coming full circle from the days when designs used mainframes to calculate design parameters and lay out chips accordingly, to now where large amount of servers eat up individual companies electricty bills—an unsustainble scenario. The next step is to bring the designs out to remote IT server farms and share server farms with others, as in the mainframe era, but now used remotely.
Raul Camposano, formerly of Synopsys and now a consultant, said cloud computing is driven by the cost-effective on-demand availability of large, scalable amounts of computing resources. "The cloud has become an established paradigm for many enterprise and consumer applications. However, in IC design its success is still limited."
John Chilton, senior vice president Marketing/Strategic Development at Synopsys, said "it is clear that the EDA industry must determine how to take best advantage of the benefits that this resource may present."
Chilton said that Synopsys views cloud computing as a way to manage and control their total cost of design, in an environment where complexity is growing relentlessly and budgets are being scrutinized and reduced. "Access to instantly and elastically available compute resources is key to solving this challenge, and like all optimization problems, carefully and thoroughly defining the cost functions will be crucial," said Chilton.
Rean Griffith, professor at University at California, Berkeley, said certain classes of applications become more compelling in the cloud: "Parallel batch processing jobs can benefit from the computing cost. For instance, it costs the same to use 1 machine for 1000 hours as it does to use 1000 machines for 1 hour, and the result can be obtained much faster."
Deepak Singh, business development manager at Amazon.com touted the company's Amazon Web Services (AWS) as an infrastructure web services platform in the cloud since 2006. "Synopsys and other EDA vendors are using AWS for providing training on its tools," said Singh. "With AWS you can rapidly provision storage and computing on-demand via web services APIs, paying only for the resources you consume."
James Colgan, founder and CEO of Xuropa, Inc. has enticed both Synopsys and Cadence to use the Xuropa plaform for cloud computing. "We just signed on Synopsys and have a couple more prospects."
Colgan strongly believes that cloud computing is the future of IC design: "We are on a technology adoption curve and migration is slow and projects are constrained." In building out the Xuropa platform Colgan realized that for IC design to transition to cloud computing the expected environment had to have "zero IT overhead, have license agreements remain the same and provide efficient and secure data transfer."
"The business case is compelling and a migration to this technology is inevitable", said Colgin.
"The success of the cloud is unquestionable for Web-compatible applications and at first glance seems like the perfect solution to apply to electronic design," said Samuel George, director of system-chip design for Cadence's Design Services.
There are plenty of challenges such as availability, performace and pricing but George belives that "if companies are willing to make tradeoffs, they can take advantage of the cloud." Cadence launched its own SaaS offering in 2008 for semiconductor design. The service scales from a few designers in a single location to multiple large design teams distributed globally.
Paul Leventis, director, Design Automation at Altera, sees two major challenges. One is security and legal complexity of placing Altera, 3rd party and foundry IP in compute farms not located on premises. "Another challenge is licensing", said Leventis. "Even if the big EDA companies provided cloud-computing friendly licensing, these would presumably be priced such that our total EDA spend would not be reduced, or might even increase.
"Despite these challenges, I will be surprised if we are not leveraging the cloud within a few years," said Leventis.